*** EinstCrazy has joined #openstack-nova | 00:01 | |
*** akshai has quit IRC | 00:01 | |
*** gjayavelu has quit IRC | 00:01 | |
*** markvoelker has quit IRC | 00:01 | |
*** mylu has quit IRC | 00:02 | |
*** mylu has joined #openstack-nova | 00:02 | |
*** ijw has quit IRC | 00:02 | |
*** xyang1 has quit IRC | 00:03 | |
*** signed8bit is now known as signed8bit_ZZZzz | 00:03 | |
*** Piet has quit IRC | 00:03 | |
*** bapalm has quit IRC | 00:03 | |
*** russellb has quit IRC | 00:03 | |
*** bapalm has joined #openstack-nova | 00:04 | |
*** russellb has joined #openstack-nova | 00:04 | |
*** mylu_ has joined #openstack-nova | 00:04 | |
*** gjayavelu has joined #openstack-nova | 00:05 | |
*** mylu has quit IRC | 00:05 | |
*** amotoki has joined #openstack-nova | 00:06 | |
*** Piet has joined #openstack-nova | 00:06 | |
*** amotoki has quit IRC | 00:06 | |
*** baoli has joined #openstack-nova | 00:07 | |
*** n0ano has joined #openstack-nova | 00:08 | |
*** penick has joined #openstack-nova | 00:10 | |
*** ssurana has quit IRC | 00:11 | |
*** mc_nair has quit IRC | 00:11 | |
*** ctrath has quit IRC | 00:12 | |
*** EinstCrazy has quit IRC | 00:12 | |
openstackgerrit | Dan Smith proposed openstack/nova: Make libvirt driver return migrate data objects for source and dest checks https://review.openstack.org/247720 | 00:14 |
---|---|---|
openstackgerrit | Dan Smith proposed openstack/nova: Add transitional support for migrate data objects to compute manager https://review.openstack.org/247719 | 00:14 |
openstackgerrit | Dan Smith proposed openstack/nova: Add xenapi support for XenapiLiveMigrateData objects https://review.openstack.org/247853 | 00:14 |
*** zhangjn has quit IRC | 00:15 | |
*** zenoway has joined #openstack-nova | 00:15 | |
*** aginwala has quit IRC | 00:17 | |
*** zenoway has quit IRC | 00:20 | |
*** mylu_ has quit IRC | 00:20 | |
*** shaohe_feng has joined #openstack-nova | 00:21 | |
*** mylu has joined #openstack-nova | 00:21 | |
*** pratikmallya has quit IRC | 00:21 | |
*** oomichi has joined #openstack-nova | 00:21 | |
*** mylu_ has joined #openstack-nova | 00:24 | |
*** rook has joined #openstack-nova | 00:25 | |
*** mylu has quit IRC | 00:25 | |
*** edtubill has joined #openstack-nova | 00:26 | |
*** annegentle has quit IRC | 00:27 | |
*** markus_z has joined #openstack-nova | 00:29 | |
*** shaohe_feng has quit IRC | 00:29 | |
*** otter768 has joined #openstack-nova | 00:30 | |
*** electrocucaracha has quit IRC | 00:31 | |
*** jamielennox|away is now known as jamielennox | 00:32 | |
*** loquacities has quit IRC | 00:33 | |
*** loquacities has joined #openstack-nova | 00:33 | |
*** markus_z has quit IRC | 00:34 | |
*** otter768 has quit IRC | 00:35 | |
*** edtubill has quit IRC | 00:36 | |
*** zenoway has joined #openstack-nova | 00:39 | |
*** electrocucaracha has joined #openstack-nova | 00:40 | |
*** mylu_ has quit IRC | 00:40 | |
*** zenoway has quit IRC | 00:44 | |
*** penick has quit IRC | 00:51 | |
*** rk4n has quit IRC | 00:53 | |
*** apuimedo has joined #openstack-nova | 00:54 | |
*** salv-orl_ has joined #openstack-nova | 00:55 | |
*** apuimedo has quit IRC | 00:55 | |
*** salv-orlando has quit IRC | 00:58 | |
*** apoorvad has quit IRC | 00:58 | |
*** aginwala has joined #openstack-nova | 00:59 | |
*** jaypipes has quit IRC | 00:59 | |
*** rook has quit IRC | 00:59 | |
*** armax_ has joined #openstack-nova | 01:00 | |
*** RichardRaseley has quit IRC | 01:01 | |
*** EinstCrazy has joined #openstack-nova | 01:01 | |
*** gjayavelu has quit IRC | 01:02 | |
*** armax has quit IRC | 01:02 | |
*** zhenq1 has quit IRC | 01:02 | |
*** armax_ is now known as armax | 01:02 | |
*** zenoway has joined #openstack-nova | 01:03 | |
*** ccard__ has quit IRC | 01:04 | |
*** zhangjn has joined #openstack-nova | 01:05 | |
*** ssurana has joined #openstack-nova | 01:07 | |
*** zenoway has quit IRC | 01:07 | |
*** zhenguo has joined #openstack-nova | 01:08 | |
*** jinxing has joined #openstack-nova | 01:08 | |
*** mylu has joined #openstack-nova | 01:09 | |
*** flip214 has quit IRC | 01:09 | |
*** ssurana has quit IRC | 01:10 | |
*** flip214 has joined #openstack-nova | 01:11 | |
*** flip214 has joined #openstack-nova | 01:11 | |
*** markvoelker has joined #openstack-nova | 01:17 | |
*** thorst has joined #openstack-nova | 01:17 | |
*** thorst has quit IRC | 01:20 | |
*** thorst has joined #openstack-nova | 01:20 | |
*** pixelbeat has quit IRC | 01:22 | |
*** markvoelker has quit IRC | 01:22 | |
*** pixelbeat has joined #openstack-nova | 01:22 | |
*** tonyb has quit IRC | 01:23 | |
*** tonyb has joined #openstack-nova | 01:23 | |
*** tonyb has quit IRC | 01:24 | |
*** thorst has quit IRC | 01:24 | |
*** tonyb has joined #openstack-nova | 01:24 | |
*** tonyb has quit IRC | 01:26 | |
*** tonyb has joined #openstack-nova | 01:26 | |
*** ijw has joined #openstack-nova | 01:26 | |
*** rook has joined #openstack-nova | 01:30 | |
*** shaohe_feng has joined #openstack-nova | 01:30 | |
*** jinxing has quit IRC | 01:30 | |
*** electrocucaracha has quit IRC | 01:31 | |
*** tonyb has quit IRC | 01:32 | |
*** tonyb has joined #openstack-nova | 01:33 | |
*** tonyb has quit IRC | 01:34 | |
*** tonyb has joined #openstack-nova | 01:34 | |
openstackgerrit | Zhang Ni proposed openstack/nova-specs: Add local-to-instance to create server api https://review.openstack.org/241066 | 01:35 |
*** ccard_ has joined #openstack-nova | 01:36 | |
*** chhavi has joined #openstack-nova | 01:37 | |
*** Daisy_ has joined #openstack-nova | 01:37 | |
*** mylu has quit IRC | 01:38 | |
*** zenoway has joined #openstack-nova | 01:39 | |
*** mriedem has joined #openstack-nova | 01:41 | |
*** daemontool_ has quit IRC | 01:41 | |
*** daemontool_ has joined #openstack-nova | 01:42 | |
*** exploreshaifali has quit IRC | 01:42 | |
*** Daisy_ has quit IRC | 01:43 | |
*** zenoway has quit IRC | 01:43 | |
*** mylu has joined #openstack-nova | 01:44 | |
*** suro-patz has quit IRC | 01:49 | |
*** chenzeng has joined #openstack-nova | 01:50 | |
*** markvoelker has joined #openstack-nova | 01:50 | |
*** pratikmallya has joined #openstack-nova | 01:51 | |
chenzeng | hello, every one! I have a question: suppose I create a volume by cinder, can I mount this volume to the host machine, and how? Thanks. | 01:55 |
*** tpatil_ has quit IRC | 01:55 | |
chenzeng | Looking forward to the answer,Thanks. | 01:56 |
*** ccard_ has quit IRC | 01:56 | |
*** aginwala has quit IRC | 01:58 | |
*** apoorvad has joined #openstack-nova | 01:58 | |
*** ccard_ has joined #openstack-nova | 01:59 | |
mriedem | http://docs.openstack.org/cli-reference/content/novaclient_commands.html#novaclient_subcommand_volume-attach | 01:59 |
mriedem | docs ftw | 01:59 |
mriedem | oh host machine | 02:00 |
*** ljxiash has joined #openstack-nova | 02:03 | |
*** apoorvad has quit IRC | 02:03 | |
*** zenoway has joined #openstack-nova | 02:05 | |
*** rfolco_ has joined #openstack-nova | 02:07 | |
*** aginwala has joined #openstack-nova | 02:08 | |
*** zenoway has quit IRC | 02:09 | |
*** aginwala has quit IRC | 02:11 | |
*** mylu has quit IRC | 02:12 | |
*** jwcroppe has quit IRC | 02:12 | |
*** otter768 has joined #openstack-nova | 02:14 | |
*** ijw has quit IRC | 02:14 | |
*** jerrygb has quit IRC | 02:14 | |
*** rfolco_ has quit IRC | 02:14 | |
*** aginwala has joined #openstack-nova | 02:15 | |
*** apoorvad has joined #openstack-nova | 02:16 | |
chenzeng | mriedem:thanks very much! | 02:16 |
*** unicell1 has quit IRC | 02:21 | |
*** jerrygb has joined #openstack-nova | 02:24 | |
*** vilobhmm has quit IRC | 02:27 | |
*** vilobhmm has joined #openstack-nova | 02:28 | |
*** vilobhmm has quit IRC | 02:29 | |
*** apoorvad has quit IRC | 02:30 | |
*** mylu has joined #openstack-nova | 02:31 | |
mriedem | jroll: i will apologize ahead of time https://review.openstack.org/#/c/237067/ | 02:31 |
*** mylu has quit IRC | 02:32 | |
*** haomaiwang has joined #openstack-nova | 02:32 | |
*** pratikmallya has quit IRC | 02:34 | |
*** pixelbeat has quit IRC | 02:35 | |
*** mylu has joined #openstack-nova | 02:35 | |
*** nic has quit IRC | 02:36 | |
*** nithyag_ has joined #openstack-nova | 02:36 | |
*** haomaiwang has quit IRC | 02:37 | |
*** nithyag__ has quit IRC | 02:37 | |
*** mylu has quit IRC | 02:38 | |
*** haomaiwang has joined #openstack-nova | 02:38 | |
*** mylu has joined #openstack-nova | 02:38 | |
*** dims_ has quit IRC | 02:39 | |
*** mylu has quit IRC | 02:43 | |
*** mylu has joined #openstack-nova | 02:44 | |
*** aginwala has quit IRC | 02:45 | |
*** klkumar has joined #openstack-nova | 02:45 | |
*** pratikmallya has joined #openstack-nova | 02:48 | |
*** aginwala has joined #openstack-nova | 02:48 | |
*** aginwala has quit IRC | 02:48 | |
*** Kevin_Zheng has joined #openstack-nova | 02:49 | |
*** markus_z has joined #openstack-nova | 02:51 | |
*** armax has quit IRC | 02:56 | |
*** sacharya has joined #openstack-nova | 02:57 | |
*** c00281451 has joined #openstack-nova | 03:00 | |
*** otter768 has quit IRC | 03:01 | |
*** tonyb has quit IRC | 03:01 | |
*** tonyb has joined #openstack-nova | 03:01 | |
*** yamahata has quit IRC | 03:01 | |
*** venkat_p has joined #openstack-nova | 03:01 | |
*** rushiagr_away is now known as rushiagr | 03:02 | |
*** chenzeng has quit IRC | 03:02 | |
*** jdurgin1 has quit IRC | 03:03 | |
*** fawadkhaliq has joined #openstack-nova | 03:08 | |
*** markus_z has quit IRC | 03:14 | |
openstackgerrit | Zhenyu Zheng proposed openstack/nova-specs: Add pagination and timestamp filtering support for os-migrations API https://review.openstack.org/239869 | 03:14 |
*** pratikmallya has quit IRC | 03:19 | |
openstackgerrit | Bo Chi proposed openstack/nova: Fix some usage for exception https://review.openstack.org/247898 | 03:19 |
*** edmondsw has quit IRC | 03:19 | |
*** dims has joined #openstack-nova | 03:21 | |
*** mylu_ has joined #openstack-nova | 03:22 | |
*** mylu has quit IRC | 03:22 | |
openstackgerrit | Zhenyu Zheng proposed openstack/nova-specs: Add pagination and changes since filter support for os-isntance-action API https://review.openstack.org/240401 | 03:23 |
*** hparekh2 has quit IRC | 03:23 | |
*** takashin has joined #openstack-nova | 03:27 | |
*** gcb has quit IRC | 03:35 | |
*** gcb has joined #openstack-nova | 03:35 | |
*** otter768 has joined #openstack-nova | 03:36 | |
*** sacharya has quit IRC | 03:37 | |
*** sacharya has joined #openstack-nova | 03:37 | |
*** sacharya has quit IRC | 03:38 | |
*** hparekh has joined #openstack-nova | 03:38 | |
*** baoli has quit IRC | 03:40 | |
openstackgerrit | Jianghua Wang proposed openstack/nova: xenapi: OVS agent updates the wrong port when using XenServer + Neutron https://review.openstack.org/242846 | 03:42 |
*** vilobhmm has joined #openstack-nova | 03:42 | |
*** chhavi has quit IRC | 03:44 | |
*** vilobhmm1 has joined #openstack-nova | 03:47 | |
*** zhenq has joined #openstack-nova | 03:48 | |
*** vilobhmm has quit IRC | 03:49 | |
*** sacharya has joined #openstack-nova | 03:50 | |
*** thangp has joined #openstack-nova | 03:51 | |
*** thangp has quit IRC | 03:53 | |
*** mriedem has quit IRC | 03:59 | |
*** gyee has quit IRC | 04:00 | |
*** fawadkhaliq has quit IRC | 04:01 | |
*** zenoway has joined #openstack-nova | 04:04 | |
*** vilobhmm1 has quit IRC | 04:06 | |
*** jhesketh has quit IRC | 04:06 | |
*** dave-mccowan has quit IRC | 04:07 | |
*** cfriesen__ has quit IRC | 04:08 | |
*** jeblair has quit IRC | 04:08 | |
*** suro-patz has joined #openstack-nova | 04:08 | |
*** cfriesen__ has joined #openstack-nova | 04:08 | |
*** zenoway has quit IRC | 04:08 | |
*** josh6627 has joined #openstack-nova | 04:08 | |
*** EmilienM has quit IRC | 04:08 | |
*** jeblair has joined #openstack-nova | 04:09 | |
*** EmilienM has joined #openstack-nova | 04:10 | |
*** Sree has joined #openstack-nova | 04:10 | |
*** Sree has quit IRC | 04:11 | |
*** Sree has joined #openstack-nova | 04:11 | |
*** venkat_p has left #openstack-nova | 04:13 | |
*** zhangjn has quit IRC | 04:14 | |
*** otter768 has quit IRC | 04:17 | |
*** zhenq has quit IRC | 04:20 | |
*** mc_nair has joined #openstack-nova | 04:23 | |
*** achanda has joined #openstack-nova | 04:23 | |
*** rajesht has quit IRC | 04:27 | |
*** gjayavelu has joined #openstack-nova | 04:28 | |
*** zhangjn has joined #openstack-nova | 04:29 | |
*** yamahata has joined #openstack-nova | 04:32 | |
*** dims has quit IRC | 04:36 | |
*** armax has joined #openstack-nova | 04:37 | |
*** zenoway has joined #openstack-nova | 04:39 | |
*** signed8bit_ZZZzz is now known as signed8bit | 04:41 | |
*** signed8bit has quit IRC | 04:42 | |
*** vilobhmm has joined #openstack-nova | 04:42 | |
*** zenoway has quit IRC | 04:43 | |
*** jinxing has joined #openstack-nova | 04:44 | |
*** rushiagr is now known as rushiagr_away | 04:44 | |
*** richm has quit IRC | 04:45 | |
*** zhangjn has quit IRC | 04:45 | |
*** achanda_ has joined #openstack-nova | 04:46 | |
*** mylu_ has quit IRC | 04:47 | |
*** jinxing has quit IRC | 04:48 | |
*** suro-patz has quit IRC | 04:49 | |
*** achanda has quit IRC | 04:50 | |
*** kutyamutya has quit IRC | 04:50 | |
*** suro-patz has joined #openstack-nova | 04:51 | |
*** suro-patz1 has joined #openstack-nova | 04:52 | |
*** suro-patz has quit IRC | 04:52 | |
*** daemontool_ has quit IRC | 04:58 | |
*** daemontool_ has joined #openstack-nova | 04:58 | |
*** gjayavelu has quit IRC | 04:58 | |
*** fawadkhaliq has joined #openstack-nova | 04:59 | |
*** gjayavelu has joined #openstack-nova | 04:59 | |
*** vilobhmm has quit IRC | 05:00 | |
*** fawadkhaliq has quit IRC | 05:00 | |
*** fawadk has joined #openstack-nova | 05:00 | |
*** armax has quit IRC | 05:00 | |
*** oomichi_ has joined #openstack-nova | 05:01 | |
*** oomichi has quit IRC | 05:03 | |
*** gjayavelu has quit IRC | 05:04 | |
*** RuiChen has quit IRC | 05:05 | |
*** RuiChen has joined #openstack-nova | 05:06 | |
*** zhangjn has joined #openstack-nova | 05:11 | |
*** zhangjn has quit IRC | 05:12 | |
*** zhangjn has joined #openstack-nova | 05:18 | |
*** mc_nair has quit IRC | 05:21 | |
*** achanda_ has quit IRC | 05:21 | |
*** sacharya has quit IRC | 05:23 | |
*** mylu has joined #openstack-nova | 05:24 | |
*** jinxing has joined #openstack-nova | 05:25 | |
*** mylu has quit IRC | 05:25 | |
*** mylu has joined #openstack-nova | 05:26 | |
*** mylu has quit IRC | 05:27 | |
*** mylu has joined #openstack-nova | 05:28 | |
*** jerrygb has quit IRC | 05:29 | |
*** sacharya has joined #openstack-nova | 05:29 | |
openstackgerrit | MD NADEEM proposed openstack/nova: Updating nova config-reference doc https://review.openstack.org/239321 | 05:29 |
*** sacharya has quit IRC | 05:30 | |
*** rushiagr_away is now known as rushiagr | 05:32 | |
*** stevemar_ has quit IRC | 05:34 | |
*** stevemar_ has joined #openstack-nova | 05:35 | |
*** armax has joined #openstack-nova | 05:35 | |
*** stevemar_ has quit IRC | 05:38 | |
*** mylu has quit IRC | 05:39 | |
*** suro-patz1 has quit IRC | 05:40 | |
*** ildikov has quit IRC | 05:43 | |
*** haomaiwang has quit IRC | 05:46 | |
*** suro-patz has joined #openstack-nova | 05:47 | |
*** mylu has joined #openstack-nova | 05:57 | |
*** CORTEX\krfa has quit IRC | 05:57 | |
*** haomaiwang has joined #openstack-nova | 05:59 | |
*** unicell has joined #openstack-nova | 06:01 | |
*** vilobhmm has joined #openstack-nova | 06:01 | |
*** sudipto has joined #openstack-nova | 06:03 | |
*** oomichi_ has quit IRC | 06:04 | |
openstackgerrit | OpenStack Proposal Bot proposed openstack/nova: Imported Translations from Zanata https://review.openstack.org/242727 | 06:04 |
*** aginwala has joined #openstack-nova | 06:07 | |
*** suro-patz has quit IRC | 06:07 | |
*** amotoki_ has joined #openstack-nova | 06:07 | |
*** jkraj has joined #openstack-nova | 06:08 | |
*** yonglihe has quit IRC | 06:09 | |
*** lpetrut has joined #openstack-nova | 06:10 | |
*** mylu has quit IRC | 06:12 | |
*** mylu has joined #openstack-nova | 06:12 | |
*** haomaiwang has quit IRC | 06:14 | |
*** suro-patz has joined #openstack-nova | 06:14 | |
*** oomichi has joined #openstack-nova | 06:15 | |
*** amotoki_ has quit IRC | 06:17 | |
*** otter768 has joined #openstack-nova | 06:18 | |
*** mylu has quit IRC | 06:18 | |
*** haomaiwa_ has joined #openstack-nova | 06:18 | |
*** penick has joined #openstack-nova | 06:23 | |
*** otter768 has quit IRC | 06:23 | |
*** eliqiao_ has joined #openstack-nova | 06:26 | |
*** achanda has joined #openstack-nova | 06:26 | |
*** suro-patz has quit IRC | 06:26 | |
*** eliqiao has quit IRC | 06:28 | |
*** rcernin has joined #openstack-nova | 06:30 | |
*** jerrygb has joined #openstack-nova | 06:30 | |
*** sacharya has joined #openstack-nova | 06:31 | |
*** jerrygb has quit IRC | 06:36 | |
*** sacharya has quit IRC | 06:36 | |
*** abhishekk has joined #openstack-nova | 06:42 | |
*** achanda has quit IRC | 06:42 | |
*** achanda has joined #openstack-nova | 06:43 | |
*** haomaiwa_ has quit IRC | 06:43 | |
*** penick has quit IRC | 06:43 | |
*** unicell has quit IRC | 06:44 | |
*** unicell has joined #openstack-nova | 06:45 | |
*** achanda has quit IRC | 06:48 | |
*** rushiagr is now known as rushiagr_away | 06:52 | |
*** haomaiwang has joined #openstack-nova | 06:53 | |
openstackgerrit | lyanchih proposed openstack/python-novaclient: Help msg about libvirt always default device names https://review.openstack.org/247951 | 06:54 |
*** salv-orlando has joined #openstack-nova | 06:54 | |
*** mjura has joined #openstack-nova | 06:57 | |
*** jinxing has quit IRC | 06:58 | |
*** salv-orl_ has quit IRC | 06:58 | |
*** jinxing has joined #openstack-nova | 06:59 | |
*** haomaiwang has quit IRC | 07:01 | |
*** haomaiwang has joined #openstack-nova | 07:02 | |
*** eliqiao has joined #openstack-nova | 07:09 | |
*** eliqiao_ has quit IRC | 07:10 | |
*** ildikov has joined #openstack-nova | 07:13 | |
*** rubasov has quit IRC | 07:14 | |
*** cfriesen__ has quit IRC | 07:15 | |
*** eliqiao has quit IRC | 07:18 | |
*** eliqiao has joined #openstack-nova | 07:20 | |
*** Kevin_Zheng has quit IRC | 07:22 | |
*** nkrinner has joined #openstack-nova | 07:23 | |
*** haomaiwang has quit IRC | 07:25 | |
*** haomaiwang has joined #openstack-nova | 07:26 | |
*** achanda has joined #openstack-nova | 07:29 | |
*** AJaeger has joined #openstack-nova | 07:31 | |
AJaeger | nova stable cores, could you give the import of translations for liberty another try, please? This change was approved several times but then failed to merge due to other problems - and it's an important metadata update: https://review.openstack.org/#/c/237411/ | 07:32 |
AJaeger | johnthetubaguy: can you help, please? ^ | 07:32 |
*** armax has quit IRC | 07:34 | |
*** sahid has joined #openstack-nova | 07:34 | |
*** ankit_ag has joined #openstack-nova | 07:37 | |
*** vilobhmm has quit IRC | 07:39 | |
*** alexschm has joined #openstack-nova | 07:40 | |
*** achanda has quit IRC | 07:41 | |
*** zhangjn has quit IRC | 07:47 | |
*** mpavone has joined #openstack-nova | 07:48 | |
garyk1 | johnthetubaguy: dhellmann: please see https://review.openstack.org/247705 for release notes for m-1 | 07:53 |
AJaeger | garyk1: could you check 237411, please? See above | 07:54 |
*** moshele has joined #openstack-nova | 07:55 | |
AJaeger | garyk1: looking at releasenotes, this looks ugly - http://docs-draft.openstack.org/05/247705/1/check/gate-nova-releasenotes/a23f3bf//releasenotes/build/html/unreleased.html | 07:55 |
AJaeger | garyk1: indeed, the release note is broken. Let me have a look quickly... | 07:55 |
*** takashin has left #openstack-nova | 07:56 | |
AJaeger | garyk1: add a bullet point before it! | 07:56 |
*** ttx has quit IRC | 07:57 | |
*** ttx has joined #openstack-nova | 07:57 | |
*** pratikmallya has joined #openstack-nova | 07:58 | |
*** haomaiwang has quit IRC | 07:59 | |
*** haomaiwang has joined #openstack-nova | 08:02 | |
garyk1 | AJaeger: thanks! | 08:05 |
*** ildikov has quit IRC | 08:06 | |
openstackgerrit | Oleg Bondarev proposed openstack/nova: Live migration: wait for vif-plugged event on pre live migration https://review.openstack.org/246910 | 08:07 |
openstackgerrit | garyk proposed openstack/nova: Add relnote for change in default setting https://review.openstack.org/247705 | 08:10 |
AJaeger | garyk1: did you see my comment above for translations and can help getting them in, please? https://review.openstack.org/#/c/237411/ | 08:10 |
*** markus_z has joined #openstack-nova | 08:11 | |
garyk1 | AJaeger: looking now. sorry i missed it before | 08:11 |
AJaeger | garyk1: thanks! | 08:11 |
garyk1 | AJaeger: done! | 08:11 |
AJaeger | garyk1: thanks! | 08:12 |
AJaeger | now waiting for an approval ;) | 08:12 |
*** rotbeard has joined #openstack-nova | 08:15 | |
*** moshele has quit IRC | 08:16 | |
*** achanda has joined #openstack-nova | 08:17 | |
garyk1 | AJaeger: its my first time doingt he reno stuff. so it make take a few iterations :) | 08:17 |
*** rubasov has joined #openstack-nova | 08:17 | |
*** zhangjn has joined #openstack-nova | 08:18 | |
*** otter768 has joined #openstack-nova | 08:18 | |
*** matrohon has joined #openstack-nova | 08:19 | |
AJaeger | garyk1: we're all learning with it. We should tell dhellmann that you fall into a trap and figure out a way to make it more robust | 08:20 |
*** paul-carlton1 has joined #openstack-nova | 08:20 | |
*** fawadk has quit IRC | 08:21 | |
openstackgerrit | garyk proposed openstack/nova: Add relnote for change in default setting https://review.openstack.org/247705 | 08:21 |
*** fawadkhaliq has joined #openstack-nova | 08:22 | |
*** otter768 has quit IRC | 08:23 | |
*** rushiagr_away is now known as rushiagr | 08:26 | |
garyk1 | AJaeger: :) | 08:26 |
*** oomichi has quit IRC | 08:27 | |
*** Savemech has quit IRC | 08:27 | |
*** markus_z has quit IRC | 08:28 | |
*** Savemech has joined #openstack-nova | 08:31 | |
openstackgerrit | Oleg Bondarev proposed openstack/nova: Live migration: wait for vif-plugged event on pre live migration https://review.openstack.org/246910 | 08:31 |
*** zhangjn has quit IRC | 08:32 | |
*** lpetrut has quit IRC | 08:35 | |
*** daemontool_ has quit IRC | 08:36 | |
*** amotoki_ has joined #openstack-nova | 08:36 | |
*** zhangjn has joined #openstack-nova | 08:36 | |
*** pratikmallya has quit IRC | 08:38 | |
*** daemontool_ has joined #openstack-nova | 08:39 | |
*** jichen has joined #openstack-nova | 08:41 | |
openstackgerrit | sahid proposed openstack/nova: libvirt: make detach_device to return an async object https://review.openstack.org/217680 | 08:42 |
*** gcb has quit IRC | 08:43 | |
*** jlanoux has joined #openstack-nova | 08:43 | |
*** gcb has joined #openstack-nova | 08:44 | |
bauzas | good morning Nova | 08:44 |
