| *** anniec has quit IRC | 00:10 | |
| devananda | NobodyCam: something to take a peek at: https://review.openstack.org/#/c/50348/3 | 00:18 |
|---|---|---|
| * NobodyCam peeks | 00:19 | |
| *** anniec has joined #openstack-ironic | 00:19 | |
| NobodyCam | :-p your +1'ing scope creep for us lol | 00:19 |
| NobodyCam | but I see the need for it | 00:20 |
| devananda | i just commented on the related bug report | 00:21 |
| devananda | not scope creep. bug fix parity, which we need ... | 00:21 |
| NobodyCam | ya I know | 00:21 |
| devananda | but ya, I +1'd more work :) | 00:21 |
| NobodyCam | lol :) | 00:22 |
| devananda | also, i have a patch that might fix it for us, too ;) | 00:22 |
| NobodyCam | :) | 00:22 |
| devananda | not the code i meant to write this afternoon, but.... | 00:24 |
| devananda | NobodyCam: for pxe.vendor._continue_deploy() | 00:31 |
| devananda | NobodyCam: shouldn't this method set the provision_state= ACTIVE? | 00:31 |
| devananda | right now, it sets it to DEPLOYDONE.... that seems wrong ... | 00:31 |
| NobodyCam | off the top of my head I'd say ya | 00:32 |
| devananda | NobodyCam: compare to ComputeManager.do_node_deploy() which, for other drivers, will set prov_state to ACTIVE | 00:32 |
| devananda | it doesn't in the case of PXE driver, since deploy isn't done yet. but I think they should both end up in same staet .... | 00:32 |
| NobodyCam | makes sense to /me | 00:33 |
| devananda | k k | 00:33 |
| NobodyCam | humm I seem to have lock up my dev env. | 00:37 |
| devananda | i'm still having the issue where if i reboot seed, it borks neutron | 00:39 |
| devananda | makes working from cafes rather difficult | 00:39 |
| NobodyCam | :( | 00:40 |
| devananda | found a work around today | 00:40 |
| NobodyCam | ??? | 00:40 |
| devananda | delete the ports, subnets, and nets from neutron. then call setup-neutron again, then delete and re-add my baremetal node | 00:40 |
| devananda | and i can heat stack-create again | 00:41 |
| NobodyCam | so it wasnn't the no-proxy setting? | 00:44 |
| devananda | taht was an issue, too | 00:44 |
| devananda | but not the same issue | 00:44 |
| NobodyCam | ahh | 00:44 |
| *** yuikotakada has joined #openstack-ironic | 00:46 | |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Improve error handling in PXE _continue_deploy https://review.openstack.org/63037 | 01:03 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Improve error handling in PXE _continue_deploy https://review.openstack.org/63037 | 01:04 |
| devananda | fixing commit message | 01:04 |
| devananda | and with that, I'm signing off for a bit and doing some shopping :) | 01:05 |
| devananda | g'night all! | 01:05 |
| NobodyCam | night devananda | 01:05 |
| *** haomeng2 has joined #openstack-ironic | 01:10 | |
| *** haomeng2 has quit IRC | 01:18 | |
| *** rwsu has quit IRC | 01:18 | |
| *** rwsu has joined #openstack-ironic | 01:19 | |
| *** yfujioka has joined #openstack-ironic | 01:22 | |
| *** anniec has quit IRC | 01:23 | |
| *** anniec has joined #openstack-ironic | 01:24 | |
| *** anniec has quit IRC | 01:24 | |
| *** nosnos has joined #openstack-ironic | 01:38 | |
| *** itooon has quit IRC | 01:45 | |
| *** harlowja has quit IRC | 01:51 | |
| *** jcooley_ has quit IRC | 01:55 | |
| *** harlowja has joined #openstack-ironic | 02:03 | |
| *** itooon has joined #openstack-ironic | 02:04 | |
| *** rongze has joined #openstack-ironic | 02:12 | |
| *** jcooley_ has joined #openstack-ironic | 02:27 | |
| *** nosnos_ has joined #openstack-ironic | 02:30 | |
| *** nosnos has quit IRC | 02:33 | |
| *** jcooley_ has quit IRC | 02:37 | |
| *** rloo has quit IRC | 02:46 | |
| *** vkozhukalov has joined #openstack-ironic | 03:06 | |