tonyb | bauzas: Good morning to you! | 08:46 |
openstackgerrit | sahid proposed openstack/nova: libvirt: add policy check to avoid using qga in realtime context https://review.openstack.org/247585 | 08:47 |
openstackgerrit | sahid proposed openstack/nova: libvirt: add realtime support https://review.openstack.org/197569 | 08:47 |
* tonyb goes to get a beer before trying to write his first spec ... what could *possibly* go wrong ;P | 08:47 | |
*** amotoki_ has quit IRC | 08:47 | |
*** oomichi has joined #openstack-nova | 08:48 | |
*** ihrachys has joined #openstack-nova | 08:50 | |
*** zhangjn has quit IRC | 08:51 | |
*** rk4n has joined #openstack-nova | 08:52 | |
*** ihrachys has quit IRC | 08:53 | |
*** fawadkhaliq has quit IRC | 08:53 | |
*** fawadkhaliq has joined #openstack-nova | 08:53 | |
*** rotbeard has quit IRC | 08:53 | |
*** markus_z has joined #openstack-nova | 08:54 | |
markus_z | garyk1: Could you check my question to your feedback in https://review.openstack.org/#/c/242064/ please? | 08:55 |
*** aginwala has quit IRC | 08:56 | |
*** tdurakov is now known as Guest14657 | 08:56 | |
markus_z | johnthetubaguy: Do you think it makes sense to split out the creation of the "nova/conf" package here: https://review.openstack.org/#/c/244177/ ? | 08:56 |
*** Guest14657 has joined #openstack-nova | 08:57 | |
markus_z | johnthetubaguy: If that would then merge, I could create other patch sets which only depend on that. | 08:57 |
*** paul-carlton1 has quit IRC | 08:57 | |
*** rk4n has quit IRC | 08:58 | |
*** nkrinner has quit IRC | 08:58 | |
*** yamahata has quit IRC | 08:58 | |
bauzas | tonyb: oh man, I also have to write a spec, thanks for the reminder ! | 08:58 |
*** nkrinner has joined #openstack-nova | 08:58 | |
tonyb | bauzas: your welcome. That was probably the most productive thing I've done today :D | 08:59 |
bauzas | that, and mentioning a beer | 09:00 |
*** ihrachys has joined #openstack-nova | 09:00 | |
*** tdurakov has joined #openstack-nova | 09:00 | |
tdurakov | paul-carlton, hi, are you around? | 09:00 |
bauzas | but drinking a beer at 10:00am doesn't sound really social, unfortunately | 09:00 |
*** haomaiwang has quit IRC | 09:01 | |
*** matrohon has quit IRC | 09:01 | |
markus_z | bauzas: We have a thing here called "morning pint". It's basically wine for breakfast and socially acceptable :) | 09:01 |
*** haomaiwang has joined #openstack-nova | 09:01 | |
* bauzas considers relocating | 09:01 | |
bauzas | oh, my manners | 09:02 |
*** bauzas is now known as bauwser | 09:02 | |
*** ndipanov has joined #openstack-nova | 09:06 | |
*** rotbeard has joined #openstack-nova | 09:07 | |
*** nkrinner has quit IRC | 09:10 | |
*** matrohon has joined #openstack-nova | 09:13 | |
*** nkrinner has joined #openstack-nova | 09:14 | |
*** achanda has quit IRC | 09:17 | |
*** yassine__ has joined #openstack-nova | 09:18 | |
*** chhavi has joined #openstack-nova | 09:19 | |
*** yassine__ has quit IRC | 09:19 | |
*** yassine__ has joined #openstack-nova | 09:19 | |
*** jnclyz12483 has joined #openstack-nova | 09:23 | |
*** danpb has joined #openstack-nova | 09:25 | |
*** achanda has joined #openstack-nova | 09:25 | |
*** haomaiwang has quit IRC | 09:25 | |
*** ihrachys_ has joined #openstack-nova | 09:25 | |
*** ihrachys has quit IRC | 09:25 | |
*** jinxing has quit IRC | 09:25 | |
*** haomaiwang has joined #openstack-nova | 09:27 | |
*** jnclyz12483 has quit IRC | 09:28 | |
*** ihrachys_ is now known as ihrachys | 09:28 | |
*** jinxing has joined #openstack-nova | 09:29 | |
*** markvoelker has quit IRC | 09:30 | |
*** e0ne has joined #openstack-nova | 09:31 | |
*** zhangjn has joined #openstack-nova | 09:31 | |
*** pkoniszewski has quit IRC | 09:35 | |
openstackgerrit | Timofey Durakov proposed openstack/nova: DO NOT MERGE NFS setup for multinode job https://review.openstack.org/247081 | 09:35 |
*** zenoway has joined #openstack-nova | 09:35 | |
*** haomaiwang has quit IRC | 09:37 | |
*** jinxing has quit IRC | 09:37 | |
*** achanda has quit IRC | 09:37 | |
*** pkoniszewski has joined #openstack-nova | 09:41 | |
johnthetubaguy | markus_z: I think thats fine as a single patch, but I could be missing something? | 09:42 |
*** haomaiwang has joined #openstack-nova | 09:42 | |
*** zhangjn has quit IRC | 09:44 | |
AJaeger | johnthetubaguy: could I bother you again for getting Nova stable translations in, please? https://review.openstack.org/#/c/237411/ That's an important metadata change... | 09:44 |
*** jistr has joined #openstack-nova | 09:44 | |
openstackgerrit | Tang Chen proposed openstack/nova-specs: Support soft reboot and poweroff in nova ironic driver. https://review.openstack.org/229282 | 09:45 |
johnthetubaguy | AJaeger: hmm, I didn't know it worked like that, I thought we would be translating trunk and backporting, but I guess that doesn't work? or that happens as well? | 09:46 |
AJaeger | johnthetubaguy: this is new since liberty release, the translators are translating on a branch since you make changes to master for strings... | 09:47 |
AJaeger | so, no backports here | 09:47 |
AJaeger | Daisy,i18n PTL, wanted to reach out to all stable teams, seems that didn't happen properly ;( | 09:47 |
openstackgerrit | John Garbutt proposed openstack/nova: config options: enhance help text of section "serial_console" https://review.openstack.org/246465 | 09:48 |
*** EinstCrazy has quit IRC | 09:49 | |
AJaeger | johnthetubaguy: the translators - and our translation server - take care of syncing translated strings between releases. | 09:49 |
johnthetubaguy | AJaeger: so I may have been told, and just forgotten, thanks for the update on that. | 09:50 |
*** amotoki_ has joined #openstack-nova | 09:50 | |
abhishekk | johnthetubaguy: hi, I have replied to your comments on nova-specs, https://review.openstack.org/135387 I will be grateful if you could have a look at it, Thank you in advance. | 09:51 |
*** Sree has quit IRC | 09:53 | |
johnthetubaguy | abhishekk: I really want alaski to take a look at that ASAP, I will try ping him again to see if he has chance to take another look a tthat | 09:53 |
AJaeger | johnthetubaguy: feel free to reach out to Daisy or the i18n mailing list and discuss further. | 09:53 |
abhishekk | johnthetubaguy: just got to know that non-priority spec freeze is approaching (3 December) and I will be happy if this gets approved for Mitaka :) | 09:54 |
AJaeger | This metadata change was important for me - getting rid of old URL for translation server | 09:54 |
abhishekk | johnthetubaguy: thank you | 09:54 |
*** alex_klimov has joined #openstack-nova | 09:55 | |
johnthetubaguy | AJaeger: so we have finish changing strings for you folks to have a chance to translate it, and that that justifies the approach, yeah, I should catch up with Daisy | 09:55 |
*** c00281451 has quit IRC | 10:00 | |
*** zhenguo has quit IRC | 10:01 | |
*** lpetrut has joined #openstack-nova | 10:01 | |
*** ljxiash has quit IRC | 10:04 | |
*** fawadkhaliq has quit IRC | 10:05 | |
bauwser | johnthetubaguy: FYI, in case you missed markus_z's email https://review.openstack.org/#/c/247775/ | 10:06 |
*** amotoki_ has quit IRC | 10:06 | |
*** aix has joined #openstack-nova | 10:07 | |
johnthetubaguy | bauwser: yeah, we need to get going with reno soon, and adding some details that we have already missed, etc | 10:07 |
bauwser | johnthetubaguy: that's my next week's duty, I promised mriedem to take a look at all UpgradeImact merged changes since we need a reno file for all of them by m-1 | 10:07 |
bauwser | johnthetubaguy: but now, we're fully gating on reno | 10:08 |
bauwser | johnthetubaguy: so we should -1 any change not providing relnotes when needed, hence my change | 10:08 |
bauwser | johnthetubaguy: I'll respin a new PS based on markus's point, by trying to explain further how to build a note | 10:09 |
markus_z | bauwser: There is also a good chance that I'm overthinking this. | 10:09 |
bauwser | heh | 10:10 |
*** haomaiwang has quit IRC | 10:10 | |
bauwser | don't tell me *you* are overthinking, it's generally my case :) | 10:10 |
markus_z | :D | 10:10 |
*** nagyz has joined #openstack-nova | 10:11 | |
johnthetubaguy | bauwser: cools, I need to read through the meeting notes, not done that yet | 10:11 |
*** lpetrut1 has joined #openstack-nova | 10:13 | |
*** jinxing has joined #openstack-nova | 10:13 | |
*** lpetrut has quit IRC | 10:13 | |
*** lpetrut1 is now known as lpetrut | 10:13 | |
*** ptm_away is now known as PaulMurray | 10:13 | |
*** haomaiwang has joined #openstack-nova | 10:16 | |
openstackgerrit | Adelina Tuvenie proposed openstack/nova: Added support for new block device format in Hyper-V https://review.openstack.org/246298 | 10:18 |
*** pixelbeat has joined #openstack-nova | 10:19 | |
*** otter768 has joined #openstack-nova | 10:19 | |
*** ihrachys has quit IRC | 10:24 | |
*** jinxing has quit IRC | 10:24 | |
*** otter768 has quit IRC | 10:24 | |
*** mpavone has quit IRC | 10:24 | |
*** rk4n has joined #openstack-nova | 10:27 | |
*** chinmay_ has joined #openstack-nova | 10:27 | |
*** gcb has quit IRC | 10:28 | |
*** ihrachys has joined #openstack-nova | 10:28 | |
*** haomaiwang has quit IRC | 10:28 | |
*** haomaiwang has joined #openstack-nova | 10:28 | |
*** gcb has joined #openstack-nova | 10:28 | |
*** amotoki_ has joined #openstack-nova | 10:29 | |
*** fawadkhaliq has joined #openstack-nova | 10:30 | |
*** amotoki_ has quit IRC | 10:30 | |
*** shaohe_feng has quit IRC | 10:30 | |
*** markvoelker has joined #openstack-nova | 10:31 | |
*** ildikov has joined #openstack-nova | 10:31 | |
openstackgerrit | Balazs Gibizer proposed openstack/nova: Add service status notification https://review.openstack.org/245678 | 10:32 |
openstackgerrit | Balazs Gibizer proposed openstack/nova: Add infra for versioned notifications https://review.openstack.org/247024 | 10:32 |
openstackgerrit | Balazs Gibizer proposed openstack/nova: Make emitting versioned notifications configurable https://review.openstack.org/247564 | 10:32 |
*** gongysh__ has joined #openstack-nova | 10:32 | |
alex_xu | johnthetubaguy: morning, re: https://review.openstack.org/245543, we can totally get rid of disk_over_commit afer we get this http://specs.openstack.org/openstack/nova-specs/specs/mitaka/approved/check-destination-on-migrations.html. So we have two choice, one is depend on that, sencond is change disk_over_commit next time. | 10:32 |
alex_xu | johnthetubaguy: hope to hear your opinion which choice is beter | 10:32 |
alex_xu | s/beter/better/ | 10:32 |
AJaeger | markus_z, current release notes are always visible at http://docs.openstack.org/releasenotes/nova/ | 10:33 |
AJaeger | index page will come later once more projects have adopted it... | 10:33 |
johnthetubaguy | alex_xu: I was thinking we just make it default to false, for now? | 10:33 |
johnthetubaguy | alex_xu: it should all be done by claims in the resource tracker, not really by the scheduler, I think ndipanov had some plans around that | 10:33 |
alex_xu | johnthetubaguy: emm...why not choice true, as the when booting instance, we count the disk_over_commit | 10:33 |
markus_z | AJaeger: Is this always current? | 10:33 |
*** moshele has joined #openstack-nova | 10:34 | |
markus_z | AJaeger: Or a regular job, done every 1 week or so? | 10:34 |
alex_xu | johnthetubaguy: emm...that sounds good, I didn't notice ndipanov works | 10:34 |
AJaeger | markus_z: it gets updated after each merge - a post job | 10:34 |
markus_z | AJaeger: Yeahhh, awesome, thanks! | 10:34 |
AJaeger | markus_z: note that releasenotes/notes/ebtables-version-fde659fe18b0e0c0.yaml renders broken, see http://docs.openstack.org/releasenotes/nova/unreleased.html. | 10:35 |
*** gongysh has quit IRC | 10:35 | |
*** gongysh_ has quit IRC | 10:35 | |
AJaeger | "ebtables and libvirt requirements" is not the subsection header... | 10:35 |
*** gongysh has joined #openstack-nova | 10:35 | |
alex_xu | johnthetubaguy: ok, I'm ok with default as false, let me update the spec | 10:35 |
johnthetubaguy | alex_xu: really? I have never really dug into disk management as we only do RAM management right now | 10:35 |
johnthetubaguy | alex_xu: I was just thinking false as it matches block_migration | 10:36 |
*** markvoelker has quit IRC | 10:36 | |
markus_z | AJaeger: What's causing this? | 10:36 |
AJaeger | markus_z: misunderstanding on what prelude does ;) | 10:36 |
alex_xu | johnthetubaguy: yea, I checked the code, we use flavor's disk size to consume the resource | 10:36 |
johnthetubaguy | alex_xu: ideally we default to "resource tracker does its thing" I guess | 10:36 |
ndipanov | alex_xu, so like johnthetubaguy says - I really don't think that check_destination makes sense | 10:36 |
ndipanov | we either claim | 10:37 |
AJaeger | markus_z: let me sent a fix... | 10:37 |
ndipanov | or we say - admin knows what they are doing so let them do it | 10:37 |
ndipanov | after talking to HP/rax folks in Tokyo | 10:37 |
johnthetubaguy | ndipanov: I think we need both, check_destination just doesn't fix the disk issue | 10:37 |
ndipanov | johnthetubaguy, what does check destination give us | 10:37 |
alex_xu | ndipanov: actually disk_over_commit option in live-migration is buggy | 10:37 |
bauwser | ndipanov: have you read the spec ? :) | 10:38 |
*** rk4n has quit IRC | 10:38 | |
johnthetubaguy | ndipanov: its for the case when the admin is making a free host, it checks the affinity | 10:38 |
ndipanov | in a racy way | 10:38 |
johnthetubaguy | ndipanov: anti-affinity is the only useful check | 10:38 |
openstackgerrit | Andreas Jaeger proposed openstack/nova: Fix ebtables-version release note https://review.openstack.org/248013 | 10:39 |
AJaeger | markus_z: ^ | 10:39 |
johnthetubaguy | ndipanov: yep | 10:39 |
ndipanov | I really don't get it - we go for obviously wrong but kinda easy to do currently | 10:39 |
bauwser | ndipanov: the idea is to make sure that we call the scheduler, even if the scheduler can make wrong decisions | 10:39 |
markus_z | AJaeger: thanks! I'm really glad we are using this. It's a great tool. | 10:40 |
ndipanov | ok I will read the spec | 10:40 |
AJaeger | markus_z: yep, it is - great work by dhellmann | 10:40 |
alex_xu | ndipanov: there is force option in bauwser's propose, so that can resolve your concern as my understand | 10:40 |
bauwser | ndipanov: because that's not the role of claims to select a destination, right? | 10:40 |
markus_z | AJaeger: And it doesn't interrupt my workflow | 10:40 |
ndipanov | again | 10:40 |
johnthetubaguy | ndipanov: its really just the starting point, we need to get the API working properly, reducing the race window, etc, totally need the resource tracker claims, etc | 10:40 |
AJaeger | markus_z: wow! That's an extra +3 ;) | 10:40 |
bauwser | ndipanov: but of course, that still needs claims able to get the Reqspec and do a good call | 10:40 |
markus_z | AJaeger: :) | 10:41 |
tdurakov | sdague, johnthetubaguy, hi | 10:41 |
ndipanov | so it;s like this - the only way to make this stuff not racy is to re-check the schduler decision on a host holding a global host lock | 10:41 |
ndipanov | currently | 10:41 |
johnthetubaguy | alex_xu: I was really just thinking about doing it all in one API microversion, I could be pushing that too far, just curious if its possible really :) | 10:41 |
ndipanov | everything else is broken | 10:41 |
tdurakov | https://review.openstack.org/#/c/247081/ - nfs shared storage for live-migration job | 10:41 |
ndipanov | (if jay moves claims to the scheduler then it's a different story) | 10:42 |
bauwser | ndipanov: I'm not sure you read the spec when talking about a check_destination() method | 10:42 |
alex_xu | johnthetubaguy: ok, no problem. | 10:42 |
ndipanov | bauwser, ok re-reading | 10:42 |
*** ihrachys has quit IRC | 10:43 | |
alex_xu | johnthetubaguy: you still prefer default value? | 10:43 |
*** e0ne has quit IRC | 10:43 | |
*** rk4n has joined #openstack-nova | 10:44 | |
alex_xu | johnthetubaguy: actually I prefer get rid of disk_over_commit, but it depends on bauwser's spec, and that spec depend requestspec obj persisnt, and that depends on requestspec obj works...:( | 10:44 |
alex_xu | such long dependence~ | 10:44 |
bauwser | yeah, that's a big and massive change | 10:44 |
bauwser | I should work on check-destinations BP soon | 10:45 |
johnthetubaguy | alex_xu: I am OK with just going for a default for now, so we break the dependency chain | 10:45 |
ndipanov | bauwser, so your spec makes no mention of the fact that the result of the check_destination is meaningless without the claim | 10:45 |
ndipanov | so I stopped reading | 10:45 |
bauwser | there is *no* check_destination call | 10:45 |
*** ihrachys has joined #openstack-nova | 10:45 | |
ndipanov | bauwser, "We should modify that logic to explicitly call the Scheduler any time a move (ie. either a live-migration or an evacuation) is requested (whether the destination host is provided or not) so that the Scheduler would verify the destination host thru all the enabled filters and if successful consume the instance usage from its internal HostState." | 10:46 |
bauwser | there is a call which is already existing, but under specific constraints, ie. select_dest() | 10:46 |
*** ankit_ag has quit IRC | 10:46 | |
ndipanov | that is literally the same as not calling it unless you plan to do the claim-retry dance | 10:46 |
alex_xu | johnthetubaguy: ok, I will choice disk_over_commit default value as True, because booting instance counting the flavor's disk size. | 10:46 |
alex_xu | johnthetubaguy: does sounds ok? | 10:46 |
bauwser | ndipanov: right, and what's the problem here? we already do that, but only if no host is provided | 10:46 |
ndipanov | so why do you want to add more code that does not fix anything? | 10:46 |
*** rushiagr is now known as rushiagr_away | 10:46 | |
*** ankit_ag has joined #openstack-nova | 10:47 | |
bauwser | okay, I'm going to take a coffee | 10:47 |
ndipanov | it literally does not fix a problem | 10:47 |
*** jianghuaw has joined #openstack-nova | 10:48 | |
openstackgerrit | Timofey Durakov proposed openstack/nova: DO NOT Merge NFS setup for live-migration job https://review.openstack.org/247081 | 10:48 |
*** sacharya has joined #openstack-nova | 10:48 | |
bauwser | ndipanov: you're persisting on thinking about a different problem IMHO | 10:48 |
ndipanov | bauwser, the only bit of the spec that makes sense is to mark the forced instances | 10:48 |
bauwser | ndipanov: I'm not saying I'll fix all the problems, but just the fact that when calling a destination, we don't verify the scheduler, that's it | 10:48 |
ndipanov | bauwser, I am reading your problem description | 10:48 |
ndipanov | so my answer to that is: "don't do it because it's useless" | 10:49 |
bauwser | well, I'm very glad to see your point now that I showed you that spec for like 6 months ago | 10:49 |
bauwser | I don't pretend to fix the claim issue | 10:49 |
ndipanov | I said the same thing several times in the past | 10:49 |
ndipanov | not sure if it was 6 months ago | 10:50 |
* bauwser whispers | 10:50 | |
*** tdurakov has quit IRC | 10:50 | |
bauwser | well, okay, what do you recommend ? | 10:50 |
*** rk4n has quit IRC | 10:50 | |
*** tdurakov has joined #openstack-nova | 10:51 | |
bauwser | stop doing that, not persisting the spec and calling the scheduler, and just fixing claims rather? | 10:51 |
*** e0ne has joined #openstack-nova | 10:51 | |
*** sacharya has quit IRC | 10:52 | |
ndipanov | bauwser, so still calling the scheduler makes sense only if you do a claim on the compute host even if that claim would normally fail | 10:52 |
ndipanov | bauwser, it needs to recalculate some stuff in the general case | 10:53 |
bauwser | I never said the contrary :) | 10:53 |
*** tdurakov has quit IRC | 10:53 | |
*** AJaeger has left #openstack-nova | 10:54 | |
*** EinstCrazy has joined #openstack-nova | 10:54 | |
ndipanov | bauwser, so the spec would make sense if it also includes modifications to the claim such that it would still happen but just not fail if it otherwise would | 10:55 |
bauwser | ndipanov: so I feel it's another spec IMHO | 10:55 |
ndipanov | bauwser, well I am not sure this one makes any sense without it really ... | 10:56 |
*** mpavone has joined #openstack-nova | 10:56 | |
bauwser | I got that - but my thoughts are that it's still a benefit anyway | 10:56 |
*** gszasz has joined #openstack-nova | 10:56 | |
ndipanov | there is benefit in the fact that now we know which instances are forced | 10:57 |
*** rk4n has joined #openstack-nova | 10:58 | |
bauwser | ndipanov: consider the AZ case | 10:58 |
*** ihrachys_ has joined #openstack-nova | 10:58 | |
bauwser | ndipanov: that's not a claim thing, right? | 10:58 |
*** lpetrut has quit IRC | 10:58 | |
bauwser | ndipanov: but we still have an AZFilter | 10:59 |
*** lpetrut has joined #openstack-nova | 10:59 | |
*** tdurakov has joined #openstack-nova | 10:59 | |
bauwser | ndipanov: https://bugs.launchpad.net/nova/+bug/1452568 | 10:59 |
openstack | Launchpad bug 1452568 in OpenStack Compute (nova) "nova allows to live-migrate instance from one availability zone to another" [Low,Confirmed] - Assigned to Sylvain Bauza (sylvain-bauza) | 11:00 |
johnthetubaguy | my take is folks have been asking for this for a while, and this is a good step in the right direction, sure it has bit limitations, but live-migrate generally has bigger ones right now, that we are working through | 11:00 |
bauwser | ndipanov: don't you feel that the spec would help that? | 11:00 |
ndipanov | bauwser, well that is a bit of a grey area because AZ of the host does not change often in reality | 11:00 |
*** ihrachys has quit IRC | 11:00 | |
bauwser | ndipanov: heh, that's an hypothesis which is really huge | 11:01 |
bauwser | ndipanov: I mean, hosts can easily move from one agg to another, nope? | 11:01 |
johnthetubaguy | hang on, this is upside down right? | 11:01 |
bauwser | that's not racy in the terms of what you think | 11:01 |
bauwser | johnthetubaguy: when moving an host from one agg to another, we check the instances | 11:02 |
bauwser | johnthetubaguy: so that's a safe operation | 11:02 |
johnthetubaguy | bauwser: right | 11:02 |
bauwser | johnthetubaguy: well, probably racy tho if we consider instances being scheduled tho | 11:03 |
ndipanov | yes | 11:03 |
ndipanov | so I am thinking about what johnthetubaguy said | 11:03 |
*** ihrachys_ has quit IRC | 11:04 | |
ndipanov | there definitely is some usefulness to the spec - the fact that we track which ones got force-migrated | 11:04 |
*** ihrachys has joined #openstack-nova | 11:05 | |
bauwser | ndipanov: tbh, I was also thinking of proposing another spec for getting rid of force_hosts and do the same for boot | 11:05 |
bauwser | ndipanov: which would mean that you'd get a requested_destination field set for boot too | 11:06 |
bauwser | atm, I'm very frustrated by how we deal with force_hosts | 11:06 |
openstackgerrit | Alex Xu proposed openstack/nova-specs: Make block_migration and host flags in live-migration API https://review.openstack.org/245543 | 11:06 |
ndipanov | what annoys me is that instead of working towards a cleaner abstraction - we keep working around these special cases | 11:06 |
*** alex_klimov has quit IRC | 11:07 | |
bauwser | I don't get what could be a cleaner abstraction ? | 11:07 |
bauwser | I got that the spec should make sure that object goes down to the compute manager, so that at the end, the claim is able to read it | 11:08 |
*** haomaiwang has quit IRC | 11:08 | |
*** fawadkhaliq has quit IRC | 11:08 | |
bauwser | but we're using an object which is well-formatted, and where we identify whether a destination was proposed or not | 11:09 |
ndipanov | so there's a ton of stuff we need to check before building on a compute host, some of them we have to recalculate and persist | 11:09 |
ndipanov | currently we don't do it for some things | 11:10 |
ndipanov | we use a completely different code path for some (instance groups) | 11:10 |
*** ihrachys has quit IRC | 11:10 | |
alex_xu | johnthetubaguy: I guess know why you think disk_over_commit default value should be false. Actually disk_over_commit isn't related to disk overcommit ratio. It's just about count disk size as image actual size or image virtual size | 11:11 |
alex_xu | here is the code reference https://github.com/openstack/nova/blob/3e0bbc332726a5df3031628b71ef3517b5209fb6/nova/virt/libvirt/driver.py#L5325 | 11:11 |
johnthetubaguy | alex_xu: yeah, sorry, I should have linked you to that, yeah | 11:11 |
johnthetubaguy | alex_xu: it seems odd really, that is not just managed by the resource tracker | 11:11 |
bauwser | ndipanov: right, that's why I think it's another spec to consider | 11:11 |
alex_xu | johnthetubaguy: yea, very odd | 11:11 |
ndipanov | bauwser, well that spec is approved so | 11:12 |
johnthetubaguy | alex_xu: don't think the other virt drivers implement it anyway, its tempting to drop it all together, but I wish I knew why it was added... | 11:12 |
ndipanov | yeah we need another spec | 11:12 |
alex_xu | johnthetubaguy: yea, hope we can drop it future | 11:13 |
* bauwser needs to go afk | 11:13 | |
*** ihrachys has joined #openstack-nova | 11:13 | |
*** tdurakov has quit IRC | 11:15 | |
openstackgerrit | Alex Xu proposed openstack/nova-specs: Make block_migration and host flags in live-migration API https://review.openstack.org/245543 | 11:19 |
johnthetubaguy | alex_xu: yeah, I wonder if we should just drop it in your spec, or does that seem too soon? | 11:20 |
*** alex_klimov has joined #openstack-nova | 11:20 | |
*** jianghuaw has quit IRC | 11:20 | |
alex_xu | johnthetubaguy: oops, sorry, I must miss understand your words again. So let's depend on bauwser's spec? | 11:21 |
alex_xu | johnthetubaguy: we can't drop it directly, there will be no disk size checking for user specify a host in the api | 11:22 |
johnthetubaguy | alex_xu: ah, I am not sure I am clear in my own head right now, I think we could just drop that, independent of bauwser's spec | 11:22 |
alex_xu | johnthetubaguy: emm...it is ok without disk size check? | 11:22 |
*** gongysh__ has quit IRC | 11:23 | |
*** ihrachys_ has joined #openstack-nova | 11:23 | |
*** rushiagr_away is now known as rushiagr | 11:23 | |
*** dims has joined #openstack-nova | 11:24 | |
*** zenoway has quit IRC | 11:24 | |
*** ihrachys has quit IRC | 11:24 | |
*** zenoway has joined #openstack-nova | 11:24 | |
bauwser | alex_xu: johnthetubaguy: sorry I wasn't following that spec so I don't have very much context | 11:24 |
bauwser | I can't really say whether my spec is needed or not, but we can discuss | 11:25 |
bauwser | and see why it came up in the discussion | 11:25 |
bauwser | alex_xu: I'll try to review your spec asap once I'm done with lunch | 11:26 |
alex_xu | johnthetubaguy: ok, I got another idea, remove the disk_over_commit, and add right disk size check in the code. | 11:26 |
johnthetubaguy | alex_xu: hmm, well it will end up defaulting to something I guess, in the libvirt driver, so it just keeping do that for new API calls? | 11:26 |
alex_xu | bauwser: thanks :) it's about remove the disk_over_commit flag in the live-migrate api | 11:26 |
bauwser | alex_xu: sure, that I understood but I need to understand more the problem | 11:26 |
bauwser | and how/if my spec could help, to answer your question | 11:27 |
johnthetubaguy | alex_xu: honestly I would OK leaving the current check, for now, and just let the other specs on the live-migrate add the proper resource tracker checks | 11:27 |
johnthetubaguy | alex_xu: so leaving the current check, but removing the config option for that check from the API, basically | 11:27 |
alex_xu | johnthetubaguy: ok, no problem, that sounds good choice | 11:28 |
johnthetubaguy | bauwser: basically, make live-migrate API be {live-migrate: null}, rather than it asking silly questions you can't really answer | 11:28 |
*** salv-orlando has quit IRC | 11:28 | |
alex_xu | bauwser: the current disk_size check in live migrate is buggy, but after you spec use scheduler check the disk size, then everything is good. | 11:28 |
*** salv-orlando has joined #openstack-nova | 11:29 | |
bauwser | johnthetubaguy: oh that, sure | 11:29 |
johnthetubaguy | can you spot my "love" of the current API in my text :) | 11:29 |
bauwser | alex_xu: the 'host' flag is mentioned in the check-destinations spec | 11:29 |
*** fawadkhaliq has joined #openstack-nova | 11:30 | |
bauwser | alex_xu: not sure it needs to be scoped by your spec? | 11:30 |
*** rushiagr is now known as rushiagr_away | 11:30 | |
*** achanda has joined #openstack-nova | 11:31 | |
alex_xu | bauwser: it looks like no conflict in our specs. I just make host optionl in the api , and the default value is None | 11:31 |
*** markus_z_bnc has joined #openstack-nova | 11:31 | |
openstackgerrit | Adelina Tuvenie proposed openstack/nova: Added support for new block device format in Hyper-V https://review.openstack.org/246298 | 11:33 |
openstackgerrit | Adelina Tuvenie proposed openstack/nova: Added support for new block device format in vmops https://review.openstack.org/246299 | 11:33 |
openstackgerrit | Adelina Tuvenie proposed openstack/nova: Add support for setting boot order in Hyper-V https://review.openstack.org/248029 | 11:33 |
bauwser | alex_xu: right my spec was vague about the difference between evacuate and live-migrate | 11:33 |
bauwser | alex_xu: evacuate makes host optional, while live-migrate provides an empty string for host IIRC | 11:33 |
*** ociuhandu has quit IRC | 11:34 | |
bauwser | alex_xu: so your spec just unifys both behaviours, which I'm fine | 11:34 |
*** Sree has joined #openstack-nova | 11:34 | |
bauwser | unifies even | 11:34 |
alex_xu | bauwser: yea :) | 11:34 |
*** smatzek has joined #openstack-nova | 11:35 | |
bauwser | alex_xu: tbh, I was a bit okay with providing the options, given that the default policy was for admins - it meant that when wanting to live-migrate, you were having to know the physical infrastructure, but since we can have nullable hosts, then it would mean that you would need to have an homogenous infrastructure | 11:37 |
bauwser | alex_xu: so your spec is okay - to be clear, the problem is more than the admin is not sure that the destination will have the options he said | 11:38 |
*** tdurakov has joined #openstack-nova | 11:39 | |
*** tdurakov has quit IRC | 11:39 | |
mdbooth | paul-carlton: Hey, I'm getting a bit ratholed trying to work out the scope of the migrate libvirt volumes spec. | 11:40 |
mdbooth | Got a sec? | 11:40 |
markus_z | pkholkin: Could it be that your bug is superseded? https://bugs.launchpad.net/nova/+bug/1510504 | 11:40 |
openstack | Launchpad bug 1510504 in OpenStack Compute (nova) "Keypairs list results not limited on database server-side" [Low,New] - Assigned to Pavel Kholkin (pkholkin) | 11:40 |
*** doug-fish has quit IRC | 11:42 | |
alex_xu | bauwser: yea, agree with you | 11:42 |
mdbooth | paul-carlton: The issue is I'm not sure where the libvirt storage pools spec ends. It's not clear to me if that spec implies that copying to the correct location on a remote host is already implemented, but using existing copy methods. | 11:42 |
mdbooth | If it does imply that, then the migrate libvirt storage pools spec becomes trivial. | 11:42 |
*** doug-fish has joined #openstack-nova | 11:42 | |
bauwser | alex_xu: so, about the disk_over_commit flag, I don't see how my spec can help, because it sounds that something which isn't read by the scheduler anyway, right? | 11:43 |
*** moshele has quit IRC | 11:44 | |
alex_xu | bauwser: emm...my understand is your spec will use scheduler to check the user specified host. Then scheduler will ensure the host whether have enough disk. | 11:44 |
bauwser | alex_xu: johnthetubaguy: meh, the disk_over_commit option sounds like a libvirt hack :/ | 11:44 |
johnthetubaguy | bauwser: +1 | 11:45 |
bauwser | that's a bit more complicated | 11:45 |
alex_xu | bauwser: yea, very strange behaviour | 11:45 |
*** salv-orlando has quit IRC | 11:45 | |
bauwser | alex_xu: so IIUC, that hacks allows you to count only the real size of the disk, not the allocated size | 11:45 |
bauwser | because QCOW2 does compression | 11:45 |
alex_xu | bauwser: yea | 11:45 |
bauwser | but that's a very terrible way to pass that info IMHO :/ | 11:46 |
bauwser | that should be a RT thing like johnthetubaguy said | 11:46 |
bauwser | ie. | 11:46 |
*** doug-fish has quit IRC | 11:47 | |
bauwser | compute nodes should have a libvirt CONF flag saying whether the disk space is taking real size or allocated size, and the virt driver should report accordingly the disk size to RT | 11:47 |
bauwser | because you could decide to have a different policy of counting disks based on your compute node | 11:47 |
bauwser | exactly like allocation ratios | 11:47 |
alex_xu | bauwser: I more prefer just get rid of it | 11:47 |
bauwser | alex_xu: but that's a feature, we can't just pretend it never existed :/ | 11:48 |
alex_xu | bauwser: just use the resource tracker and scheduler way to count resource | 11:48 |
bauwser | alex_xu: sure, but you need to provide that ability to either count on real size or allocated size | 11:48 |
bauwser | (speaking of the metrics that the libvirt driver reports to Rt) | 11:48 |
bauwser | oh man, I don't know when it has been introduced in the REST API, but I would have been preferred to have never seen it | 11:49 |
johnthetubaguy | bauwser: yup, we have a few specs around "improving" that somewhere | 11:50 |
johnthetubaguy | the metrics API that is | 11:50 |
bauwser | johnthetubaguy: any idea of what happens if you have a different hypervisor ? it's passed as part of the body, so I guess that if you don't care for your virt driver, you ignore it, right? | 11:50 |
johnthetubaguy | bauwser: yeah, its generally ignored by everyone else, AFAIK | 11:50 |
bauwser | argh | 11:51 |
johnthetubaguy | bauwser: to me, what it should be is like the CPU and RAM allocation ratios | 11:51 |
bauwser | totally agreed | 11:51 |
johnthetubaguy | its just totally broken in its current form | 11:51 |
bauwser | now the ratios are per compute | 11:51 |
johnthetubaguy | right | 11:51 |
bauwser | but ideally we should even not report them to the scheduler | 11:51 |
ndipanov | johnthetubaguy, bauwser this might be related https://bugs.launchpad.net/nova/+bug/1517442 | 11:51 |
openstack | Launchpad bug 1517442 in OpenStack Compute (nova) "libvirt/xenapi: disk_available_least reported by the driver does not take into account instances being migrated to/from the host" [High,New] | 11:51 |
bauwser | johnthetubaguy: we should just give the calculated size of the disk to the RT | 11:52 |
johnthetubaguy | so I could see a case where you might want to push a few things together real close for a little bit, but thats not what this really enables | 11:52 |
bauwser | johnthetubaguy: no need to pass a ratio | 11:52 |
bauwser | johnthetubaguy: so there are 2 possibilites | 11:52 |
johnthetubaguy | ndipanov: thats back to needing claims I think | 11:52 |
ndipanov | yes | 11:52 |
bauwser | johnthetubaguy: either we consider that we provide the disk size *and* the flag info to the scheduler so that the DiskFilter does the right thing - but we also need to claim that correctly like ndipanov said | 11:53 |
*** tdurakov has joined #openstack-nova | 11:53 | |
*** fawadkhaliq has quit IRC | 11:53 | |
bauwser | johnthetubaguy: or, we just provided a consolidated view of the disk size hiden by the virt logic | 11:53 |
bauwser | the latter has my preference | 11:54 |
openstackgerrit | dstepanenko proposed openstack/nova: WIP: This is 3rd part of changes according to pci-generate-stats blueprint. https://review.openstack.org/247692 | 11:54 |
johnthetubaguy | so the feature makes no sense, I think we just drop it | 11:54 |
bauwser | johnthetubaguy: well, the feature can make sense | 11:54 |
johnthetubaguy | regardless, we need to fix up the claims to match migrate/resize | 11:54 |
bauwser | johnthetubaguy: but people can use a disk allocation ratio for that | 11:54 |
ndipanov | bauwser, that bug is precisely about the fact that we can't do that | 11:55 |
bauwser | johnthetubaguy: agreed | 11:55 |
*** ihrachys_ has quit IRC | 11:55 | |
* bauwser reads the bug | 11:55 | |
johnthetubaguy | what I mean is, that API isn't needed in any of the sensible features here | 11:55 |
openstackgerrit | Timofey Durakov proposed openstack/nova: NFS setup for live-migration job https://review.openstack.org/247081 | 11:56 |
bauwser | johnthetubaguy: oh totally agreed | 11:56 |
bauwser | by no way, it should have been landed as an API option | 11:56 |
bauwser | ndipanov: so agree with you too | 11:57 |
*** zenoway has quit IRC | 11:57 | |
*** ihrachys has joined #openstack-nova | 11:57 | |
bauwser | ndipanov: disk_available_least is terrible for operators | 11:57 |
bauwser | ndipanov: because the logic is different based on which virt driver you use | 11:57 |
*** zenoway has joined #openstack-nova | 11:57 | |
tdurakov | johnthetubaguy, can i ask you to review this https://review.openstack.org/#/c/247081/ | 11:58 |
*** markus_z is now known as markus_z_meeting | 11:58 | |
*** daemontool_ has quit IRC | 11:58 | |
pkholkin | markus_z: yes, it relates to my spec but it is not approved | 11:58 |
pkholkin | I spoke to Matt some days ago and he marked it -1 because of related spec, see comments there | 11:59 |
openstackgerrit | Alex Xu proposed openstack/nova-specs: Make block_migration and host flags in live-migration API https://review.openstack.org/245543 | 11:59 |
bauwser | alex_xu: johnthetubaguy: so the problem I see with removing disk_over_commit in the request body is that we don't provide a migration path for operators wanting that feature | 11:59 |
pkholkin | johnthetubaguy: Hello John, I have a question about use_slave concept, is it true that it removed? | 12:00 |
alex_xu | bauwser: emm...does admin really need migrate instance without consider overcommit? | 12:00 |
pkholkin | johnthetubaguy: it relates to our enginefacade patches | 12:00 |
bauwser | alex_xu: that's not a request thing | 12:00 |
bauwser | alex_xu: that's a resource usage thing | 12:01 |
johnthetubaguy | pkholkin: I haven't had chance to dig into that properly | 12:01 |
johnthetubaguy | pkholkin: I hope we don't remove it, but it looked like the current patches were going that way | 12:01 |
*** daemontool_ has joined #openstack-nova | 12:01 | |
alex_xu | bauwser: yea, we need resource tracker to do that right? | 12:01 |
johnthetubaguy | bauwser: what does the feature do for them? | 12:01 |
bauwser | alex_xu: that's one proposal | 12:02 |
pkholkin | johnthetubaguy: yes, so the question is that in enginefacade there is async decorator, please look this patch https://review.openstack.org/#/c/243496/2/ | 12:02 |
bauwser | johnthetubaguy: eg. say that you asked for a 10GB instance | 12:02 |
*** zhangjn has joined #openstack-nova | 12:02 | |
*** zhangjn has quit IRC | 12:02 | |
pkholkin | johnthetubaguy: using this decorator we will try to use slave_connection if it is in conf file | 12:02 |
*** zenoway has quit IRC | 12:02 | |
bauwser | johnthetubaguy: the first time you're booting your instance, the real consumed disk space on the hypervisor is 2GB | 12:02 |
bauwser | johnthetubaguy: because QCOW2 is a compressed format | 12:03 |
tdurakov | alex_xu, read your spec about api for live-migration, imho it's good idea to make things optional, not sure about overcommit flag, will leave comment later:) | 12:03 |
johnthetubaguy | pkholkin: but thats a problem, it covers every use of that DB API | 12:03 |
pkholkin | johnthetubaguy: but previously it can be managed by develepor using use_slave parameter | 12:03 |
*** zhangjn has joined #openstack-nova | 12:03 | |
alex_xu | tdurakov: thanks :) | 12:03 |
bauwser | johnthetubaguy: so I guess that operators want to have the ability to say "I don't care, only count the real disk size" | 12:03 |
pkholkin | johnthetubaguy: yes, using async will always try to use slave_connection | 12:03 |
bauwser | johnthetubaguy: here, the features makes that decision per-request | 12:04 |
johnthetubaguy | pkholkin: right that my problem with it, I feel we need to explicitly select that on a per call basis, which is why we added the parameter | 12:04 |
bauwser | but since RT claims don't do that, the feature is broken | 12:04 |
alex_xu | bauwser: if mean RT can only count real disk size? | 12:04 |
*** haomaiwa_ has joined #openstack-nova | 12:05 | |
alex_xu | s/if/you/ | 12:05 |
bauwser | alex_xu: RT counts what the driver says | 12:05 |
tdurakov | alex_xu, tbh, I don't like live-migration with user-provided host, as it could break a lot of things | 12:05 |
* bauwser *really* needs to eat some food | 12:06 | |
pkholkin | johnthetubaguy: do you think if use_slave really helps with performance or maybe we can continue removing it? | 12:06 |
alex_xu | bauwser: go for food first :) | 12:06 |
johnthetubaguy | pkholkin: I totally didn't see this in the spec somehow :( just came up in discussion on IRC the other day when I -1ed a patch for adding use_slave | 12:06 |
johnthetubaguy | pkholkin: so the async decorator is doing a similar thing, as I understand it, its the slightly lagging read slave | 12:07 |
johnthetubaguy | pkholkin: its really about readability, in some ways | 12:08 |
johnthetubaguy | pkholkin: the change you linked is good, because there is no major change, as I understand it | 12:08 |
johnthetubaguy | pkholkin: the problem is when someone re-uses that DB API call in some other bit of code, and it starts doing use_slave=False, and that causes bugs | 12:09 |
*** zenoway has joined #openstack-nova | 12:09 | |
johnthetubaguy | pkholkin: well I mean when it gets used, and they didn't realise that was accessing the async read slave | 12:10 |
*** ihrachys has quit IRC | 12:11 | |
*** zenoway has quit IRC | 12:12 | |
*** ihrachys has joined #openstack-nova | 12:12 | |
*** zenoway has joined #openstack-nova | 12:12 | |
*** hparekh has quit IRC | 12:12 | |
tdurakov | johnthetubaguy, ping | 12:13 |
johnthetubaguy | tdurakov: hey | 12:13 |
tdurakov | there is a patch for nfs shared storage of dedicated l-m job | 12:13 |
tdurakov | could you review it? | 12:13 |
johnthetubaguy | tdurakov: we really, really, really need to specify the host for live-migrate, from an operator perspective, we should make sure we document those usecases better, I think PaulMurray has a patch up for that somewhere. | 12:14 |
johnthetubaguy | tdurakov: I can do, whats the link? | 12:14 |
johnthetubaguy | tdurakov: I don't see when I am added to reviews, that has turned into SPAM | 12:14 |
tdurakov | johnthetubaguy, https://review.openstack.org/#/c/247081/ | 12:15 |
PaulMurray | johnthetubaguy, we do need the host - absolutely | 12:15 |
tdurakov | about target host, i'm not talking that we should remove this option | 12:16 |
tdurakov | but we could rewrite implementation for that in conductor | 12:16 |
PaulMurray | johnthetubaguy, I spent some time this morning looking at that disk_over_commit option BTW | 12:16 |
johnthetubaguy | tdurakov: oh, totally agreed with that | 12:16 |
PaulMurray | johnthetubaguy, we have used it and I think I understand why now - adding to the list of reviews on it | 12:16 |
openstackgerrit | jichenjc proposed openstack/nova-specs: Add ips-add-max-and-type.rst https://review.openstack.org/247496 | 12:17 |
*** salv-orlando has joined #openstack-nova | 12:17 | |
johnthetubaguy | PaulMurray: the first idea was making it optional, and defaulting to False | 12:17 |
pkholkin | johnthetubaguy: it would be nice if we remove using use_slave and use only @reader, but as I understand it is not good for performance | 12:17 |
tdurakov | i'm big + for making all params these param in api optional | 12:18 |
PaulMurray | johnthetubaguy, it looks out of whack with the usual way of controlling over commit | 12:18 |
johnthetubaguy | pkholkin: yes, the idea is when folk shave a DB read slave, we can send some of the traffic there, so the primary DB nodes have less load | 12:18 |
pkholkin | johnthetubaguy: if we still want it we can try to manage it in layer above but it is not beautiful or maybe try to change smth in oslo.db | 12:18 |
openstackgerrit | Ankit Agrawal proposed openstack/nova: Remove redundant migration status update https://review.openstack.org/247519 | 12:18 |
PaulMurray | and seems that is because libvirt checks available disk in the check_destinations callsa | 12:18 |
PaulMurray | so its needed to control that one check | 12:18 |
*** e0ne has quit IRC | 12:19 | |
PaulMurray | I'd like to see it go personally | 12:19 |
johnthetubaguy | PaulMurray: yeah, we were discuss how that should all be resource tracker really | 12:19 |
*** e0ne has joined #openstack-nova | 12:19 | |
johnthetubaguy | yeah, I think the debate is optional or kill it, at this point | 12:19 |
johnthetubaguy | I am leaning towards kill it | 12:19 |
johnthetubaguy | force is a better metaphor, where that means: if its at all possible, put it there | 12:20 |
*** Sree_ has joined #openstack-nova | 12:20 | |
*** otter768 has joined #openstack-nova | 12:20 | |
tdurakov | PaulMurray, correct me if i wrong disk_over_commit is used only here: https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L5325 | 12:20 |