| *** hstimer has quit IRC | 03:23 | |
| *** hstimer has joined #openstack-ironic | 03:23 | |
| *** jcooley_ has joined #openstack-ironic | 04:39 | |
| *** rongze has quit IRC | 04:44 | |
| *** jimjiang has quit IRC | 04:50 | |
| *** jimjiang has joined #openstack-ironic | 04:50 | |
| *** nosnos_ has quit IRC | 05:06 | |
| *** nosnos has joined #openstack-ironic | 05:07 | |
| *** urulama has joined #openstack-ironic | 05:07 | |
| *** itooon has quit IRC | 05:11 | |
| *** rongze has joined #openstack-ironic | 05:18 | |
| *** rongze has quit IRC | 05:23 | |
| *** harlowja is now known as harlowja_away | 05:40 | |
| *** harlowja_away is now known as harlowja | 05:59 | |
| openstackgerrit | Jenkins proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/61920 | 06:00 |
| *** darrylms has joined #openstack-ironic | 06:18 | |
| *** rongze has joined #openstack-ironic | 06:20 | |
| *** hstimer has quit IRC | 06:23 | |
| *** rongze has quit IRC | 06:25 | |
| *** rongze has joined #openstack-ironic | 06:27 | |
| *** rongze has quit IRC | 06:28 | |
| *** rongze has joined #openstack-ironic | 06:37 | |
| *** jimjiang has quit IRC | 06:39 | |
| *** jimjiang has joined #openstack-ironic | 06:40 | |
| *** vkozhukalov has quit IRC | 06:52 | |
| *** rongze has quit IRC | 07:49 | |
| *** rongze has joined #openstack-ironic | 07:49 | |
| *** prekarat has joined #openstack-ironic | 08:02 | |
| *** harlowja is now known as harlowja_away | 08:04 | |
| *** vkozhukalov has joined #openstack-ironic | 08:29 | |
| *** romcheg has joined #openstack-ironic | 08:38 | |
| *** kpavel has joined #openstack-ironic | 08:39 | |
| *** kpavel_ has joined #openstack-ironic | 08:40 | |
| openstackgerrit | A change was merged to openstack/ironic: Strip '/' from api_url string for PXE driver https://review.openstack.org/60205 | 08:42 |
| *** kpavel has quit IRC | 08:44 | |
| *** kpavel_ is now known as kpavel | 08:44 | |
| *** jimjiang has quit IRC | 08:48 | |
| *** jimjiang has joined #openstack-ironic | 08:49 | |
| *** jcooley_ has quit IRC | 08:54 | |
| openstackgerrit | Yuriy Zveryanskyy proposed a change to openstack/ironic: Allow clean reservation at update node https://review.openstack.org/55549 | 08:55 |
| *** ifarkas has joined #openstack-ironic | 09:01 | |
| *** yfujioka has quit IRC | 09:07 | |
| *** lucasagomes has joined #openstack-ironic | 09:24 | |
| *** agordeev has joined #openstack-ironic | 09:27 | |
| *** jimjiang has quit IRC | 09:30 | |
| *** jimjiang has joined #openstack-ironic | 09:31 | |
| *** kpavel_ has joined #openstack-ironic | 09:32 | |
| *** kpavel has quit IRC | 09:33 | |
| *** kpavel_ is now known as kpavel | 09:34 | |
| *** jimjiang has quit IRC | 09:36 | |
| *** jimjiang has joined #openstack-ironic | 09:37 | |
| *** jcooley_ has joined #openstack-ironic | 09:38 | |
| *** mdurnosvistov has joined #openstack-ironic | 09:47 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Convert trycmd() to oslo's processutils https://review.openstack.org/61709 | 09:53 |
| *** urulama has quit IRC | 09:56 | |
| *** mdurnosvistov_ has joined #openstack-ironic | 09:59 | |
| *** mdurnosvistov has quit IRC | 10:03 | |
| *** jcooley_ has quit IRC | 10:09 | |
| *** rongze has quit IRC | 10:11 | |
| lucasagomes | lifeless, ping | 10:14 |
| *** derekh has joined #openstack-ironic | 10:15 | |
| lucasagomes | lifeless, re ur comment at https://review.openstack.org/#/c/61960/2/ironic/tests/api/test_nodes.py | 10:15 |
| lifeless | lucasagomes: hi | 10:17 |
| lucasagomes | lifeless, the new patchset is not passing any interval to that function it's using the whatever default in the get_active_driver_dict() method... and testing if the conductor didn't check-in for a while and the driver is now inactive (out of the list) is already being tested by the tests on the get_active_driver_dict() | 10:17 |