*** Sree_ is now known as Guest43431 | 12:20 | |
*** ihrachys has quit IRC | 12:21 | |
PaulMurray | johnthetubaguy, I saw something else in there too, | 12:21 |
PaulMurray | tdurakov, ^ | 12:22 |
johnthetubaguy | tdurakov: thats the only of it I found | 12:22 |
tdurakov | PaulMurray, well, i couldn't find any | 12:23 |
johnthetubaguy | so I should run away and get some lunch, getting low blood sugar, back soon ish | 12:23 |
PaulMurray | From discussions with mdbooth and danpb about the way qcow2 is handled I believe they make files to the full size of the image dispite the qcow2 being smaller | 12:23 |
pkholkin | johnthetubaguy: I also want to point that NO patches removing use_slave were already merged | 12:23 |
PaulMurray | I wonder if that is related | 12:23 |
*** Sree has quit IRC | 12:23 | |
PaulMurray | So even if you have enough disk space for the actual size, libvirt might grab more, so its hard to over commit | 12:24 |
PaulMurray | hard to allow it ^^ | 12:24 |
johnthetubaguy | pkholkin: ah, OK, thats good, so we should decide about this ASAP, have you spoken to dansmith about this? | 12:24 |
PaulMurray | (That is more of a question than a statement) | 12:24 |
PaulMurray | johnthetubaguy, ^^ | 12:24 |
tdurakov | johnthetubaguy, PaulMurray, if nfs is being merged what about moving l-m job from experimental to check pipeline | 12:24 |
danpb | PaulMurray: yep, that's correct, if preallocation is turned on | 12:24 |
pkholkin | johnthetubaguy: that's why I write to you to come to the decision. No, I speak only to you right now | 12:25 |
*** otter768 has quit IRC | 12:25 | |
johnthetubaguy | PaulMurray: yeah, its probably broken anyways, although depends on your storage system, as zeros may take up no space | 12:25 |
johnthetubaguy | pkholkin: cool, appreciate you reaching out | 12:25 |
pkholkin | we marked all suspicious patches as wip | 12:26 |
johnthetubaguy | pkholkin: so the use case was really that we needed a way to move DB load from the primary servers to the read slaves, those slaves will have a slight lag in updates, depending on how they are setup, so we need to be careful about how they are used, if that makes sense? | 12:26 |
pkholkin | johnthetubaguy: yes, the idea was clear | 12:27 |
PaulMurray | tdurakov, are there any tests being exercised yet? | 12:28 |
*** jianghuaw has joined #openstack-nova | 12:28 | |
johnthetubaguy | pkholkin: just checking what we wrote in here: http://specs.openstack.org/openstack/nova-specs/specs/juno/implemented/juno-slaveification.html | 12:28 |
PaulMurray | or would it be effetively an empty job | 12:28 |
tdurakov | PaulMurray, well, it's already tests l-m with no-shared storage and nfs patch is on review | 12:29 |
*** sudipto has quit IRC | 12:29 | |
tdurakov | so, it's definitely not empty job:) | 12:29 |
*** alejandrito has joined #openstack-nova | 12:30 | |
PaulMurray | tdurakov, I don't actual know the the policy for adding a non-voting check job - johnthetubaguy ? | 12:30 |
johnthetubaguy | pkholkin: I think the main argument against the decorator approach, is really just that it fixes that DB call to always read from the slave, which might not make sense, and its less explicit where its used that its possibly reading stale data, hence the use_slave parameter approach, its very ugly, but was the best way to meet all the requirements | 12:30 |
markus_z_meeting | melwitt: Is your name here still valid? https://wiki.openstack.org/wiki/Nova/BugTriage#Step_2:_Triage_Tagged_Bugs | 12:30 |
johnthetubaguy | PaulMurray: usually its do experimental first to make sure its working | 12:30 |
PaulMurray | tdurakov, what were the next things you were going to add - nfs was the first of a list wasn't it? | 12:31 |
markus_z_meeting | melwitt: There are 7 bugs for the novaclient which aren't triaged, would be cool if you could have a look. If you're the wrong contact here, please let me know. | 12:31 |
tdurakov | PaulMurray, it's only cep left | 12:31 |
tdurakov | *ceph | 12:32 |
pkholkin | johnthetubaguy: ok, we will think for some time how to support use_slave and make the code good | 12:32 |
* johnthetubaguy really has to go get food now, hungry | 12:32 | |
*** markvoelker has joined #openstack-nova | 12:32 | |
johnthetubaguy | pkholkin: thanks, it would be good to end up with something better than the current approach | 12:32 |
*** ihrachys has joined #openstack-nova | 12:32 | |
pkholkin | johnthetubaguy: ok, we will try and share the results | 12:32 |
*** eglynn has joined #openstack-nova | 12:32 | |
johnthetubaguy | pkholkin: maybe a new method called slave_something_something, so its explicit, is possible? | 12:32 |
pkholkin | johnthetubaguy: we are thinking about it right now) | 12:33 |
tdurakov | johnthetubaguy, do we have some criteria that job works? stats maybe? | 12:33 |
*** Sree has joined #openstack-nova | 12:33 | |
PaulMurray | tdurakov, I'd be inclined to add ceph first - and ideally tests that use nfs and ceph before moving to check | 12:35 |
PaulMurray | as a way to make sure everything is working | 12:35 |
PaulMurray | then move to check, then start increasing coverage | 12:35 |
PaulMurray | does that sound reasonable tdurakov johnthetubaguy ? | 12:35 |
*** Guest43431 has quit IRC | 12:36 | |
*** aix has quit IRC | 12:36 | |
tdurakov | PaulMurray, well, there is no technical blockers to wait for ceph, but I'm ok with finishing all of this | 12:36 |
*** markvoelker has quit IRC | 12:37 | |
*** raildo-afk is now known as raildo | 12:39 | |
*** ihrachys has quit IRC | 12:39 | |
*** ihrachys has joined #openstack-nova | 12:40 | |
PaulMurray | tdurakov, I'd like someone who's gone through adding jobs before to make a suggestion, but that seems to be sensible to me | 12:40 |
PaulMurray | The other idea is to start testing as soon as a test works - which is what you have now, so I can go either way | 12:40 |
johnthetubaguy | tdurakov: not seeming broken, I guess, its depends how new the job is really, and how many resources it would burn | 12:41 |
*** rotbeard has quit IRC | 12:41 | |
johnthetubaguy | PaulMurray: we already have a ceph job, I guess, the NFS job would be new I guess? | 12:41 |
* johnthetubaguy really should get some food, actually goes this time... | 12:41 | |
*** baoli has joined #openstack-nova | 12:41 | |
PaulMurray | tdurakov, your call, what would you prefer? | 12:42 |
tdurakov | johnthetubaguy, it's not separate job for nfs, but separate job for live-migration here is description: https://github.com/openstack/nova/blob/master/nova/tests/live_migration/hooks/run_tests.sh | 12:42 |
PaulMurray | tdurakov, what's in my mind is it would be good not to start using resources on every commit before we are reasonably sure it will work | 12:43 |
tdurakov | PaulMurray, well, there is another way, we could leave it in experimental and make announcement for mailing list | 12:44 |
PaulMurray | tdurakov, and it would be good not see failures | 12:44 |
*** jianghuaw has quit IRC | 12:44 | |
tdurakov | so evereyone who need to check live-migration could type "check experimental" | 12:44 |
PaulMurray | so I think johnthetubaguy's comment about ceph vs nfs is that if there is a ceph job already we might be more sure that adding ceph is unlikely to have surprises | 12:45 |
*** jianghuaw has joined #openstack-nova | 12:45 | |
*** Sree has quit IRC | 12:45 | |
PaulMurray | but adding nfs has value | 12:45 |
tdurakov | unfortunately it's single-node ceph job | 12:45 |
*** Sree has joined #openstack-nova | 12:46 | |
tdurakov | but yes, it's already done, so no surprises should be:) | 12:46 |
*** baoli has quit IRC | 12:46 | |
*** rotbeard has joined #openstack-nova | 12:46 | |
PaulMurray | So if there is a risk that it may start spuriously breaking because we haven't got the configuration quite right first time, it would be better to keep it in experimental | 12:46 |
*** thorst has joined #openstack-nova | 12:46 | |
PaulMurray | (I mean start breaking when you add ceph) | 12:46 |
tdurakov | PaulMurray, ok, let's go with my proposal about about mailing list announcement then? | 12:47 |
PaulMurray | tdurakov, sounds like a good plan | 12:47 |
tdurakov | as folks merge my nfs patch i'll write mail | 12:47 |
tdurakov | call it beta-testing, lol | 12:47 |
*** baoli has joined #openstack-nova | 12:48 | |
PaulMurray | now I'm getting lunch too - its that time around here | 12:48 |
tdurakov | bon appétit | 12:48 |
PaulMurray | tdurakov, BTW - good job so far - I'm very glad to see your progress | 12:49 |
pkholkin | johnthetubaguy: what about some custom decorator in nova? I mean smth like @choose_reader_or_async depending on use_slave | 12:54 |
*** salv-orl_ has joined #openstack-nova | 12:54 | |
*** jianghuaw has quit IRC | 12:55 | |
*** wuhg has joined #openstack-nova | 12:55 | |
*** jwcroppe has joined #openstack-nova | 12:55 | |
*** lucas-dinner is now known as lucas-hungry | 12:56 | |
*** salv-orlando has quit IRC | 12:57 | |
*** signed8bit has joined #openstack-nova | 12:58 | |
openstackgerrit | jichenjc proposed openstack/nova: Change some wording on server_concepts.rst https://review.openstack.org/248063 | 12:58 |
*** jaypipes has joined #openstack-nova | 13:00 | |
*** achanda has quit IRC | 13:01 | |
*** achanda has joined #openstack-nova | 13:02 | |
*** MVenesio has joined #openstack-nova | 13:04 | |
johnthetubaguy | pkholkin: its the approach of using a decorator that I think is problematic, not what the decorator says | 13:08 |
*** tdurakov is now known as tdurakov_afk | 13:08 | |
mdbooth | johnthetubaguy pkholkin: The problem with removing use_slave is that I didn't consider what happens when a request crosses an rpc boundary | 13:08 |
johnthetubaguy | mdbooth: ah, good point, I forgot that important detail! | 13:09 |
mdbooth | The decorator itself is ok, because you can put it in the appropriate scope | 13:09 |
*** salv-orl_ has quit IRC | 13:09 | |
mdbooth | The problem is that if that scope does rpc, the context is lost | 13:09 |
*** salv-orlando has joined #openstack-nova | 13:09 | |
mdbooth | So you still need use_slave to be passed so that the context isn't lost | 13:09 |
*** claudiub has joined #openstack-nova | 13:09 | |
mdbooth | Because we're tracking it in the request context automatically, though, we might still be able to get rid of a bunch of use_slave arguments | 13:10 |
mdbooth | Because we can just pull it out of the request context immediately before making an rpc call | 13:10 |
johnthetubaguy | mdbooth: so the current proposal is OK with that actually, it doesn't cross RPC, but yeah, we can't just store it in the context near where its called because of that | 13:10 |
johnthetubaguy | https://review.openstack.org/#/c/243496/2 | 13:11 |
johnthetubaguy | is the link | 13:11 |
*** tdurakov_afk has quit IRC | 13:12 | |
mdbooth | It occurs to me that there's a potential race here which we thought we'd addressed before | 13:13 |
mdbooth | When using Galera | 13:14 |
mdbooth | If you write data on host a, which goes to write connection db node X | 13:14 |
mdbooth | And then rpc to host b, which reads from db node Y | 13:14 |
mdbooth | The rpc can read read state before it was written on host a | 13:15 |
claudiub | johnthetubaguy: hello. we've talked before about implementing a wait for a neutron event in the hyper-v driver, so it can start the VM after its ports have been bound. You said it can be a specless blueprint, since it is already implemented in libvirt. Here's the blueprint: https://blueprints.launchpad.net/nova/+spec/hyper-v-spawn-on-neutron-event | 13:15 |
*** signed8bit is now known as signed8bit_ZZZzz | 13:15 | |
*** salv-orlando has quit IRC | 13:16 | |
mdbooth | We addressed that in the single node case by caching a write context in the current request context, and ensuring that a subsequent read context would always go to the same db connection | 13:16 |
mdbooth | But I don't think we can do that across rpc | 13:16 |
mdbooth | Incidentally, this problem isn't just theoretical | 13:16 |
johnthetubaguy | claudiub: is that on the list of proposed spec-less blueprints yet? | 13:16 |
mdbooth | It's pretty easy to reproduce | 13:16 |
*** salv-orlando has joined #openstack-nova | 13:16 | |
claudiub | johnthetubaguy: no, sorry. | 13:16 |
*** tdurakov_afk has joined #openstack-nova | 13:16 | |
*** tdurakov_afk has quit IRC | 13:17 | |
claudiub | johnthetubaguy: putting it now. | 13:17 |
*** mkoderer has quit IRC | 13:18 | |
johnthetubaguy | claudiub: cool, thank you | 13:19 |
mdbooth | Multi-master galera is a nightmare for races | 13:19 |
*** sacharya has joined #openstack-nova | 13:20 | |
*** ryandahp_ has joined #openstack-nova | 13:20 | |
pkholkin | mdbooth: Matthew, so what do you think about supporting use_slave? | 13:20 |
mdbooth | pkholkin: I think we have to, because there's no other way to retain that context when doing an rpc. | 13:21 |
mdbooth | We're not using it well right now, but that's not the point. The enginefacade is a single node solution. | 13:22 |
*** mkoderer has joined #openstack-nova | 13:22 | |
mdbooth | We *can* remove use_slave for non-remotable apis | 13:22 |
mdbooth | Because the recipient can decorate the request context immediately, and that will propagate until there's another rpc | 13:23 |
*** rpodolyaka1 has joined #openstack-nova | 13:23 | |
*** ryandahp has quit IRC | 13:24 | |
*** jkraj has quit IRC | 13:24 | |
johnthetubaguy | so there are methods that only get used with use_slave=True | 13:24 |
johnthetubaguy | the proposed decorator approach would work just fine right now | 13:24 |
bauwser | yup | 13:24 |
*** sacharya has quit IRC | 13:24 | |
*** rk4n has quit IRC | 13:24 | |
johnthetubaguy | as the remote method is the object call, and the decorator is in the DB api that is called on the conductor | 13:25 |
*** jkraj has joined #openstack-nova | 13:25 | |
johnthetubaguy | but my worry is that its very unclear in the code that you are accessing a read slave, if we hide use_slave=True | 13:25 |
johnthetubaguy | now we do have the RPC context issue, which constrains the possible alternative approaches we could take | 13:25 |
*** jkraj has quit IRC | 13:26 | |
johnthetubaguy | at least that is what is in my head around this right now | 13:26 |
mdbooth | johnthetubaguy: Anything running under the async_reader context is readonly | 13:26 |
*** abhishekk has quit IRC | 13:26 | |
mdbooth | If it wasn't, it wouldn't work | 13:26 |
mdbooth | So the code just need to return data | 13:26 |
johnthetubaguy | mdbooth: yes, sorry, not sure I understand why thats important? | 13:26 |
mdbooth | It's up to the caller to say they can handle old data | 13:27 |
johnthetubaguy | agreed | 13:27 |
mdbooth | The decorator doesn't need to be in the db/api.py | 13:27 |
mdbooth | It can be in objects/instance.py | 13:27 |
mdbooth | This is one of the great benefits of this patch | 13:27 |
johnthetubaguy | so it could be in the DB API for things that are *always* assumed to be reading from the slave | 13:27 |
johnthetubaguy | which is the current proposal being made | 13:27 |
mdbooth | I don't think the db api should ever assume that | 13:28 |
openstackgerrit | Alexis Lee proposed openstack/nova: Remove SoftDeleteMixin from BandwidthUsage https://review.openstack.org/240361 | 13:28 |
mdbooth | It should use its current context | 13:28 |
mdbooth | If it's async, it will read from async | 13:28 |
johnthetubaguy | so I agree, it shouldn't assume that, hence my push to keep read_slave=True | 13:28 |
*** achanda has quit IRC | 13:28 | |
*** rk4n has joined #openstack-nova | 13:28 | |
mdbooth | read_slave isn't required in the db api, though | 13:28 |
mdbooth | because that's not an rpc call | 13:28 |
johnthetubaguy | agreed | 13:29 |
mdbooth | And the caller can already decorate the current context with async reader if that's what they want | 13:29 |
mdbooth | before calling the db api, that is | 13:29 |
mdbooth | The beauty of that is that it will automatically use the same async connection for all db calls in the current context | 13:30 |
*** aix has joined #openstack-nova | 13:30 | |
mdbooth | Even if the db call is embedded in an object read | 13:30 |
*** dave-mccowan has joined #openstack-nova | 13:30 | |
mdbooth | As long as it wasn't remoted :) | 13:30 |
*** achanda has joined #openstack-nova | 13:31 | |
*** zenoway has quit IRC | 13:31 | |
*** dane-fichter has joined #openstack-nova | 13:33 | |
mdbooth | johnthetubaguy: So let's be specific: _check_instance_build_time | 13:34 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: refines the exceptions raised in the driver https://review.openstack.org/212102 | 13:34 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: adds os-win library https://review.openstack.org/247729 | 13:34 |
johnthetubaguy | mdbooth: so this is the patch pkholkin and I were discussing: https://review.openstack.org/#/c/243496 | 13:34 |
mdbooth | johnthetubaguy: Even better :) | 13:35 |
pkholkin | mdbooth: if I understand you correctly, you want to use reader as default and the decorate method as async if needed in objects, yes? | 13:35 |
*** markvoelker has joined #openstack-nova | 13:35 | |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: removes *Utils modules and unit tests https://review.openstack.org/215917 | 13:35 |
mdbooth | pkholkin: No. I think we would decorate the function as async. If the admin configures async it will be used. If not it will use the reader connection. | 13:36 |
johnthetubaguy | pkholkin: so my issue here is really to keep the objects contract the same, however best we do that, I don't mind too much. | 13:36 |
* mdbooth reads the review in question | 13:36 | |
*** bochi-michael has joined #openstack-nova | 13:36 | |
johnthetubaguy | mdbooth: pkholkin: so I am totally confused now... | 13:36 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Converting nova.virt.hyperv to py3 https://review.openstack.org/232554 | 13:36 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Replaces longs with ints https://review.openstack.org/238239 | 13:37 |
johnthetubaguy | feels like we are talking at cross purposes, maybe lets add comments in the review, and loop back here? | 13:37 |
*** achanda has quit IRC | 13:37 | |
mdbooth | johnthetubaguy: k | 13:37 |
*** zenoway has joined #openstack-nova | 13:38 | |
mdbooth | pkholkin: I think we may hit an invalid assumption in the oslo.db stuff here | 13:38 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Fixes dict keys and items references for Python 3 https://review.openstack.org/232585 | 13:38 |
mdbooth | pkholkin: Which we may have to work round | 13:38 |
*** ociuhandu has joined #openstack-nova | 13:38 | |
pkholkin | mdbooth: the problem in our patch right now is that if slave_connection is configured we will always use it and ignore any developer's use_slave parameter in code | 13:38 |
*** achanda has joined #openstack-nova | 13:38 | |
*** richm has joined #openstack-nova | 13:39 | |
*** tdurakov has joined #openstack-nova | 13:39 | |
mdbooth | Right. So lets get specific again. Lets consider _check_instance_build_time in compute manager | 13:39 |
mdbooth | pkholkin: ^^^ | 13:39 |
mdbooth | I believe that function should be decorated with async | 13:40 |
mdbooth | Because that's where that context is relevant | 13:40 |
mdbooth | It then does: | 13:40 |
mdbooth | building_insts = objects.InstanceList.get_by_filters(context, | 13:40 |
mdbooth | filters, expected_attrs=[], use_slave=True) | 13:40 |
*** leseb has quit IRC | 13:40 | |
mdbooth | That use_slave is redundant, because it's already there in the context | 13:40 |
*** ihrachys has quit IRC | 13:40 | |
mdbooth | However, that function is presumably also remotable | 13:41 |
mdbooth | If we assume that it gets rpc'd to conductor, conductor needs a way to know that the context is async | 13:41 |
mdbooth | So we need to put it back for the rpc call. | 13:42 |
mdbooth | Thinking on my feet, it might actually be possible to do that automatically in the remotable decorator... | 13:42 |
mdbooth | Anyway, when it crosses the wire, use_slave needs to be there | 13:42 |
mdbooth | When that results in a call into the db api, it should use the current context. If that's async it should use that. | 13:43 |
*** edleafe is now known as figleaf | 13:43 | |
lxsli | Linked http://docs.openstack.org/developer/nova/code-review.html on https://wiki.openstack.org/wiki/How_To_Contribute#Reviewing and https://wiki.openstack.org/wiki/Nova/CoreTeam#Review_Expectations | 13:43 |
*** klkumar has quit IRC | 13:43 | |
*** Piet has quit IRC | 13:44 | |
mdbooth | The problem I think we'll hit is that iirc we assume a hierarchy of writer->reader->async reader | 13:44 |
raildo | hey guys, me and some guys are working in the nested quota driver (https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/nested-quota-driver-api,n,z) in addition, we want start discuss the re-design of the quota implementation on nova and in other projects, like cinder and neutron. I now that we already have a base spec for this here in nova: https://review.openstack.org/#/c/182445/4/specs/backlog/approved/quotas- | 13:44 |
raildo | reimagined.rst | 13:44 |
raildo | *https://review.openstack.org/#/c/182445/4/specs/backlog/approved/quotas-reimagined.rst | 13:44 |
mdbooth | So I have a sneaking suspicion that if an async context hits @reader it will raise an exception | 13:44 |
raildo | So was I thinking on reate a subteam to speed up the code review in the nested quota implementation and discuss this re-design of quotas. Someone have interest to participate in this subteam? | 13:44 |
pkholkin | mdbooth: yes it will be | 13:45 |
rpodolyaka1 | it will | 13:45 |
mdbooth | If so, we'll need to either fix that, or introduce a new decorator | 13:45 |
mdbooth | which checks to see if something is async already, and makes it reader only if it isn't | 13:45 |
pkholkin | https://github.com/openstack/oslo.db/blob/master/oslo_db/sqlalchemy/enginefacade.py#L560-L561 | 13:46 |
rpodolyaka1 | maybe decorators just don't make sense at db/sqlalchemy/api level? | 13:46 |
rpodolyaka1 | as we don't have enough information there | 13:46 |
rpodolyaka1 | if it should be async vs reader | 13:46 |
mdbooth | rpodolyaka: I think you're right | 13:46 |
mdbooth | We obviously know reader vs writer | 13:46 |
rpodolyaka1 | yeah | 13:46 |
mdbooth | But reader vs async reader is a higher context thing | 13:46 |
rpodolyaka1 | so the only problem with that is that objects level will know about the implementation details, i.e. sqlalchemy based db api impl | 13:47 |
mdbooth | Hehe | 13:48 |
*** tdurakov has quit IRC | 13:48 | |
johnthetubaguy | rpodolyaka1: mdbooth: thats my main complaint, async vs reader should not be hidden down at the DB API layer | 13:48 |
mdbooth | We should totally ditch the notion that we might replace sqlalchemy :) | 13:48 |
rpodolyaka1 | heh | 13:48 |
*** ihrachys has joined #openstack-nova | 13:48 | |
rpodolyaka1 | I thought, rax or someone else had their own implementation | 13:48 |
alex_xu | bauwser: actually disk_over_commit is removed by microverions, so I think we needn't deprecated it first | 13:48 |
johnthetubaguy | rpodolyaka1: use_slave is fairly abstract though, in concept, I wish it was allow_update_lag or something | 13:49 |
mdbooth | johnthetubaguy: Yup. This is actually *better* at that. | 13:49 |
rpodolyaka1 | agreed | 13:49 |
mdbooth | Because you can create a db context at any scope | 13:49 |
mdbooth | It's not confined to the db api | 13:49 |
johnthetubaguy | rpodolyaka1: not really, for nova-cells workers a few calls have been re-implemented to speed it up, but mostly because we don't use pymysql | 13:49 |
alex_xu | bauwser: the disk_over_commit is almost at here very long time in old version | 13:49 |
rpodolyaka1 | johnthetubaguy: ack, I must have confused rax with someone else then | 13:50 |
*** ankit_ag has quit IRC | 13:50 | |
bauwser | alex_xu: I'd prefer a more gentle approach where we keep it optional but raise a warning if someone uses it | 13:51 |
*** rk4n has quit IRC | 13:51 | |
bauwser | alex_xu: I don't want to have the operators thrown under the bus because of that | 13:51 |
bauwser | tbh, I don't feel the warning right, but a relnote is certainly good | 13:51 |
bauwser | alex_xu: for sure we could say "eh, you just have to ask for an earlier microversion" but that's a bit bad IMHO - because what if they want a bugfix ? | 13:51 |
bauwser | alex_xu: sure but imagine the following case : disk_over_commit is removed by v2.25 but a new feature or bugfix is modifying an API by v2.26 | 13:52 |
*** rk4n has joined #openstack-nova | 13:52 | |
bauwser | alex_xu: is this fromage or desert? | 13:52 |
rpodolyaka1 | mdbooth: ok, so looks like we'll need something similar to async/reader/writer in db/api and allow to use that up the stack at objects level | 13:52 |
bauwser | oops | 13:52 |
bauwser | cheese or desert I mean | 13:52 |
*** salv-orlando has quit IRC | 13:52 | |
*** markus_z_bnc has quit IRC | 13:52 | |
rpodolyaka1 | mdbooth: I'm just a bit confused, why that wouldn't work with remote calls? | 13:52 |
*** salv-orl_ has joined #openstack-nova | 13:52 | |
* bauwser not sure that this expression is well known abroad | 13:52 | |
alex_xu | bauwser: emm...probably more clear your point | 13:53 |
*** Sree has quit IRC | 13:53 | |
mdbooth | rpodolyaka: Because the db context isn't serialised on the wire | 13:53 |
mdbooth | The remote end doesn't have the db context of the caller | 13:53 |
*** lucas-hungry is now known as lucasagomes | 13:53 | |
mdbooth | It'll have to be explicit | 13:53 |
rpodolyaka1 | aha | 13:53 |
bauwser | alex_xu: so, sure the operator can get the option by calling an earlier version, right? | 13:53 |
*** tdurakov has joined #openstack-nova | 13:54 | |
rpodolyaka1 | so, maybe leave use_salve in rpc calls | 13:54 |
rpodolyaka1 | and then use context managers? | 13:54 |
mdbooth | Yup | 13:54 |
rpodolyaka1 | on the remote end | 13:54 |
alex_xu | let me think more minutes | 13:54 |
mdbooth | Yes. The receiver can just set the context according to what it received | 13:54 |
*** annegentle has joined #openstack-nova | 13:54 | |
rpodolyaka1 | aha, makes sense now, thanks! | 13:54 |
*** klkumar has joined #openstack-nova | 13:54 | |
alex_xu | bauwser: yes, and that old behavior will keep at here | 13:54 |
*** ihrachys has quit IRC | 13:55 | |
bauwser | alex_xu: okay, so they will be stuck with that earlier version if they want to use disk_over_commit, right? | 13:55 |
*** paul-carlton has quit IRC | 13:55 | |
*** jianghuaw has joined #openstack-nova | 13:55 | |
alex_xu | bauwser: yes | 13:56 |
*** claudiub has quit IRC | 13:57 | |
alex_xu | bauwser: just to be clear, you expect RT can provide similar functionality in the future, right? | 13:57 |
bauwser | alex_xu: before talking about the implementation, I'm just trying to explain why I want to deprecate :) | 13:58 |
*** jaypipes has quit IRC | 13:59 | |
bauwser | alex_xu: so if the operator is stuck with an earlier version, how can he get the latest API features, given it's monotonic? | 13:59 |
*** jianghuaw has quit IRC | 13:59 | |
alex_xu | bauwser: if operator want to upgrade his client to support latest api, he need prepare all the things he need change, including disk_overcommit get rid of. | 14:00 |
bauwser | alex_xu: to be clear, I'm thinking of a 'grace period' where we could admit that we made something wrong but that it would need time to implement the feature differently before removing the posibility to use it | 14:00 |
bauwser | alex_xu: I think I see your point, what just worries me is that we remove a feature without waiting a deprecation cycle | 14:00 |
bauwser | (to summarize) | 14:01 |
bauwser | alex_xu: that's not like if we change the API contract but we're still providing the possibility to count against real disk usage | 14:01 |
*** rlrossit has joined #openstack-nova | 14:01 | |
*** klkumar has quit IRC | 14:02 | |
alex_xu | bauwser: yea, I see your point. I'm also think about we already have microversion that is used to advert api change, why we still need other way to deprecated feature | 14:02 |
*** ihrachys has joined #openstack-nova | 14:02 | |
bauwser | alex_xu: it's your expertise, so lemme ask you a question | 14:02 |
alex_xu | bauwser: but API contract changed, something means the feature changed, like add feature and remove feature | 14:02 |
bauwser | alex_xu: did you already bump a microversion for removing a feature without providing another way to get that feature? | 14:02 |
bauwser | not speaking of the EC2 endpoint for example | 14:03 |
bauwser | just microversions | 14:03 |
*** alex_klimov has quit IRC | 14:03 | |
alex_xu | bauwser: at least for now, we don't have such case | 14:03 |
bauwser | alex_xu: tbc, I'm not particularly concerned by that | 14:03 |
bauwser | alex_xu: I just want to make sure it's something agreed | 14:04 |
alex_xu | bauwser: yea, I see | 14:05 |
*** klkumar has joined #openstack-nova | 14:05 | |
bauwser | alex_xu: maybe I'm overthinking tho :) | 14:05 |
*** zhenq has joined #openstack-nova | 14:06 | |
alex_xu | bauwser: so what is deprecated propose for? deprecated is for admin after upgrade his code, his tools or client won't break, and get chance find something instead | 14:06 |
bauwser | yeah that | 14:06 |
bauwser | a migration path | 14:07 |
alex_xu | bauwser: emm...next thing is think about, whether microversion provide that. | 14:07 |
bauwser | alex_xu: so, my take is that it's just a matter of providing a release note | 14:08 |
bauwser | alex_xu: you keep the attribute but you make it optional | 14:08 |
alex_xu | microversion is different, microversion tell people there are something removed, but you still can use old version before you are ready | 14:08 |
*** ildikov has quit IRC | 14:08 | |
*** ijuwang has joined #openstack-nova | 14:08 | |
bauwser | alex_xu: so that when the new cycle begins, you provide a microversion for removing that attribute | 14:08 |
bauwser | alex_xu: a-ha I see | 14:08 |
*** claudiub has joined #openstack-nova | 14:09 | |
bauwser | well, if you feel microversioning the removal is fine, then I'm okay :) | 14:09 |
*** mjura has quit IRC | 14:10 | |
*** klkumar has quit IRC | 14:11 | |
*** achanda has quit IRC | 14:11 | |
*** ildikov has joined #openstack-nova | 14:12 | |
alex_xu | bauwser: after thinking, yea, I think it is ok. And the new feature is thing to encourage people upgrade to deprecated the old api. | 14:12 |
*** stevemar_ has joined #openstack-nova | 14:13 | |
*** achanda has joined #openstack-nova | 14:13 | |
*** eharney has joined #openstack-nova | 14:15 | |
*** signed8bit_ZZZzz is now known as signed8bit | 14:16 | |
*** mriedem has joined #openstack-nova | 14:16 | |
*** salv-orl_ has quit IRC | 14:17 | |
*** signed8bit is now known as signed8bit_ZZZzz | 14:17 | |
*** mdrabe has joined #openstack-nova | 14:18 | |
bauwser | alex_xu: okay, you mean we can go straight and just remove the attribute in a microversion ? fair enough then | 14:20 |
*** salv-orlando has joined #openstack-nova | 14:21 | |
alex_xu | bauwser: yea, that is my understand microversion is used to | 14:21 |
*** jerrygb has joined #openstack-nova | 14:21 | |
*** doug-fish has joined #openstack-nova | 14:21 | |
*** otter768 has joined #openstack-nova | 14:21 | |
*** jichen has quit IRC | 14:22 | |
*** burt has joined #openstack-nova | 14:22 | |
alex_xu | PaulMurray: just saw your commit, do you mean hp need that feature? or that is a feature can be depreacted by microversion? | 14:23 |
*** signed8bit_ZZZzz is now known as signed8bit | 14:23 | |
*** mrkz has joined #openstack-nova | 14:24 | |
alex_xu | maybe I should bring this up to livemigration weekly meeting | 14:25 |
*** klkumar has joined #openstack-nova | 14:26 | |
*** otter768 has quit IRC | 14:26 | |
*** ihrachys has quit IRC | 14:26 | |
*** artom has joined #openstack-nova | 14:26 | |
*** achanda has quit IRC | 14:26 | |
*** rubasov has quit IRC | 14:29 | |
*** Piet has joined #openstack-nova | 14:29 | |
mriedem | bauwser: i'm spilling the reno madness over into other projects now :) https://review.openstack.org/#/c/247906/ | 14:30 |
bauwser | \o/ | 14:30 |
bauwser | mriedem: FYI, I just saw the current relnotes in a draft build, and I'm a bit sad about the current prelude section, so I'll also try to update the YAML files next week | 14:31 |
bauwser | for the moment, trying to get my brain focused on specs more than 10 mins | 14:32 |
*** ihrachys has joined #openstack-nova | 14:32 | |
mriedem | good luck | 14:32 |
mriedem | i do spec reviews at night | 14:32 |
bauwser | I know | 14:32 |
mriedem | and you work at night :) | 14:32 |
bauwser | I'd rather say I work at Australia morning | 14:32 |
*** jianghuaw has joined #openstack-nova | 14:32 | |
*** jianghuaw has quit IRC | 14:33 | |
*** daemontool has joined #openstack-nova | 14:35 | |
*** rpodolyaka1 has quit IRC | 14:35 | |
*** smatzek has quit IRC | 14:35 | |
*** akshai has joined #openstack-nova | 14:35 | |
*** tdurakov has quit IRC | 14:37 | |
*** rpodolyaka1 has joined #openstack-nova | 14:37 | |
*** daemontool_ has quit IRC | 14:37 | |
*** annegentle has quit IRC | 14:38 | |
*** tdurakov has joined #openstack-nova | 14:38 | |
mriedem | superdan?! | 14:41 |
*** fawadkhaliq has joined #openstack-nova | 14:41 | |
dansmith | omg is it friday? | 14:41 |
*** dansmith is now known as superdan | 14:41 | |
mriedem | it's tgif | 14:41 |
sdague | johnthetubaguy: why did you want this as an action instead of a resource - https://review.openstack.org/#/c/228828/ ? | 14:43 |
*** paul-carlton has joined #openstack-nova | 14:43 | |
sdague | on the live migrations cancel | 14:43 |
*** achanda has joined #openstack-nova | 14:44 | |
*** chinmay_ has quit IRC | 14:45 | |
*** burgerk has joined #openstack-nova | 14:46 | |
*** xyang1 has joined #openstack-nova | 14:46 | |
*** cfriesen__ has joined #openstack-nova | 14:46 | |
*** aix has quit IRC | 14:47 | |
*** aix has joined #openstack-nova | 14:49 | |
*** ctrath has joined #openstack-nova | 14:49 | |
PaulMurray | alex_xu, do you mean the comment about disk_over_commit ? | 14:50 |
*** paul-carlton1 has joined #openstack-nova | 14:50 | |
*** ihrachys has quit IRC | 14:51 | |
*** ihrachys has joined #openstack-nova | 14:54 | |
*** paul-carlton has left #openstack-nova | 14:54 | |
*** paul-carlton has joined #openstack-nova | 14:54 | |
*** xek has quit IRC | 14:55 | |
*** nkrinner has quit IRC | 14:55 | |
openstackgerrit | Balazs Gibizer proposed openstack/nova: Handle race in allocate_for_instance https://review.openstack.org/221803 | 14:55 |
*** paul-carlton has left #openstack-nova | 14:55 | |
*** gwei3 has joined #openstack-nova | 14:56 | |
*** pumaranikar has joined #openstack-nova | 14:56 | |
*** ccarmack has joined #openstack-nova | 14:57 | |
*** vladikr has joined #openstack-nova | 14:58 | |
*** mhickey has joined #openstack-nova | 14:58 | |
openstackgerrit | Yingxin Cheng proposed openstack/nova: Use stevedore for scheduler host manager https://review.openstack.org/246476 | 14:58 |
*** markus_z_meeting is now known as markus_z | 14:59 | |
*** paul-carlton2 has joined #openstack-nova | 14:59 | |
johnthetubaguy | sdague: good question, I think its that I wanted to drop the migration endpoint really | 14:59 |
dims | mhickey : can you please paste a github link to the deprecated opt and we can ask folks here? | 14:59 |
sdague | johnthetubaguy: right, that's fine, but as a sub resource on servers this seems to make more sense | 15:00 |
sdague | I just hate openning up another action if there is a restful way to model this | 15:00 |
*** mpavone has quit IRC | 15:01 | |
*** boris-42 has joined #openstack-nova | 15:01 | |
mhickey | dims: Sure. Here it is: https://github.com/openstack/nova/blob/master/nova/network/neutronv2/api.py#L132 | 15:01 |
*** mgoddard_ has joined #openstack-nova | 15:02 | |
dims | johnthetubaguy : when do we clean up deprecated options? like the one above ^^^ | 15:02 |
johnthetubaguy | sdague: true | 15:02 |
*** kfarr has joined #openstack-nova | 15:02 | |
johnthetubaguy | dims: depends when we deprecated them | 15:02 |
*** breitz has joined #openstack-nova | 15:03 | |
johnthetubaguy | sdague: my thinking we add the action, so it matches everything else for now, while we sort out tasks more generally | 15:03 |
dims | johnthetubaguy : "Aug 31, 2014" | 15:03 |
sdague | johnthetubaguy: then we have to support the action largely forever | 15:03 |
sdague | dims: + 3 months | 15:03 |
johnthetubaguy | dims: so we can do that last release then, I guess | 15:03 |
sdague | oh, 2014, | 15:03 |
dims | sdague :) | 15:04 |
johnthetubaguy | sdague: yeah, we do, but at least is consistently terrible as other things, rather than a new thing on the side, that gets turned into another new ish thing later? | 15:05 |
sdague | johnthetubaguy: I really think we should avoid the actions interface whenever possible | 15:05 |
johnthetubaguy | sdague: maybe I am over thinking this | 15:05 |
*** smatzek has joined #openstack-nova | 15:05 | |
sdague | you don't think we'd continue to want to have access to the migrations this way? | 15:05 |
*** mgoddard has quit IRC | 15:05 | |
sdague | also, at this point I consider tasks to be duke nukem 3 | 15:05 |
sdague | I don't believe it's happening in any reasonable time frame | 15:06 |
johnthetubaguy | so we have instance actions for some things and migrations for other things | 15:06 |
johnthetubaguy | now thats not quite the same thing, granted | 15:06 |
johnthetubaguy | but its a way to get status on an async task | 15:06 |
sdague | a migration is a subresource of a server | 15:07 |
sdague | we model it that way in our database even | 15:07 |
johnthetubaguy | http://developer.openstack.org/api-ref-compute-v2.1.html#os-migrations-v2.1 | 15:08 |
sdague | right, the search is not | 15:08 |
*** jasondotstar is now known as jasondotstar_afk | 15:08 | |
sdague | I don't want to add it to that resource | 15:08 |
johnthetubaguy | ah, so thats the proposal I was against | 15:09 |
johnthetubaguy | at least in my head | 15:09 |
sdague | I want that resource to return search results which include links to /server/{id}/migrations/{mid} | 15:09 |
sdague | and DELETE / GET on /server/{id}/migrations/{mid} is a thing | 15:09 |
paul-carlton2 | the problem as I see it is that the migrations list returns migration ids, we'd need to convert this to migration uuid to move forward with this plan | 15:10 |
johnthetubaguy | oh, didn't know that existed, my bad | 15:10 |
*** stevemar_ has quit IRC | 15:10 | |
sdague | johnthetubaguy: it doesn't | 15:10 |
openstackgerrit | Pavel Kholkin proposed openstack/nova: enginefacade: 'migration' https://review.openstack.org/243496 | 15:10 |
johnthetubaguy | oh, now I get you | 15:10 |
sdague | that's the way I think this API should work though, instead of a new action | 15:10 |
*** yonglihe has joined #openstack-nova | 15:10 | |
johnthetubaguy | what about GET on /server/{uuid} | 15:11 |
johnthetubaguy | arg | 15:11 |
sdague | paul-carlton2: why would we need to change the id? | 15:11 |
paul-carlton2 | I dropped the idea of doing cancel migration on migration object when https://review.openstack.org/#/c/243587/ got rejected | 15:11 |
johnthetubaguy | GET on /server/{uuid}/action/{request-id} | 15:11 |
*** e0ne has quit IRC | 15:11 | |
paul-carlton2 | We hse uuids everywhere else so I assumed we'd want to do that for migrations too | 15:12 |
*** haomaiwa_ has quit IRC | 15:12 | |
sdague | paul-carlton2: we use them because of potential uniqueness constraints, there is no reason why you'd have to change it here | 15:12 |
sdague | johnthetubaguy: and GET /server/{uuid}/actions returns what? | 15:13 |
johnthetubaguy | basically you get something like this, but complete: /v2.1/​{tenant_id}​/servers/​{server_id}​/os-instance-actions | 15:13 |
johnthetubaguy | a record for every action call on the instance, and its current status | 15:13 |
sdague | ok, so now this is basically tasks light | 15:14 |
johnthetubaguy | live-migrate, snapshots, soft-reboot, can then all get cancelled in a similar way | 15:14 |
johnthetubaguy | I suppose, yeah | 15:14 |
sdague | ok, so what is the objection to exposing /server/{id}/migrations long term? | 15:15 |
sdague | I guess that's the resistance I don't understand | 15:15 |
paul-carlton2 | so I'd get this back | 15:15 |
paul-carlton2 | RESP BODY: {"instanceActions": [{"instance_uuid": "9d266594-5f13-43b9-b9dd-28c8ee92a534", "user_id": "5bf0029a20f24149810cafed6d820ca7", "start_time": "2015-11-19T07:24:18.000000", "request_id": "req-7bd61401-433f-48cc-9419-a9d32a8fd5de", "action": "live-migration", "message": null, "project_id": "bf941e97250e49cab47d07767c62c15a"}, {"instance_uuid": "9d266594-5f13-43b9-b9dd-28c8ee92a534", "user_id": "0e86d465eadf46 | 15:15 |
johnthetubaguy | sdague: we could, I just slightly prefer a full list of tasks, rather that having to go to different places depending on the action you take | 15:16 |
*** krtaylor has joined #openstack-nova | 15:16 | |
johnthetubaguy | list of actions, I guess I mean | 15:16 |
sdague | johnthetubaguy: ok, but that's the tasks api (more or less), which has been stalled for 2 years | 15:16 |
*** sneti has joined #openstack-nova | 15:16 | |
sdague | also, with instance actions you have the visibility concerns as well right. A normal user probably shouldn't know their instance is getting live migrated | 15:18 |
paul-carlton2 | sdague: The cat is out of the bag when they do get on server, task state of migrating gives it away! | 15:19 |
johnthetubaguy | sdague: so I thought about that more, they actually have to know, because it stops a snapshot being taken, but, pathologically, you sort of don't want to tell them the live-migrate happened if they didn't spot it while it was happening | 15:19 |
johnthetubaguy | paul-carlton2: and that, good point | 15:20 |
sdague | yeh, honestly, I think that if we start overloading instance actions to be tasks light we're just going to end up with gorp that we're all going to regret down the road | 15:20 |
*** rlrossit has quit IRC | 15:20 | |
*** pradk has joined #openstack-nova | 15:21 | |
*** annegentle has joined #openstack-nova | 15:21 | |
sdague | I personally hate actions because all the logging by url becomes useless | 15:21 |
sdague | so you have to build a whole additional toolchain to figure out what's going on postmortem | 15:21 |
lxsli | Have requirements on the tasks API been written up somewhere? | 15:21 |
bauwser | lxsli: on it | 15:21 |
sdague | made worse so by putting things under apache | 15:21 |
bauwser | lxsli: for the moment, we have old stuff and my manifesto | 15:22 |
*** rlrossit has joined #openstack-nova | 15:22 | |
bauwser | lxsli: so I'm writing a Tasks API spec | 15:22 |
lxsli | bauwser: cool cool | 15:22 |
sdague | bauwser: for which I salute you | 15:22 |
bauwser | (keeping in mind the deadline, for sure) | 15:22 |
sdague | however, it's still a long ways off | 15:22 |
*** jerrygb has quit IRC | 15:22 | |
bauwser | sdague: totally agreed | 15:23 |
bauwser | sdague: I don't pretend to boil the ocean | 15:23 |
bauwser | sdague: :) | 15:23 |
*** jerrygb has joined #openstack-nova | 15:23 | |
lxsli | Anyone for https://review.openstack.org/#/c/240361/ please? | 15:23 |
*** pratikmallya has joined #openstack-nova | 15:23 | |
*** kfarr has quit IRC | 15:23 | |
bauwser | so, yeah, it seems there is a consensus on having a Tasks object - that's why I'm writting a spec to discuss that and confirm that | 15:23 |
mriedem | bauwser: hmmm | 15:24 |
mriedem | so, | 15:24 |
paul-carlton2 | We really need to re do every action on an instance as a task but we can't stop all enhancements till task API is done | 15:24 |
bauwser | my personal target for Mitaka is just write the object, that's it | 15:24 |
mriedem | lxsli: if that's never soft deleted, how do we ever purge it? | 15:24 |
bauwser | no further actions | 15:24 |
lxsli | mriedem: huh? | 15:24 |
sdague | lxsli: a table drop like that probably should have a reno | 15:25 |
mriedem | lxsli: see the thread on archive | 15:25 |
bauwser | I mean, I could do some tyromancy to know if I'm right | 15:25 |
bauwser | http://closetprofessor.blogspot.fr/2011/01/word-of-week-tyromancy.html | 15:25 |
mriedem | lxsli: for tables that are never soft deleted, what are we doing to allow operators to actually prune those tables? | 15:25 |
sdague | mriedem: yeh, I don't think this ever deleted | 15:25 |
mriedem | right, like instance_actions | 15:25 |
mriedem | i'm finding out we have more skeletons in the closet than i though | 15:25 |
mriedem | *thought | 15:25 |
lxsli | mriedem: it's hard-deleted sometimes, I think? I'd have to check to be sure | 15:26 |
*** pratikma_ has joined #openstack-nova | 15:26 | |
*** kfarr has joined #openstack-nova | 15:26 | |
mriedem | doubtful | 15:26 |
mriedem | unless there is a task that's purging them after they are so old | 15:26 |
sdague | mriedem: there is no delete interface at all from what I can see | 15:27 |
sdague | this is another one that should be a cascade delete from instances | 15:27 |
lxsli | mriedem: is that really my problem though? Today they're not deleted, after my patch they're still not deleted | 15:27 |
mriedem | lxsli: well, | 15:27 |
mriedem | as is posed in the ML, | 15:28 |
mriedem | the same is true for instance_acitons | 15:28 |
mriedem | and the questoin is, should we start soft deleting those | 15:28 |
mriedem | so the same question applies for this table now | 15:28 |
*** sneti has quit IRC | 15:28 | |
mriedem | bauwser: do you have a link handy to the ML thread where dhellmann talks about integrating reno with each project? | 15:28 |
*** ctrath has quit IRC | 15:28 | |
sdague | mriedem: I actually don't think we should softdelete subresources of instances | 15:28 |
*** pratikmallya has quit IRC | 15:28 | |
sdague | we should delete them when archiving instances | 15:29 |
mriedem | sdague: that was one of the options i laid out in the ML | 15:29 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: refines the exceptions raised in the driver https://review.openstack.org/212102 | 15:29 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: adds os-win library https://review.openstack.org/247729 | 15:29 |
mriedem | and is the easier to implement | 15:29 |
sdague | mriedem: yeh, I responded on ML with basically that this morning as well | 15:29 |
bauwser | mriedem: sure, sec | 15:29 |
mriedem | ah http://lists.openstack.org/pipermail/openstack-dev/2015-November/078301.html | 15:30 |
sdague | mriedem: or, archive them with the instances | 15:30 |
mriedem | just found it | 15:30 |
bauwser | mriedem: http://lists.openstack.org/pipermail/openstack-dev/2015-November/078301.html | 15:30 |
sdague | instance_actions honestly is probably useful in the archive | 15:30 |
bauwser | busted AGAIN | 15:30 |
sdague | this table is definitely not | 15:30 |
*** ihrachys has quit IRC | 15:30 | |
bauwser | come'on ! | 15:30 |
mriedem | bauwser: thanks though, i've been looking for it for like 20 minutes | 15:30 |
bauwser | mriedem: okay, ping me anytime you look for your keys or something like that, it seems my call makes you find it faster | 15:31 |
mriedem | ha | 15:31 |
mriedem | i don't lose things at home b/c i use a system, like a sane OCD person | 15:31 |
*** ihrachys has joined #openstack-nova | 15:31 | |
mriedem | unlike my wife that leaves things where she last used them | 15:31 |
mriedem | so we have 20 bottles of glass cleaner stashed around the house | 15:31 |
*** ctrath has joined #openstack-nova | 15:31 | |
*** mc_nair has joined #openstack-nova | 15:32 | |
sdague | mriedem: it's high availability | 15:32 |
mriedem | it's maddening | 15:32 |
*** nikhil has quit IRC | 15:32 | |
mriedem | costco doesn't help because you can buy glass cleaner in bulk | 15:32 |
*** nikhil has joined #openstack-nova | 15:32 | |
*** thumpba has joined #openstack-nova | 15:33 | |
*** annegentle has quit IRC | 15:33 | |
*** alexschm has quit IRC | 15:33 | |
*** e0ne has joined #openstack-nova | 15:33 | |
lxsli | sdague: I'm probably being dumb but I can't see a relation between BandwidthUsage and Instance from looking at db/sqlalchemy/models.py ? Can you explain please? | 15:33 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: removes *Utils modules and unit tests https://review.openstack.org/215917 | 15:34 |
mriedem | lxsli: there isn't one | 15:34 |
lxsli | mriedem: then how is this class like instance_actions? | 15:34 |
mriedem | it's just a periodic task that is only enabled for xenserver to poll stats from the hypervisor | 15:34 |
*** rcernin has quit IRC | 15:34 | |
*** tdurakov has quit IRC | 15:34 | |
mriedem | lxsli: in that it's not ever deleted | 15:34 |
*** gwei3 has quit IRC | 15:34 | |
*** thangp has joined #openstack-nova | 15:35 | |
mriedem | lxsli: the bigger picture is we are gd pack rats | 15:35 |
mriedem | archive was meant to help with part of that, but we're finding we have other tables that we don't soft delete so archive won't archive them | 15:35 |
*** yamahata has joined #openstack-nova | 15:35 | |
mriedem | which is also why we have ctrath's purge command work | 15:35 |
lxsli | mriedem: right, so the question is whether removing SoftDeleteMixin should be blocked behind having a purge strategy | 15:35 |
sdague | it's looking up by instance mac address, right? | 15:35 |
*** hparekh has joined #openstack-nova | 15:35 | |
mriedem | but the purge command only purges soft deleted things | 15:35 |
*** nithyag_ has quit IRC | 15:36 | |
sdague | https://github.com/openstack/nova/blob/1734ce7101982dd95f8fab1ab4815bd258a33744/nova/notifications.py#L315-L327 | 15:36 |
sdague | the uuid in the table is the instance uuid | 15:36 |
garyk1 | mriedem: https://review.openstack.org/#/c/247705/ | 15:36 |
mriedem | with no foreign key? | 15:36 |
mriedem | nice | 15:36 |
sdague | mriedem: remember, all the db experts said foreign keys are evil | 15:37 |
lxsli | hmmm... I might add a foreign key and pick an easier class to start with then | 15:37 |
*** sneti has joined #openstack-nova | 15:37 | |
lxsli | o.O | 15:37 |
mriedem | sdague: what's more evil is being completely incosistent about using them | 15:37 |
sdague | mriedem: sure, I just never understood the fk are evil standpoint myself :) | 15:37 |
sdague | maybe we can throw mordred under a bus on that one again | 15:37 |
mordred | aroo? | 15:38 |
mordred | ok. fk's are not scalable - and you can't shard vertically when you use db fks | 15:38 |
mordred | app level fks are great | 15:38 |
sdague | sure, we don't shard vertically anyway | 15:38 |
mordred | right | 15:38 |
*** tdurakov has joined #openstack-nova | 15:38 | |
mordred | but as people talk abot GIANT SCALE things, one of the ways to alleviate db scaling is by sharding | 15:39 |
sdague | app level fks mean you have to implement cascade delete manually | 15:39 |
mordred | which fks remove as an option | 15:39 |
mordred | and yes. they do mean that | 15:39 |
sdague | mordred: ok, it still seems like very premature scaling optimization | 15:39 |
openstackgerrit | Pavel Kholkin proposed openstack/nova: enginefacade: 'migration' https://review.openstack.org/243496 | 15:40 |
mordred | sdague: we have a rabbitmq in our architecture | 15:40 |
mordred | sdague: that was a premature scaling optimization | 15:40 |
sdague | mordred: sure | 15:40 |
mordred | and is also a scaling bottleneck | 15:40 |
sdague | but lets not do more of them | 15:40 |
mordred | right | 15:40 |
sdague | because this has created an actual operator problem | 15:40 |
mordred | really? | 15:40 |
sdague | in their db filling up with crap | 15:40 |
mriedem | sdague: ftr, we don't implement cascading delete automatically in our db schema either :/ | 15:41 |
sdague | mriedem: I know | 15:41 |
sdague | but a bunch of these would be well served by that | 15:41 |
mriedem | we implement pain | 15:41 |
mriedem | PaaS | 15:41 |
mordred | well, it would be served well by something implementing delete | 15:41 |
*** ihrachys has quit IRC | 15:41 | |
mordred | that could be implemented as a FK with cascading delete | 15:41 |
mordred | or it could just be implemeted at the data model layer | 15:41 |
openstackgerrit | Andrey Kurilin proposed openstack/nova-specs: Update novaclient-api-microversions spec https://review.openstack.org/211206 | 15:42 |
openstackgerrit | Andrey Kurilin proposed openstack/nova-specs: Move novaclient-api-microversions spec to implemented https://review.openstack.org/248142 | 15:42 |
mordred | which, btw, is in python | 15:42 |
mordred | and given how little everyone around here claims to know sql - I'd recommend locating the logic in python personally | 15:42 |
sdague | yes, but now needs to handle all the transaction logic | 15:42 |
mordred | yeah. "begin transaction. delete tihng. delete second thing. commit" | 15:42 |
mordred | that should be a delete method on the data model | 15:43 |
*** ihrachys has joined #openstack-nova | 15:43 | |
sdague | also lock the table to ensure no new rows show up that could break the constraint | 15:43 |
mordred | nope | 15:43 |
mordred | you do not need to do that | 15:43 |
mordred | and should not | 15:43 |
mordred | the MVCC support in innodb does the right thing quite happily | 15:44 |
sdague | if the db has no constraints on it, how do you prevent that coming in on a different worker? | 15:44 |
*** mgoddard_ has quit IRC | 15:44 | |
*** mgoddard has joined #openstack-nova | 15:44 | |
*** MVenesio has quit IRC | 15:45 | |
pkholkin | johnthetubaguy, mdbooth: guys, take a look please https://review.openstack.org/#/c/243496/, fixed this patch as an example | 15:46 |
mordred | oh - sorry - a row being added that references a thing that the transaction is deleting? yeah, you still have to protect for that app side, but you can do it without a table lock | 15:46 |
*** ihrachys has quit IRC | 15:46 | |
pkholkin | leave your comments and questions there | 15:47 |
mordred | (basically, ANY time the words "table lock" are uttered, it's the wrong approach) | 15:47 |
*** shaohe_feng1 has joined #openstack-nova | 15:47 | |
sdague | mordred: sure, but now you have to synchronize across API servers | 15:47 |
*** ihrachys has joined #openstack-nova | 15:47 | |
openstackgerrit | Claudiu Belu proposed openstack/nova: Converting nova.virt.hyperv to py3 https://review.openstack.org/232554 | 15:47 |
mordred | no you don't | 15:47 |
sdague | mordred: ok | 15:47 |
sdague | explain please | 15:47 |
mriedem | didn't people talk about DLM at the summit? | 15:47 |
mdbooth | pkholkin: I think you need a decorator | 15:48 |
mriedem | i heard we're converting to java? | 15:48 |
mdbooth | I was thinking that the new decorator would replace *all* reader decorators | 15:48 |
mdbooth | Which is why we might be better of fixing it in oslo.db, tbh | 15:48 |
sdague | API1 starts this transaction, res1 plus subres1,2,3,4 being deleted. API2 gets a request to add subres5 in the middle of it. | 15:48 |
mdbooth | Although an expedient fix here would be fine | 15:48 |
mordred | sdague: you have the add record logic check the referant after the insert, and if it's been invalidated, you delete the record and send back a fail | 15:48 |
mordred | sdague: in either case it's a thing trying to add a record that references something someone else is trying to delete | 15:49 |
sdague | mordred: right | 15:49 |
mordred | so it's an application logic race condition | 15:49 |
mordred | and has to be dealt with in either case | 15:49 |
mordred | in terms of expectations of the consumes | 15:49 |
mordred | consumers | 15:49 |
sdague | so without the constraint, the delete succeeds, you don't know anything is wrong. If the delete finishes before the insert, you can fail the insert | 15:50 |
sdague | if the insert gets in first it looks successufl | 15:50 |
*** ihrachys has quit IRC | 15:50 | |
sdague | so you need to check again on the delete that life is really actually good | 15:50 |
sdague | and stop or fail or leak | 15:51 |
mdbooth | As a general design principal, if a compute host knows it can side-load an image from another compute host, should it do that, or go straight to glance anyway? | 15:51 |
mordred | yup. | 15:51 |
mdbooth | That is, both options are open, which is better? | 15:51 |
openstackgerrit | Alexis Lee proposed openstack/nova: Note that BandwidthUsage relates to Instance https://review.openstack.org/248146 | 15:51 |
mordred | it is more logic in the application | 15:51 |
mordred | but the application is already scale out | 15:51 |
sdague | I guess I'm just old school and like dbs that you can't fill with crud | 15:51 |
mdbooth | Getting it from glance seems cleaner, but side-loading seems like it scales better. | 15:51 |
mordred | right | 15:51 |
mordred | but that's the old-school oracle/db model where you have a GIANT machine your database is on and there are no apps that will ever exceed the dbs capabilities | 15:52 |
sdague | assuming the db is the bottleneck | 15:52 |
mordred | everything logic-wise emantes from the db, and in fact stored procedures are used to further put the model layer directly in the db | 15:52 |
mriedem | sdague: replied to the ML, i'm basically in agreement. we archive instance_actions when archiving instances - it's a yucky side path but it is what it is, it's the right thing to do for postmortems. and i'll push up a spec for making os-instance-actions read deleted instances | 15:52 |
*** ihrachys has joined #openstack-nova | 15:52 | |
*** stevemar_ has joined #openstack-nova | 15:52 | |
mordred | sdague: the DB is always going to be the bottleneck ultimately in any scale out application | 15:52 |
sdague | mordred: ok | 15:53 |
mordred | or, rather, "the persistance layer" - because it's the thing that's constrained by the physics of physical media | 15:53 |
pkholkin | mdbooth: I think that integration with use_slave should be on nova side, oslo.db doesn't know anything about use_slave | 15:53 |
mordred | at some point _something_ has to write the data to an actual local and then guarantee that it's there | 15:53 |
pkholkin | could you please leave comments in the patch | 15:53 |
mdbooth | pkholkin: Yes, I mean sync/async | 15:53 |
mdbooth | pkholkin: Ok | 15:54 |
pkholkin | or maybe give some your example | 15:54 |
mordred | that's the part that mongo played fast and loose with - although they're doing much better with that these days | 15:54 |
pkholkin | we will continue working on it | 15:54 |
sdague | mordred: sure, it just seems that there are many many many layers of bottlenecks before we get there | 15:54 |
sdague | we seem very focussed on step 37 | 15:54 |
mordred | sure. but it we've designed the application to be scale out (which we have) | 15:54 |
mordred | designing the db interaction to be consistent with that seems non-insane | 15:55 |
mordred | if we hadn't - if we'd designed this as a more synchornous system without a ton of scale out components | 15:55 |
mordred | then I would not beat this drum as often | 15:55 |
pkholkin | johnthetubaguy: please look at the updated patch too, thanks! | 15:55 |
mordred | foreign keys behind vmware vcenter for instance? makes total sense | 15:55 |
*** bochi-michael has quit IRC | 15:55 | |
sdague | mordred: sure, it's just we've introduced actual performance issues in people's environments because of no sane cleanup | 15:56 |
mordred | yes. I agree. we should have sane cleanup | 15:56 |
sdague | because after FK were more or less dropped no one followed up with that | 15:56 |
mordred | well, yeah. there are many problems at the DB layer | 15:56 |
mordred | and to be fair - I'm not actually pushing hard on getting rid of them or blocking them | 15:57 |
mriedem | i hope to have sane cleanup by the end of this release :) | 15:57 |
mriedem | or die trying | 15:57 |
*** rpodolyaka1 has quit IRC | 15:57 | |
mordred | I'm just saying in the corner that they'r enot scalable | 15:57 |
sdague | mriedem: ok, so on cleanup, it actually seems like we should probably come up with a way of annotating the models | 15:57 |
mordred | sometimes to get to where we need to be, taking a step backwards is super useful | 15:57 |
mdbooth | pkholkin: Replied | 15:57 |
sdague | because some set of them are "archive with instances" and another set are "delete when instances archive" | 15:57 |
paul-carlton2 | sdague: re cancel live migration, not sure I understand what you are suggesting, the os-instance-actions only returns a list of actions on the instance, with no indication of the status and there is no task id, just a req-id | 15:57 |
*** e0ne_ has joined #openstack-nova | 15:57 | |
mdbooth | I believe the lack of FKs has caused us a great deal of pain. | 15:58 |
* mdbooth would prefer to see the code fail at source, so we can fix it. | 15:58 | |
*** e0ne has quit IRC | 15:58 | |
*** stevemar_ has quit IRC | 15:58 | |
mriedem | sdague: instance_actions are the former and bw_usage_cache is the latter | 15:59 |
sdague | paul-carlton2: I'm not suggesting anything really about os-instance-actions | 15:59 |
sdague | mriedem: right | 15:59 |
*** stevemar_ has joined #openstack-nova | 15:59 | |
sdague | mriedem: which I think are 2 cases the tool needs to handle | 15:59 |
mriedem | yeah, and will | 15:59 |
mriedem | archive and purge | 15:59 |
sdague | and once it knows how to do that generically, we can just annotate things | 15:59 |
sdague | and hopefully it's not massive gross to make that all come together | 15:59 |
*** rcernin has joined #openstack-nova | 16:00 | |
*** rcernin is now known as rcernin|dinner | 16:00 | |
mriedem | sdague: i have to write the changes to archive first to see how gross it is | 16:00 |
pkholkin | mdbooth: ok, thank you! will work on it) | 16:00 |
*** rpodolyaka1 has joined #openstack-nova | 16:00 | |
mriedem | then we can iterate and make it generic and nifty | 16:00 |
mdbooth | pkholkin: Does that make sense? | 16:00 |
sdague | like fixed ips | 16:00 |
sdague | that's a purge | 16:00 |
*** mgoddard_ has joined #openstack-nova | 16:01 | |
pkholkin | mdbooth: I understand your idea but I don't know if we can do it in oslo.db, will also speak to rpodolyaka | 16:01 |
mdbooth | pkholkin: K | 16:02 |
johnthetubaguy | sdague: your totally right about the url logging thing, hmm, I think I was fundamentally just being a coward with the live-migrate thing, trying to not open this debate, but you are right, it would be better this way. | 16:02 |
lxsli | sdague mriedem: so if we're purging bw_usage_cache, can I remove SoftDeleteMixin from it? | 16:02 |
lxsli | great discussion btw, +1 to all that | 16:03 |
*** rpodolyaka1 has quit IRC | 16:03 | |
andrearosa | lxsli: agree, very intersting discussion | 16:03 |
mriedem | sdague: we actually don't touch fixed_ips | 16:03 |
*** rpodolyaka1 has joined #openstack-nova | 16:03 | |
mriedem | sdague: nova-network creates those on startup | 16:03 |
mriedem | sdague: if you delete those, upgrade fails | 16:04 |
*** mgoddard has quit IRC | 16:04 | |
mriedem | i found that out when writing the db migration that made instances.uuid non-nullable | 16:04 |
*** rpodolyaka1 has quit IRC | 16:04 | |
*** rpodolyaka1 has joined #openstack-nova | 16:04 | |
sdague | mriedem: oh, right, we just need to make sure that uuid is nulled out | 16:05 |
*** ihrachys has quit IRC | 16:05 | |
mriedem | yup | 16:05 |
*** rpodolyaka1 has quit IRC | 16:05 | |
sdague | sorry, I was just looking through model definitions | 16:05 |
*** albertom has joined #openstack-nova | 16:05 | |
mriedem | so...do we want to add a fkey on bw_usage_cache? or just know it's a thing that needs to be handled special? | 16:06 |
*** rpodolyaka1 has joined #openstack-nova | 16:06 | |
*** rpodolyaka1 has quit IRC | 16:06 | |
mriedem | i kind of like consistency in the data model... | 16:06 |
*** rpodolyaka1 has joined #openstack-nova | 16:06 | |
mriedem | that could also be a separate ML thread | 16:06 |
mriedem | ccarmack: if you're looking for something to do, i've been meaning to scrub through python-novaclient looking for any python 2.6 compat code that we can remove | 16:07 |
*** zenoway has quit IRC | 16:08 | |
*** rpodolyaka1 has quit IRC | 16:08 | |
*** hparekh has quit IRC | 16:10 | |
*** zenoway has joined #openstack-nova | 16:11 | |
*** ihrachys has joined #openstack-nova | 16:13 | |
*** rlrossit has quit IRC | 16:14 | |
*** mhickey has quit IRC | 16:14 | |
mdbooth | danpb ndipanov: Can the libvirt driver have a disk with a backing file which isn't an image? If not, is that ever likely to change? | 16:14 |
* mdbooth doesn't want to over-engineer for a situation which will never happen | 16:15 | |
mdbooth | This is a local disk, btw, not a volume | 16:15 |
ndipanov | mdbooth, what do you mean a backing file which isn't an image? | 16:15 |
*** zenoway has quit IRC | 16:15 | |
*** ccarmack has quit IRC | 16:15 | |
mdbooth | Specifically a local disk which would have to be copied in cold migration with no shared storage | 16:16 |
danpb | mdbooth: i don't think we ever do that currently | 16:16 |
mdbooth | Is it possible that one of those disks might be an overlay with a backing file which isn't an image? | 16:16 |
ndipanov | mdbooth, no | 16:16 |
*** venkat_p has joined #openstack-nova | 16:16 | |
mdbooth | Didn't think so. Can you think of anything which might change that? | 16:16 |
*** ctrath has quit IRC | 16:17 | |
mdbooth | It would be convenient of the answer was no :) | 16:17 |
ndipanov | mdbooth, why would you want to change that | 16:17 |
*** venkat_p has quit IRC | 16:17 | |
mc_nair | sahid: can you review https://review.openstack.org/#/c/227851/ when you get a chance | 16:17 |
mdbooth | ndipanov: I wouldn't | 16:17 |
danpb | mdbooth: i cna't imagine us doing that | 16:17 |
mdbooth | Cool. That means I can have an optional backing file attribute which is just an image id | 16:17 |
mdbooth | Rather than something more complex | 16:18 |
*** armax has joined #openstack-nova | 16:18 | |
*** rlrossit has joined #openstack-nova | 16:19 | |
*** ihrachys has quit IRC | 16:20 | |
*** jianghuaw has joined #openstack-nova | 16:21 | |
*** vageli has joined #openstack-nova | 16:21 | |
*** otter768 has joined #openstack-nova | 16:22 | |
mriedem | mordred: didn't you have a patch that fixes this typo? https://github.com/openstack/nova/blob/12.0.0/nova/api/openstack/compute/servers.py#L322 | 16:22 |
mriedem | s/images/instances/ | 16:22 |
mordred | mriedem: yes | 16:23 |
mordred | mriedem: I believe it is in recheck hell | 16:23 |
mordred | mriedem: was approved a month ago | 16:23 |
mriedem | oh nvm, i'm looking at 12.0.0 | 16:24 |
mriedem | :( ignore | 16:24 |
mordred | mriedem: oh! it landed | 16:24 |
mordred | https://review.openstack.org/#/c/242102/ | 16:24 |
mriedem | hot dog | 16:24 |
mriedem | ATC! | 16:24 |
*** ccarmack has joined #openstack-nova | 16:24 | |
*** ctrath has joined #openstack-nova | 16:25 | |
mriedem | ha | 16:26 |
mriedem | nova list --deleted returns a 500 | 16:26 |
mriedem | nice | 16:26 |
ccarmack | mriedem: I'll work on that. Does it mean we don't support 2.6 anymore? | 16:26 |
*** stevemar_ has quit IRC | 16:26 | |
mriedem | ccarmack: we're moving to drop support for py26 in libs and clients | 16:26 |
mriedem | oslo is already dropping it | 16:26 |
*** otter768 has quit IRC | 16:27 | |
ccarmack | Are all the projects scrubbing their code? | 16:27 |
mriedem | i assume so, or should | 16:27 |
ccarmack | ok | 16:27 |
mriedem | since most projects rely on oslo and if oslo drops it, then the consuming project can't claim it supports it | 16:27 |
*** mdrabe has quit IRC | 16:27 | |
*** jianghuaw has quit IRC | 16:27 | |
dims | ccarmack : infra is dropping testing nodes for py26, so oslo has stopped testing against py26 and will release versions without 2.7 support next week | 16:28 |
dims | right | 16:28 |
ccarmack | dims: I guess all the users are on 2.7 | 16:29 |
*** jdurgin1 has joined #openstack-nova | 16:31 | |
*** ihrachys has joined #openstack-nova | 16:31 | |
*** eharney has quit IRC | 16:33 | |
dims | y | 16:33 |
*** armax has quit IRC | 16:34 | |
*** mylu has joined #openstack-nova | 16:34 | |
*** jlanoux has quit IRC | 16:34 | |
*** ctrath has quit IRC | 16:34 | |
*** paul-carlton1 has quit IRC | 16:35 | |
openstackgerrit | ShaoHe Feng proposed openstack/nova-specs: Attach/detach SR-IOV interface https://review.openstack.org/139910 | 16:35 |
*** mylu has quit IRC | 16:38 | |
*** stevemar_ has joined #openstack-nova | 16:39 | |
openstackgerrit | Balazs Gibizer proposed openstack/nova: Adds json sample for the versioned notifications https://review.openstack.org/248167 | 16:41 |
*** ildikov has quit IRC | 16:41 | |
*** mdrabe has joined #openstack-nova | 16:41 | |
mriedem | this is fun https://bugs.launchpad.net/nova/+bug/1518382 | 16:41 |
openstack | Launchpad bug 1518382 in OpenStack Compute (nova) "nova list --deleted fails with a 500 response" [Medium,Triaged] | 16:41 |
*** burgerk has quit IRC | 16:41 | |
*** jlanoux has joined #openstack-nova | 16:41 | |
openstackgerrit | Andrey Kurilin proposed openstack/python-novaclient: Drop python 2.6 classifier https://review.openstack.org/248168 | 16:42 |
mriedem | i think lazy load of the instance flavor is trying to load the instance which is deleted, but it's not using read_deleted='yes' | 16:42 |
*** rcernin|dinner is now known as rcernin | 16:42 | |
*** stevemar_ has quit IRC | 16:44 | |
*** unicell1 has joined #openstack-nova | 16:44 | |
*** dims_ has joined #openstack-nova | 16:44 | |
*** dims has quit IRC | 16:44 | |
*** mylu has joined #openstack-nova | 16:44 | |
*** stevemar_ has joined #openstack-nova | 16:45 | |
johnthetubaguy | mriedem: we have a --deleted flag, oh my, I forgot about that one | 16:45 |
mriedem | yup | 16:45 |
mriedem | remember melwitt schooling superdan about that in YVR? | 16:45 |
mriedem | :P | 16:45 |
superdan | I think that was dansmith, not me | 16:45 |
mriedem | oh right | 16:45 |
johnthetubaguy | heh | 16:45 |
mriedem | clark kent | 16:46 |
*** unicell has quit IRC | 16:46 | |
johnthetubaguy | I bet loads of people do that do double check the instance got deleted | 16:46 |
mriedem | this is the same devstack i was testing arhive on, so i'm not sure yet if that screwed up the db to the point that it's causing this to fail | 16:47 |
mriedem | *archive | 16:47 |
*** mylu has quit IRC | 16:48 | |
*** mylu has joined #openstack-nova | 16:48 | |
mriedem | but no, i see the issue | 16:48 |
*** mylu has quit IRC | 16:50 | |
*** mylu_ has joined #openstack-nova | 16:50 | |
*** ihrachys has quit IRC | 16:51 | |
sahid | mc_nair, mriedem done | 16:52 |
*** burgerk has joined #openstack-nova | 16:52 | |
*** Piet has quit IRC | 16:53 | |
*** sahid has quit IRC | 16:54 | |
*** kfarr has quit IRC | 16:55 | |
johnthetubaguy | mriedem: there is a vague plan to use the doc generation to check our tempest coverage, I hope we get that working | 16:55 |
*** shaohe_feng1 has quit IRC | 16:56 | |
*** ihrachys has joined #openstack-nova | 16:56 | |
*** ctrath has joined #openstack-nova | 16:56 | |
openstackgerrit | Andrew Laski proposed openstack/nova: Add persistence to the RequestSpec object https://review.openstack.org/211753 | 16:57 |
*** ssurana has joined #openstack-nova | 16:57 | |
*** sbezverk has joined #openstack-nova | 16:57 | |
*** e0ne_ has quit IRC | 16:58 | |
*** daemontool has quit IRC | 16:59 | |
*** stevemar_ has quit IRC | 16:59 | |
*** stevemar_ has joined #openstack-nova | 17:00 | |
*** chhavi has quit IRC | 17:00 | |
*** vageli has quit IRC | 17:00 | |
*** flyingtt has quit IRC | 17:01 | |
*** nic has joined #openstack-nova | 17:01 | |
*** daemontool has joined #openstack-nova | 17:02 | |
*** armax has joined #openstack-nova | 17:02 | |
*** eharney has joined #openstack-nova | 17:03 | |
*** stevemar_ has quit IRC | 17:04 | |
*** tdurakov has quit IRC | 17:04 | |
*** suro-patz has joined #openstack-nova | 17:04 | |
*** mgoddard_ has quit IRC | 17:05 | |
*** mgoddard_ has joined #openstack-nova | 17:05 | |
*** jlanoux has quit IRC | 17:05 | |
*** tdurakov has joined #openstack-nova | 17:06 | |
sbezverk | Hello, appreciate if somebody from nova could comment if nova still requires a port to have associated IP address, before it can be used by instance. | 17:06 |
*** suro-patz has quit IRC | 17:09 | |
openstackgerrit | Pushkar Umaranikar proposed openstack/nova: Modify VM's updated_at field on volume actions https://review.openstack.org/247176 | 17:09 |
*** pratikma_ is now known as pratikmallya | 17:10 | |
*** ctrath has quit IRC | 17:13 | |
*** scheuran has quit IRC | 17:13 | |
*** paul-carlton has joined #openstack-nova | 17:14 | |
*** ctrath has joined #openstack-nova | 17:15 | |
*** mylu_ has quit IRC | 17:16 | |
*** mylu has joined #openstack-nova | 17:16 | |
*** ildikov has joined #openstack-nova | 17:16 | |
*** PaulMurray is now known as ptm_away | 17:16 | |
*** bnemec has quit IRC | 17:20 | |
*** tdurakov has quit IRC | 17:23 | |
*** tdurakov has joined #openstack-nova | 17:23 | |
*** tdurakov has quit IRC | 17:27 | |
openstackgerrit | Ludovic Beliveau proposed openstack/nova: Update binding:profile for SR-IOV ports https://review.openstack.org/242573 | 17:29 |
openstackgerrit | Claudiu Belu proposed openstack/nova: Hyper-V: adds os-win library https://review.openstack.org/247729 | 17:29 |
albertom | Hi | 17:29 |
*** yonglihe has quit IRC | 17:29 | |
albertom | who is working with wsgi nova-api ? | 17:29 |
*** jdurgin1 has quit IRC | 17:29 | |
albertom | who can take a look at https://bugs.launchpad.net/nova/+bug/1506958 | 17:29 |
openstack | Launchpad bug 1506958 in OpenStack Compute (nova) "TypeError: object.__new__(thread.lock) is not safe, use thread.lock.__new__()" [Undecided,New] | 17:30 |
*** bnemec has joined #openstack-nova | 17:30 | |
*** wuhg has quit IRC | 17:31 | |
*** tdurakov has joined #openstack-nova | 17:32 | |
*** ihrachys has quit IRC | 17:32 | |
*** edtubill has joined #openstack-nova | 17:32 | |
*** nic has quit IRC | 17:32 | |
*** nic has joined #openstack-nova | 17:33 | |
*** claudiub has quit IRC | 17:33 | |
*** suro-patz has joined #openstack-nova | 17:33 | |
*** unicell1 has quit IRC | 17:34 | |
*** sudipto has joined #openstack-nova | 17:34 | |
*** leseb has joined #openstack-nova | 17:37 | |
*** tdurakov has quit IRC | 17:38 | |
*** tdurakov has joined #openstack-nova | 17:39 | |
*** lucasagomes is now known as lucas-dinner | 17:39 | |
*** pratikmallya has quit IRC | 17:41 | |
*** sacharya has joined #openstack-nova | 17:43 | |
*** devananda has quit IRC | 17:44 | |
*** devananda has joined #openstack-nova | 17:45 | |
*** mylu has quit IRC | 17:46 | |
*** mylu has joined #openstack-nova | 17:46 | |
*** paul-carlton has quit IRC | 17:46 | |
*** Piet has joined #openstack-nova | 17:47 | |
*** stackdump has joined #openstack-nova | 17:47 | |
*** garyk1 has quit IRC | 17:50 | |
*** mylu has quit IRC | 17:51 | |
*** rotbeard has quit IRC | 17:51 | |
*** zhangjn has quit IRC | 17:51 | |
*** yamahata has quit IRC | 17:52 | |
*** tdurakov has quit IRC | 17:52 | |
*** lykinsbd has quit IRC | 17:53 | |
*** matrohon has quit IRC | 17:55 | |
*** RichardRaseley has joined #openstack-nova | 17:56 | |
*** lykinsbd has joined #openstack-nova | 17:57 | |
*** ericksonsantos has quit IRC | 17:58 | |
*** penick has joined #openstack-nova | 17:58 | |
*** suro-patz has quit IRC | 17:58 | |
*** josecastroleon has quit IRC | 17:59 | |
*** mylu has joined #openstack-nova | 18:00 | |
*** mylu has quit IRC | 18:01 | |
*** jistr has quit IRC | 18:01 | |
*** achanda has quit IRC | 18:01 | |
*** mylu has joined #openstack-nova | 18:03 | |
*** unicell has joined #openstack-nova | 18:04 | |
*** daemontool has quit IRC | 18:05 | |
*** dims has joined #openstack-nova | 18:05 | |
*** suro-patz has joined #openstack-nova | 18:06 | |
*** dims_ has quit IRC | 18:06 | |
*** apoorvad has joined #openstack-nova | 18:06 | |
*** daemontool has joined #openstack-nova | 18:07 | |
*** smatzek has quit IRC | 18:07 | |
*** fawadkhaliq has quit IRC | 18:07 | |
*** mylu has quit IRC | 18:09 | |
*** pratikmallya has joined #openstack-nova | 18:09 | |
*** mylu has joined #openstack-nova | 18:09 | |
*** aix has quit IRC | 18:10 | |
openstackgerrit | Zoltan Arnold Nagy proposed openstack/nova-specs: Encryption support for rbd-backed volumes https://review.openstack.org/239798 | 18:10 |
superdan | man, I'm so stupid | 18:10 |
superdan | I should never have signed up to do this migration data refactor | 18:10 |
*** e0ne has joined #openstack-nova | 18:11 | |
openstackgerrit | Dan Smith proposed openstack/nova: Add xenapi support for XenapiLiveMigrateData objects https://review.openstack.org/247853 | 18:11 |
openstackgerrit | Dan Smith proposed openstack/nova: Make libvirt driver return migrate data objects for source and dest checks https://review.openstack.org/247720 | 18:11 |
openstackgerrit | Dan Smith proposed openstack/nova: Add transitional support for migrate data objects to compute manager https://review.openstack.org/247719 | 18:11 |
openstackgerrit | Dan Smith proposed openstack/nova: Actually pass the migration data object down to the virt drivers https://review.openstack.org/248211 | 18:11 |
*** mylu has quit IRC | 18:11 | |
*** mylu has joined #openstack-nova | 18:12 | |
danpb | superdan: oooh, you da man ! i was wondering if anyone would ever dare tackle that dragon | 18:13 |
*** paul-carlton has joined #openstack-nova | 18:13 | |
superdan | danpb: it's going to be the death of me, unfortunately | 18:13 |
*** w_verdugo has quit IRC | 18:13 | |
*** mriedem has quit IRC | 18:13 | |
*** tdurakov has joined #openstack-nova | 18:14 | |
*** mylu has quit IRC | 18:14 | |
*** vilobhmm has joined #openstack-nova | 18:14 | |
*** mylu has joined #openstack-nova | 18:15 | |
*** vilobhmm has quit IRC | 18:15 | |
*** vilobhmm has joined #openstack-nova | 18:15 | |
*** pratikma_ has joined #openstack-nova | 18:15 | |
*** klkumar has quit IRC | 18:15 | |
*** mylu_ has joined #openstack-nova | 18:17 | |
*** mriedem has joined #openstack-nova | 18:18 | |
*** mylu has quit IRC | 18:18 | |
*** pratikmallya has quit IRC | 18:19 | |
mriedem | superdan: this shouldn't be true in liberty anymore right? https://github.com/openstack/nova/blob/master/nova/objects/instance.py#L753 | 18:21 |
mriedem | i tried doing this: https://gist.github.com/mriedem/fdf0edfa19632f151f6d | 18:22 |
mriedem | and got this: https://gist.github.com/mriedem/38a602f0d6bc7d38d6e6 | 18:22 |
mriedem | since the instance context is nulled out | 18:22 |
*** e0ne has quit IRC | 18:22 | |
*** markmc has quit IRC | 18:22 | |
*** w_verdugo has joined #openstack-nova | 18:23 | |
*** otter768 has joined #openstack-nova | 18:23 | |
superdan | mriedem: is it reading a deleted instance? | 18:24 |
superdan | mriedem: we never migrated deleted instances' flavor information | 18:24 |
*** markus_z has quit IRC | 18:24 | |
superdan | so if it's deleted I think you'll need to just look up the flavor by id and call it good | 18:24 |
mriedem | it is reading a deleted instance, yes | 18:25 |
*** paul-carlton has quit IRC | 18:25 | |
mriedem | this is a new devstack, so not an upgrade from kilo | 18:25 |
superdan | so you hit that if you don't have flavor on the resulting instance, which is intentional to avoid a loop | 18:25 |
superdan | so you're not getting a flavor on that deleted instance | 18:26 |
*** achanda has joined #openstack-nova | 18:26 | |
superdan | I don't remember all the details of what we do, so look and see what its extra record looks like | 18:26 |
*** danpb has quit IRC | 18:27 | |
*** edtubill has quit IRC | 18:27 | |
*** otter768 has quit IRC | 18:28 | |
jroll | mriedem: that spec review you did for me was useful, no need to apologize. though you could have just said "gimme the details yo" | 18:28 |
jroll | hard to write specs when you have all the context, end up leaving things out | 18:29 |
*** thorst_ has joined #openstack-nova | 18:29 | |
*** moshele has joined #openstack-nova | 18:31 | |
*** tdurakov has quit IRC | 18:32 | |
*** tdurakov has joined #openstack-nova | 18:33 | |
*** thorst has quit IRC | 18:33 | |
*** tdurakov has quit IRC | 18:33 | |
*** vilobhmm has quit IRC | 18:33 | |
*** e0ne has joined #openstack-nova | 18:34 | |
*** zhipeng has joined #openstack-nova | 18:35 | |
*** mgoddard_ has quit IRC | 18:35 | |
*** mylu_ has quit IRC | 18:39 | |
*** mylu has joined #openstack-nova | 18:39 | |
*** edmondsw has joined #openstack-nova | 18:39 | |
*** vilobhmm has joined #openstack-nova | 18:41 | |
*** nickchase has joined #openstack-nova | 18:42 | |
mriedem | superdan: heh | 18:42 |
mriedem | mysql> select instance_uuid,flavor,deleted from nova.instance_extra; | 18:42 |
mriedem | Empty set (0.00 sec) | 18:42 |
superdan | I wanna say we delete the extra when we delete the instance | 18:42 |
superdan | like actually delete it for reals | 18:43 |
mriedem | i also ran archive on this before doing this | 18:43 |
superdan | ah | 18:43 |
mriedem | so that could have removed extra | 18:43 |
mriedem | but b/c of the fkey issues, the instances are still there | 18:43 |
mriedem | will try to recreate w/o archive | 18:43 |
superdan | dude, don't you know? fkeys kill cloud kittens | 18:43 |
mriedem | well, | 18:43 |
mriedem | they do if you (1) aren't consistent about using them and (2) consistent about soft deleting all things in the fkey chain | 18:44 |
mriedem | which we aren't | 18:44 |
superdan | that's not the story I heard | 18:44 |
*** jbernard has quit IRC | 18:44 | |
*** pixelbeat has quit IRC | 18:45 | |
*** stevemar_ has joined #openstack-nova | 18:46 | |
*** moshele has quit IRC | 18:47 | |
*** sbezverk has quit IRC | 18:47 | |
*** gjayavelu has joined #openstack-nova | 18:48 | |
*** stevemar_ has quit IRC | 18:48 | |
*** stevemar_ has joined #openstack-nova | 18:48 | |
ccarmack | mriedem: does removing py26 compat code from python-novaclient also mean removing things like the tox py26 env? | 18:50 |
mriedem | ccarmack: i think so | 18:51 |
superdan | sdague: so I thought the existing multinode job did one live migration test? I don't see it in here: http://logs.openstack.org/20/247720/2/check/gate-grenade-dsvm-multinode/004d380/logs/testr_results.html.gz | 18:51 |
ccarmack | ok | 18:51 |
superdan | oh wait | 18:51 |
superdan | maybe multinode-full? | 18:51 |
*** aginwala has joined #openstack-nova | 18:51 | |
superdan | yep, there we go | 18:52 |
*** salv-orl_ has joined #openstack-nova | 18:54 | |
*** e0ne has quit IRC | 18:57 | |
*** salv-orlando has quit IRC | 18:57 | |
*** ijw has joined #openstack-nova | 19:00 | |
zhipeng | hey do we still have spec review today | 19:00 |
mriedem | is it just me, or is it dumb that you have to be an admin to list your tenant's deleted instances? | 19:04 |
*** ijw has quit IRC | 19:04 | |
mriedem | i understand why all-tenants is admin-only | 19:04 |
*** ijw has joined #openstack-nova | 19:04 | |
mriedem | but nova list --deleted should work for non-admins i'd think | 19:04 |
*** lpetrut has quit IRC | 19:06 | |
*** rk4n has quit IRC | 19:07 | |
*** aginwala has quit IRC | 19:09 | |
melwitt | mriedem: all should be policy controllable, in my opinion. and I can't immediately think why a tenant shouldn't be able to see their own deleted instances | 19:10 |
mriedem | yeah, i guess that's another spec for another time | 19:11 |
*** stevemar_ has quit IRC | 19:13 | |
*** aginwala has joined #openstack-nova | 19:13 | |
openstackgerrit | Ryan Rossiter proposed openstack/nova: Use ContainerFormat instead of strings https://review.openstack.org/245835 | 19:13 |
openstackgerrit | Ryan Rossiter proposed openstack/nova: Add ContainerFormat field https://review.openstack.org/245834 | 19:13 |
*** mc_nair has quit IRC | 19:13 | |
*** mylu has quit IRC | 19:13 | |
doffm | sdague: Is there any way I can help with the live-migration work or is it all under control? | 19:13 |
*** pumaranikar has quit IRC | 19:15 | |
*** stevemar_ has joined #openstack-nova | 19:16 | |
*** aginwala has quit IRC | 19:16 | |
*** baoli has quit IRC | 19:17 | |
*** smatzek has joined #openstack-nova | 19:18 | |
*** sudipto has quit IRC | 19:18 | |
*** mylu has joined #openstack-nova | 19:21 | |
*** jdurgin has quit IRC | 19:22 | |
*** eglynn has quit IRC | 19:22 | |
*** aginwala has joined #openstack-nova | 19:24 | |
*** jdurgin has joined #openstack-nova | 19:25 | |
*** rlrossit has left #openstack-nova | 19:28 | |
*** zhipeng has quit IRC | 19:30 | |
*** liusheng has quit IRC | 19:30 | |
mriedem | i'm not even sure where non-admin is filtered out for nova list --deleted | 19:30 |
mriedem | the db api doesn't seem to care | 19:30 |
*** gjayavelu has quit IRC | 19:31 | |
*** liusheng has joined #openstack-nova | 19:32 | |
melwitt | mriedem: I think it might be remove_invalid_options in servers.py that strips out the search_opt if context isn't admin | 19:33 |
mriedem | nope | 19:33 |
mriedem | for key in ('sort_key', 'sort_dir', 'limit', 'marker'): | 19:33 |
*** sacharya has quit IRC | 19:34 | |
*** achanda has quit IRC | 19:34 | |
melwitt | hm | 19:35 |
*** rook0- has quit IRC | 19:35 | |
mriedem | i see it now | 19:35 |
mriedem | https://github.com/openstack/nova/blob/12.0.0/nova/api/openstack/compute/servers.py#L335-L340 | 19:36 |
mriedem | wait, nvm | 19:37 |
mriedem | that's only if you're searching on vm_state == 'deleted' | 19:37 |
mriedem | which i'm not | 19:37 |
*** lykinsbd has quit IRC | 19:40 | |
*** yamahata has joined #openstack-nova | 19:41 | |
*** jdurgin has quit IRC | 19:41 | |
sdague | superdan: grenade does not | 19:41 |
sdague | grenade only runs the smoke target | 19:41 |
sdague | which is pretty stripped down | 19:42 |
mtreinish | sdague: we could add an experimental job that does a full tempest run, but I'm not sure it provides much more | 19:44 |
*** mc_nair has joined #openstack-nova | 19:44 | |
sdague | we have a different full run | 19:45 |
*** pumaranikar has joined #openstack-nova | 19:45 | |
sdague | it's just not passing regularly | 19:45 |
mtreinish | oh, is the desire just for multinode full I didn't read the backscroll | 19:45 |
*** mylu has quit IRC | 19:46 | |
mtreinish | yeah that job exists and doesnt work | 19:46 |
mtreinish | because of the shelve thing right? | 19:46 |
*** mylu has joined #openstack-nova | 19:46 | |
*** zenoway has joined #openstack-nova | 19:46 | |
*** lykinsbd has joined #openstack-nova | 19:46 | |
openstackgerrit | Bob Ball proposed openstack/nova: XenAPI: Cope with more Cinder backends https://review.openstack.org/244789 | 19:47 |
*** achanda has joined #openstack-nova | 19:48 | |
superdan | sdague: oh, I see I was looking at grenade.. unintentionally | 19:48 |
superdan | sdague: the full job shows that I didn't break live migration for that config at least | 19:49 |
*** ociuhandu has quit IRC | 19:49 | |
sdague | superdan: \o/ | 19:49 |
sdague | heh, we landed the liberty backport of the ebtables hack in devstack | 19:51 |
*** lykinsbd has quit IRC | 19:51 | |
*** mylu has quit IRC | 19:51 | |
sdague | now we just see kilo fails | 19:51 |
*** sacharya has joined #openstack-nova | 19:51 | |
*** liusheng has quit IRC | 19:52 | |
*** liusheng has joined #openstack-nova | 19:52 | |
openstackgerrit | Bob Ball proposed openstack/nova: XenAPI: Workaround for 6.5 iSCSI bug https://review.openstack.org/231103 | 19:53 |
superdan | xen's live migration path is not very well unit tested, so I'm fairly sure I'm going to break it | 19:53 |
superdan | but I have no way to test it, so.. | 19:53 |
*** lykinsbd has joined #openstack-nova | 19:55 | |
*** eharney has quit IRC | 19:57 | |
*** mylu has joined #openstack-nova | 19:57 | |
*** aginwala has quit IRC | 19:58 | |
figleaf | mriedem: it's here: https://github.com/openstack/nova/blob/12.0.0/nova/api/openstack/compute/servers.py#L1114-L1121 | 19:58 |
mriedem | figleaf: yeah, but what does that have to do with 'deleted'? | 19:58 |
*** aginwala has joined #openstack-nova | 19:58 | |
mriedem | figleaf: oh you mean b/c it's NOT in that list | 19:58 |
figleaf | mriedem: yup | 19:59 |
mriedem | got it | 19:59 |
figleaf | mriedem: from the log: Removing options 'deleted' from query from (pid=28732) remove_invalid_options /opt/stack/nova/nova/api/openstack/compute/servers.py:1176 | 19:59 |
figleaf | mriedem: that's when I run 'nova list --deleted' as a regular user | 20:00 |
mriedem | yup | 20:00 |
mriedem | easy fix | 20:00 |
mriedem | with a microversion | 20:00 |
mriedem | i already have a bp filed https://blueprints.launchpad.net/nova/+spec/non-admin-list-deleted-instances | 20:00 |
mriedem | just need to find the time to write a spec | 20:00 |
*** ctrath has quit IRC | 20:00 | |
figleaf | mriedem: I can do it | 20:00 |
superdan | we want to further the use of deleted things? | 20:00 |
figleaf | superdan: no, but it's already there in the API | 20:00 |
superdan | I thought we specifically wanted to cut down on and maybe remove it eventually | 20:01 |
figleaf | just making it work as advertised | 20:01 |
superdan | figleaf: yeah, aware | 20:01 |
superdan | figleaf: that doesn't mean we have to add a microversion to enable more of it | 20:01 |
figleaf | superdan: so maybe a microversion to remove it altogether? | 20:01 |
mriedem | i've sort of lost track of what are the million hoops we have to jump through to drop soft deleted | 20:01 |
mriedem | but yeah, removing access to deleted things in the API is a first step | 20:02 |
superdan | mriedem: it mostly hinges on the api telling the users they can do it | 20:02 |
superdan | so adding more surface that allows it seems like a bad idea to me | 20:02 |
*** mylu has quit IRC | 20:02 | |
mriedem | heh, you won't like this either then https://blueprints.launchpad.net/nova/+spec/os-instance-actions-read-deleted-instances | 20:02 |
*** mylu has joined #openstack-nova | 20:02 | |
*** mylu has joined #openstack-nova | 20:02 | |
superdan | leaving the actions for a longer period than the instance makes sense to me | 20:03 |
superdan | i.e. instance is deleted but the actions are not yet deleted, that's fine | 20:03 |
mriedem | well, ^ is for allowing os-instance-actions to read deleted instances | 20:04 |
mriedem | since that seems useful | 20:04 |
*** signed8bit is now known as signed8bit_ZZZzz | 20:04 | |
superdan | well, I'm not sure why we need to read deleted instances to read deleted actions | 20:04 |
superdan | maybe an implementation detail/requirement | 20:04 |
superdan | anyway | 20:04 |
mriedem | these aren't deleted actions | 20:05 |
mriedem | b/c we don't soft delete actions | 20:05 |
superdan | I know | 20:05 |
mriedem | so this is postmortem on the deleted instance | 20:05 |
superdan | sorry, I meant actions for deleted instances, not deleted actions themselves | 20:05 |
superdan | right | 20:05 |
*** mylu has quit IRC | 20:05 | |
*** mylu has joined #openstack-nova | 20:07 | |
mriedem | so you disagree with that bp? | 20:07 |
superdan | I think that bp is saying we need to read (soft) deleted instances in order to show actions for them right? | 20:07 |
mriedem | yes | 20:08 |
*** burgerk has quit IRC | 20:08 | |
superdan | and I'm saying I don't see why that is, unless there is some mechanical reason that is hard to get rid of | 20:08 |
superdan | i.e. if I have the instance uuid, we should be able to look up actions for that uuid, | 20:08 |
melwitt | it's a bit silly, the api does an instance lookup by uuid, passes the instance object to the actions api where the actions api passes only the uuid. the instance object isn't actually needed and doesn't need to be looked up | 20:08 |
superdan | or if I want to see changes-since actions for my tenant, we should be able to do that | 20:08 |
mriedem | superdan: os-instance-actions takes the instance uuid | 20:09 |
superdan | melwitt: right, my point | 20:09 |
mriedem | to look up the actions | 20:09 |
*** ctrath has joined #openstack-nova | 20:09 | |
superdan | mriedem: uuid is fine, but we shouldn't need to actually have to locate the whole instance, I think | 20:09 |
mriedem | yeah, i see where you're going now | 20:10 |
mriedem | just do the db query | 20:10 |
superdan | yar | 20:10 |
superdan | now, we might have to store something else like project_id with the action or something, | 20:10 |
superdan | but I'd rather make the BP about that than reading more deleted instance data | 20:10 |
*** smurke has joined #openstack-nova | 20:11 | |
superdan | we'll also have the instance map table in the api, | 20:11 |
superdan | which could give us our security check, | 20:11 |
superdan | but I don't know what the plans are for purging that on instance delte | 20:11 |
superdan | er, delete | 20:11 |
superdan | so maybe it's better to just store project with the actions, I dunno | 20:11 |
mriedem | i think that's already stored | 20:11 |
superdan | well then it should be "easy" | 20:12 |
superdan | SMOP | 20:12 |
mriedem | https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/models.py#L1248 | 20:12 |
*** signed8bit_ZZZzz is now known as signed8bit | 20:12 | |
superdan | perfect | 20:12 |
superdan | I'm going to go get lunch now | 20:14 |
superdan | just thought everyone should know that | 20:15 |
*** smurke_ has joined #openstack-nova | 20:15 | |
melwitt | I was thinking some more about why not let non-admin list deleted instances and if they're "using the cloud" and creating/deleting a lot, every user could have potentially 100+ deleted instances and listing them could get unwieldy on the api load | 20:15 |
mriedem | melwitt: you could using paging, but yeah.. | 20:17 |
mriedem | i think i'm giving up on that one | 20:17 |
mriedem | given the gauntlet that was thrown down in YVR | 20:17 |
mriedem | i guess we should put that in our devref somewhere that says, as a policy, we don't want more API changes to read deleted resources | 20:17 |
*** ctrath has quit IRC | 20:18 | |
figleaf | mriedem: if you're going to do that, it seems that it should be coupled with a policy that says deleted stuff hanging around is going to go away | 20:19 |
mriedem | figleaf: but we don't really have consensus on what that's going to be | 20:19 |
*** thumpba has quit IRC | 20:20 | |
*** apoorvad has quit IRC | 20:21 | |
figleaf | mriedem: is there such consensus on not fixing broken APIs if they deal with deleted instances? | 20:21 |
*** vilobhmm has quit IRC | 20:21 | |
*** vilobhmm has joined #openstack-nova | 20:22 | |
BobBall | superdan: I can test the live migration for xen :) This for the XenapiLiveMigrateData change? | 20:22 |
*** thumpba has joined #openstack-nova | 20:22 | |
*** dims has quit IRC | 20:23 | |
*** otter768 has joined #openstack-nova | 20:24 | |
*** thangp has quit IRC | 20:26 | |
mriedem | figleaf: not sure i'm following, but before we could get rid of soft delete we'd have to disable reading deleted instances in the db | 20:26 |
mriedem | but i'm not sure how we can really do that with just a microversion | 20:26 |
mriedem | since if we drop the deleted column in the db, the api microversion don't get a rat's ass | 20:27 |
mriedem | s/get/give/ | 20:27 |
figleaf | mriedem: I meant fixing 'nova list --deleted' | 20:27 |
mriedem | figleaf: i guess it's a won't fix | 20:28 |
figleaf | mriedem: and yeah, removing the deleted column means that at the very least, --deleted will be toast | 20:28 |
figleaf | mriedem: ok. I was just asking if there already was consensus on doing this | 20:28 |
*** otter768 has quit IRC | 20:28 | |
mriedem | which i guess gets us back to i don't know how we drop the deleted column | 20:29 |
mriedem | since that would seem to be a backward incompatible change that affects the API | 20:29 |
mriedem | regardless of microversion | 20:29 |
mtreinish | mriedem: you could add a config opt and have it write all the deleted instances to a new table. So if you want to preserve the old api | 20:30 |
mtreinish | not sure that really helps though | 20:30 |
*** thumpba_ has joined #openstack-nova | 20:31 | |
mriedem | that wouldn't really be 'def coreable' | 20:31 |
mriedem | bad UX | 20:31 |
mtreinish | do you want it to be def coreable? | 20:31 |
mtreinish | like you don't need to include a test that lists deleted instances | 20:32 |
mtreinish | and if you microversion against adding the option newer versions dont even acknowlege it as a thing | 20:32 |
*** rpodolyaka1 has joined #openstack-nova | 20:32 | |
*** thumpba has quit IRC | 20:33 | |
mriedem | yeah | 20:33 |
mriedem | so it's not great | 20:33 |
mriedem | anyway, have to run out for awhile | 20:33 |
mriedem | bbl | 20:33 |
mriedem | what i want is for my 2016 health care elections website to work | 20:33 |
*** mriedem has quit IRC | 20:33 | |
*** eharney has joined #openstack-nova | 20:36 | |
*** tdurakov has joined #openstack-nova | 20:37 | |
*** artom has quit IRC | 20:38 | |
*** aginwala has quit IRC | 20:40 | |
*** yassine__ has quit IRC | 20:42 | |
openstackgerrit | Samuel Matzek proposed openstack/nova: Reschedule boot when attach volume fails https://review.openstack.org/246505 | 20:46 |
*** aginwala has joined #openstack-nova | 20:47 | |
*** dane-fichter has quit IRC | 20:49 | |
*** thumpba_ has quit IRC | 20:50 | |
*** jdurgin has joined #openstack-nova | 20:52 | |
*** achanda has quit IRC | 20:53 | |
*** achanda has joined #openstack-nova | 20:56 | |
superdan | BobBall: yes and cool :) | 20:58 |
superdan | BobBall: not worth doing it just yet though | 20:58 |
superdan | still working through those | 20:58 |
*** openstackstatus has quit IRC | 21:02 | |
*** openstack has joined #openstack-nova | 21:04 | |
*** pratikm__ has joined #openstack-nova | 21:05 | |
*** pratikma_ has quit IRC | 21:05 | |
*** ijw has joined #openstack-nova | 21:06 | |
*** rpodolyaka1 has quit IRC | 21:06 | |
openstackgerrit | Brian Rosmaita proposed openstack/nova-specs: Consistent Instance Actions https://review.openstack.org/248248 | 21:07 |
*** tdurakov has quit IRC | 21:07 | |
*** mylu has quit IRC | 21:07 | |
*** mylu has joined #openstack-nova | 21:07 | |
*** pratikmallya has quit IRC | 21:08 | |
*** jianghuaw has joined #openstack-nova | 21:09 | |
*** tdurakov has joined #openstack-nova | 21:10 | |
bauwser | alaski: around? | 21:11 |
bauwser | alaski: trying to see why http://logs.openstack.org/53/211753/35/check/gate-nova-tox-functional/2fddd1f/console.html#_2015-11-20_17_28_37_751 doesn't work | 21:11 |
*** mylu has quit IRC | 21:11 | |
*** jianghuaw has quit IRC | 21:14 | |
*** marcusvrn_ has quit IRC | 21:15 | |
tpeoples | melwitt: do you want me to respin for your comment on https://review.openstack.org/#/c/220634/ ? | 21:16 |
*** sfinucan has joined #openstack-nova | 21:17 | |
*** daemontool has quit IRC | 21:19 | |
*** daemontool has joined #openstack-nova | 21:19 | |
openstackgerrit | Brian Rosmaita proposed openstack/nova-specs: Consistent Instance Actions https://review.openstack.org/248248 | 21:20 |
*** ociuhandu has joined #openstack-nova | 21:20 | |
*** apoorvad has joined #openstack-nova | 21:21 | |
*** mylu has joined #openstack-nova | 21:25 | |
*** apoorvad has quit IRC | 21:27 | |
*** gszasz has quit IRC | 21:30 | |
*** jbernard has joined #openstack-nova | 21:31 | |
*** stevemar_ has quit IRC | 21:31 | |
*** thorst_ has quit IRC | 21:31 | |
*** smatzek has quit IRC | 21:31 | |
*** mylu has quit IRC | 21:31 | |
*** mylu has joined #openstack-nova | 21:32 | |
*** mylu_ has joined #openstack-nova | 21:33 | |
*** mylu has quit IRC | 21:33 | |
*** vladikr has quit IRC | 21:33 | |
*** vilobhmm has quit IRC | 21:40 | |
*** vilobhmm has joined #openstack-nova | 21:41 | |
*** RichardRaseley has quit IRC | 21:41 | |
*** ijw has quit IRC | 21:42 | |
*** mriedem has joined #openstack-nova | 21:46 | |
openstackgerrit | Stephen Finucane proposed openstack/nova-specs: Rework policies for virt-driver CPU thread pinning https://review.openstack.org/244198 | 21:49 |
*** rcernin has quit IRC | 21:50 | |
mriedem | did everyone get everything figured out in the last hour? | 21:52 |
mriedem | re: deleting things and reading them? | 21:52 |
*** lykinsbd has quit IRC | 21:52 | |
melwitt | tpeoples: no need to respin IMO, I just mentioned it since I noticed it | 21:53 |
*** penick has quit IRC | 21:56 | |
sfinucan | Happy Friday. Is there anyone I can poke to get another +2 on Nikola's bugfixes? https://bugs.launchpad.net/nova/+bug/1250066 They're blocking my rework :( | 21:58 |
openstack | Launchpad bug 1250066 in OpenStack Compute (nova) "Need change instance's property updated_at when attach/detach volume" [Low,In progress] - Assigned to Pushkar Umaranikar (pushkar-umaranikar) | 21:58 |
openstackgerrit | Ed Leafe proposed openstack/nova: WIP: Add better help text to scheduler options https://review.openstack.org/247181 | 21:59 |
openstackgerrit | Ed Leafe proposed openstack/nova: Config options: centralize section "scheduler" https://review.openstack.org/245891 | 21:59 |
*** pradk has quit IRC | 22:02 | |
*** dave-mccowan has quit IRC | 22:03 | |
*** mylu_ has quit IRC | 22:05 | |
*** mylu has joined #openstack-nova | 22:05 | |
*** tdurakov has quit IRC | 22:06 | |
*** ZZelle_ has joined #openstack-nova | 22:07 | |
*** mylu_ has joined #openstack-nova | 22:07 | |
*** mylu has quit IRC | 22:07 | |
*** richm has left #openstack-nova | 22:11 | |
mriedem | superdan: on that instance actions API change, it does look up the instance to use that for policy auth, i'm not sure what removing that does | 22:11 |
mriedem | we'd just be authorizing by the context i think | 22:11 |
*** ijw has joined #openstack-nova | 22:12 | |
openstackgerrit | Brian Rosmaita proposed openstack/nova-specs: Lightweight Nova Service Profiling https://review.openstack.org/247616 | 22:12 |
*** pumaranikar has quit IRC | 22:13 | |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'hw_cpu_threads_policy' to ImageMetaProps https://review.openstack.org/202647 | 22:13 |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'cpu_policy' and 'cpu_threads_policy' fields https://review.openstack.org/202648 | 22:13 |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'hw:cpu_threads_policy=avoid' filtering https://review.openstack.org/202649 | 22:13 |
openstackgerrit | Stephen Finucane proposed openstack/nova: trivial: Add some logs to 'numa_topology_filter' https://review.openstack.org/202650 | 22:13 |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'hw:cpu_threads_policy=require' scheduling https://review.openstack.org/202651 | 22:13 |
*** aginwala has quit IRC | 22:14 | |
*** vilobhmm has quit IRC | 22:15 | |
*** vilobhmm has joined #openstack-nova | 22:15 | |
*** aginwala has joined #openstack-nova | 22:16 | |
*** smurke has quit IRC | 22:17 | |
*** bauwser is now known as bauzas | 22:17 | |
*** smurke_ has quit IRC | 22:18 | |
*** edtubill has joined #openstack-nova | 22:18 | |
*** yamahata has quit IRC | 22:18 | |
melwitt | mriedem: hm, it might be trying to make sure you're an instance owner or admin in order to authorize | 22:19 |
mriedem | yeah i have to assume that's what it's doing, i'm having a hard time tracing down how the instance target is used in the policy rules enforcement | 22:20 |
openstackgerrit | Stephen Finucane proposed openstack/nova: test_fields: Remove all 'Enum' subclass tests https://review.openstack.org/244372 | 22:20 |
mriedem | so the default policy is: "os_compute_api:os-instance-actions": "", | 22:21 |
mriedem | which i guess defaults to rule:admin_or_owner ? | 22:22 |
melwitt | I think so | 22:22 |
bauzas | no, anyone | 22:23 |
bauzas | oh sec | 22:23 |
bauzas | no | 22:23 |
bauzas | https://github.com/openstack/nova/blob/master/etc/nova/policy.json#L4 | 22:23 |
mriedem | so if user A in project A knows about an instance uuid for user B project B, if we remove that policy check then user A could get the instance actions for the instance in project B ? | 22:23 |
mriedem | referring to this https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/instance_actions.py#L57 | 22:24 |
mriedem | but, | 22:24 |
bauzas | it does the lookup here https://github.com/openstack/nova/blob/master/etc/nova/policy.json#L3 | 22:24 |
mriedem | if we just change that code to lookup the instance action by instance uuid and the project_id from the context, that would handle it, right? | 22:25 |
*** otter768 has joined #openstack-nova | 22:25 | |
*** doug-fish has quit IRC | 22:25 | |
mriedem | if we do ^, | 22:25 |
bauzas | mriedem: sec | 22:26 |
mriedem | then if project A is looking up project B's instance uuid, it wouldn't find it | 22:26 |
mriedem | and we return 404 | 22:26 |
bauzas | I have to remember the oslo.policy code | 22:26 |
melwitt | yeah, I'm confused about who the "owner" of a set of instance actions is, from an instance actions records perspective | 22:27 |
melwitt | the current code is checking ownership of the instance itself (we assume) | 22:27 |
*** stackdump has quit IRC | 22:29 | |
*** ijw has quit IRC | 22:29 | |
*** otter768 has quit IRC | 22:30 | |
melwitt | mriedem: if you do the lookup that way you would miss actions that were initiated by admin (or some other tenant), I think | 22:30 |
bauzas | so yeah the target is where we verify against the project_id | 22:31 |
*** RichardRaseley has joined #openstack-nova | 22:31 | |
mriedem | melwitt: so like a tenant user looking up instance actions performed on their instance but done by the admin? | 22:31 |
melwitt | mriedem: yeah, I'm thinking the way it currently is, it policy auths that you're the owner or an admin, and then you get to see all actions on that instance, no matter what tenant did them | 22:32 |
mriedem | bauzas: yeah so in the case of target=instance thta's the project_id on the instance | 22:32 |
*** yamahata has joined #openstack-nova | 22:32 | |
bauzas | correct, that's what I verified in oslo.policy | 22:32 |
melwitt | if for example admin rebooted your instance at some point, you would see that | 22:32 |
mriedem | yeah | 22:33 |
mriedem | so.... | 22:33 |
mriedem | we have to do the instance lookup | 22:33 |
melwitt | yeah. there's a reason it does that :( | 22:33 |
bauzas | indeed | 22:33 |
mriedem | alright, well, that still makes for an even easier spec | 22:33 |
bauzas | mriedem: I answered to that thread by saying that looking up at an known uuid is already a good thing | 22:34 |
*** thorst has joined #openstack-nova | 22:34 | |
mriedem | yup saw that | 22:34 |
mriedem | i was just looking into this policy auth thing before starting on the spec | 22:35 |
bauzas | we could tho imagine a separate policy for getting any action | 22:35 |
bauzas | but it would be an admin one | 22:35 |
mriedem | so if context.is_admin(): bypass looking up the instance for auth and just get all actions for that instance uuid | 22:36 |
mriedem | ? | 22:36 |
melwitt | I was trying to think if there's a way we could effectively pass read_deleted for just the instance_actions lookup call | 22:37 |
bauzas | mriedem: if we use https://github.com/openstack/nova/blob/master/etc/nova/policy.json#L92 then it only checks the context, yes | 22:38 |
*** thorst has quit IRC | 22:39 | |
bauzas | mriedem: and we could leave an empty target IIRC | 22:39 |
bauzas | sec, verifying | 22:39 |
mriedem | melwitt: i was just thinking this https://gist.github.com/mriedem/891cc00c528e7ba2d99b | 22:39 |
mriedem | crap, well, around the instance_get actually | 22:40 |
melwitt | mriedem: I didn't think you could need it there, the instance_actions aren't the thing that's deleted. right? | 22:40 |
mriedem | right | 22:40 |
mriedem | mutate around common.get_instance | 22:40 |
melwitt | yeah, that's a lot simpler than I was thinking it would be | 22:41 |
mriedem | so https://gist.github.com/mriedem/ab9cdde3d7e3f0911112 | 22:41 |
bauzas | mriedem: so, confirmed, if we have an admin policy, we can just authorize(context) | 22:42 |
melwitt | mriedem: yeah, I thought that would be an easy way to solve it. not expose read deleted everywhere, but only for these instance_actions auth | 22:43 |
*** thorst has joined #openstack-nova | 22:43 | |
*** jwcroppe has quit IRC | 22:44 | |
melwitt | being that I think the instance owner policy enforcement here does make sense now that we've talked about it | 22:44 |
bauzas | mriedem: lol, I missed your consensus with sdague, so I don't want again to say YUUUUUP | 22:44 |
bauzas | I mean in the ML | 22:45 |
bauzas | but ack | 22:45 |
mriedem | melwitt: yeah | 22:46 |
bauzas | heh sec | 22:47 |
bauzas | mriedem: melwitt: https://github.com/openstack/nova/blob/master/nova/api/openstack/extensions.py#L347-L349 | 22:47 |
bauzas | I remember now | 22:48 |
bauzas | by default we take the project_id from the user context | 22:48 |
*** thorst has quit IRC | 22:48 | |
*** junjie has quit IRC | 22:48 | |
*** betherly has quit IRC | 22:49 | |
*** kgalanov has quit IRC | 22:49 | |
*** tpeoples has quit IRC | 22:49 | |
mriedem | yup | 22:49 |
mriedem | which is pretty much how we auth everywhere | 22:49 |
mriedem | there are very few authorize(context, target=foo) APIs | 22:50 |
melwitt | yeah. this thing actually setting the target was something I hadn't seen before | 22:50 |
*** auggy has quit IRC | 22:50 | |
bauzas | IIUC, that only makes sure that we have a project_id in the context | 22:51 |
*** josh6627 is now known as jhesketh | 22:51 | |
*** sneti has quit IRC | 22:51 | |
*** mc_nair has quit IRC | 22:52 | |
*** krtaylor has quit IRC | 22:55 | |
*** aginwala has quit IRC | 22:56 | |
*** signed8bit is now known as signed8bit_ZZZzz | 22:57 | |
*** mdrabe has quit IRC | 22:58 | |
*** ijuwang has quit IRC | 22:58 | |
*** burt has quit IRC | 22:59 | |
bauzas | mriedem: melwitt: yeah, correct, it's for checking %(project_id)s % target == project_id | 22:59 |
bauzas | https://github.com/openstack/oslo.policy/blob/master/oslo_policy/_checks.py#L332 | 23:00 |
* bauzas goes in weekend | 23:02 | |
*** stevemar has joined #openstack-nova | 23:02 | |
*** xyang1 has quit IRC | 23:03 | |
*** stevemar_ has joined #openstack-nova | 23:04 | |
*** stevemar_ has quit IRC | 23:04 | |
*** stackdump has joined #openstack-nova | 23:05 | |
openstackgerrit | Stephen Finucane proposed openstack/nova: Rework 'limited' and 'get_limit_and_marker' https://review.openstack.org/241361 | 23:07 |
*** krtaylor has joined #openstack-nova | 23:08 | |
*** stackdump has quit IRC | 23:09 | |
*** aginwala has joined #openstack-nova | 23:09 | |
*** mrkz has quit IRC | 23:10 | |
*** pratikmallya has joined #openstack-nova | 23:11 | |
*** jhesketh has quit IRC | 23:13 | |
*** pratikm__ has quit IRC | 23:14 | |
*** betherly has joined #openstack-nova | 23:14 | |
*** jwcroppe has joined #openstack-nova | 23:14 | |
*** jhesketh has joined #openstack-nova | 23:15 | |
*** rk4n has joined #openstack-nova | 23:15 | |
*** pratikmallya has quit IRC | 23:19 | |
*** edtubill has quit IRC | 23:19 | |
*** vilobhmm has quit IRC | 23:22 | |
*** vilobhmm has joined #openstack-nova | 23:22 | |
*** changbl has quit IRC | 23:23 | |
*** jwcroppe has quit IRC | 23:23 | |
*** aginwala has quit IRC | 23:24 | |
*** changbl has joined #openstack-nova | 23:27 | |
*** kgalanov has joined #openstack-nova | 23:28 | |
*** liusheng has quit IRC | 23:28 | |
*** liusheng has joined #openstack-nova | 23:29 | |
*** alejandrito has quit IRC | 23:29 | |
*** vilobhmm has quit IRC | 23:31 | |
*** vilobhmm has joined #openstack-nova | 23:31 | |
*** changbl has quit IRC | 23:33 | |
openstackgerrit | Matt Riedemann proposed openstack/nova-specs: Make os-instance-actions read deleted instances https://review.openstack.org/248288 | 23:34 |
*** markvoelker has quit IRC | 23:34 | |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'hw_cpu_threads_policy' to ImageMetaProps https://review.openstack.org/202647 | 23:38 |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'cpu_policy' and 'cpu_threads_policy' fields https://review.openstack.org/202648 | 23:38 |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'hw:cpu_threads_policy=avoid' filtering https://review.openstack.org/202649 | 23:38 |
openstackgerrit | Stephen Finucane proposed openstack/nova: trivial: Add some logs to 'numa_topology_filter' https://review.openstack.org/202650 | 23:38 |
openstackgerrit | Stephen Finucane proposed openstack/nova: Add 'hw:cpu_threads_policy=require' scheduling https://review.openstack.org/202651 | 23:38 |
*** jwcroppe has joined #openstack-nova | 23:39 | |
*** pixelbeat has joined #openstack-nova | 23:40 | |
*** mylu_ has quit IRC | 23:43 | |
openstackgerrit | Lianhao Lu proposed openstack/nova-specs: Need a way to list compute node metric names https://review.openstack.org/180049 | 23:44 |
*** sfinucan has quit IRC | 23:44 | |
*** vilobhmm has quit IRC | 23:46 | |
*** vilobhmm has joined #openstack-nova | 23:46 | |
*** aginwala has joined #openstack-nova | 23:49 | |
*** ZZelle_ has quit IRC | 23:50 | |
*** liusheng has quit IRC | 23:50 | |
*** liusheng has joined #openstack-nova | 23:51 | |
*** rk4n has quit IRC | 23:53 | |
openstackgerrit | Lianhao Lu proposed openstack/nova-specs: Need a way to list compute node metric names https://review.openstack.org/180049 | 23:56 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!