| lucasagomes | I could add a similar test ofc, but sounds redundant to me | 10:18 |
| lucasagomes | I'm not very convinced | 10:18 |
| *** prekarat has quit IRC | 10:18 | |
| *** jimjiang has quit IRC | 10:19 | |
| *** jcooley_ has joined #openstack-ironic | 10:19 | |
| *** jimjiang has joined #openstack-ironic | 10:19 | |
| *** rongze has joined #openstack-ironic | 10:23 | |
| *** martyntaylor has joined #openstack-ironic | 10:28 | |
| lifeless | lucasagomes: so consider, if someone comes along and changes the code to pass in 1000000 as the interval | 10:29 |
| lifeless | lucasagomes: will a test fail ? | 10:29 |
| lifeless | lucasagomes: in fact, try that :). See what happens. | 10:29 |
| lucasagomes | hah ok | 10:30 |
| lifeless | If no test fails, and you agree that that code would be wrong, I think my point is made. | 10:30 |
| *** jimjiang has quit IRC | 10:30 | |
| lifeless | If on the other hadn you think that that code would be ok, then I grant you your point :). | 10:30 |
| *** yuikotakada has quit IRC | 10:31 | |
| *** jimjiang has joined #openstack-ironic | 10:31 | |
| *** greghaynes has quit IRC | 10:39 | |
| lifeless | lucasagomes: I'm heading off to bed -if the existing two tests catch this bad behaviour - *and* do so in a way tht is reasonably clear, I'll be happy to +2 it tomorrow morning | 10:40 |
| *** max_lobur_afk is now known as max_lobur | 10:40 | |
| *** greghaynes has joined #openstack-ironic | 10:41 | |
| *** ndipanov has joined #openstack-ironic | 10:43 | |
| lucasagomes | lifeless, right I will give it a think, just wondering cause in the API tests we mock the return value of that method so it do not introduce a new dependency on the dbapi layer in the api test module (it was requested in a previous review) | 10:44 |
| lucasagomes | so whatver interval you use won't actually matter | 10:44 |
| lucasagomes | lifeless, I will see here, have a good night :) | 10:44 |
| *** prekarat has joined #openstack-ironic | 10:46 | |
| *** martyntaylor has quit IRC | 10:48 | |
| lifeless | lucasagomes: right, so the general principle is that when you have an API you are mocking, you can assume it behaves well, but you can't assume that *you use it well* | 10:49 |
| lifeless | lucasagomes: if the API permits bad combinations of values, we need to be sure that we sent it good combinations | 10:49 |
| lifeless | and with that, gnight! | 10:49 |
| *** jcooley_ has quit IRC | 10:49 | |
| *** urulama has joined #openstack-ironic | 10:50 | |
| lucasagomes | lifeless, :) night | 10:54 |
| *** martyntaylor has joined #openstack-ironic | 10:59 | |
| *** max_lobur is now known as max_lobur_afk | 10:59 | |
| *** nosnos has quit IRC | 11:00 | |
| openstackgerrit | Imre Farkas proposed a change to openstack/ironic: Passing nodes more consistently https://review.openstack.org/62389 | 11:11 |
| *** rongze has quit IRC | 11:13 | |
| *** martyntaylor has quit IRC | 11:15 | |
| *** jcooley_ has joined #openstack-ironic | 11:32 | |
| *** romcheg1 has joined #openstack-ironic | 11:47 | |
| *** romcheg has quit IRC | 11:47 | |
| *** martyntaylor has joined #openstack-ironic | 11:50 | |
| *** jcooley_ has quit IRC | 12:04 | |
| *** rongze has joined #openstack-ironic | 12:04 | |
| *** prekarat has quit IRC | 12:17 | |
| *** prekarat has joined #openstack-ironic | 12:22 | |
| *** romcheg1 has quit IRC | 12:24 | |
| *** rongze_ has joined #openstack-ironic | 12:24 | |
| *** rongze has quit IRC | 12:28 | |
| *** max_lobur_afk is now known as max_lobur | 12:45 | |
| *** urulama has quit IRC | 12:46 | |
| *** jdob has joined #openstack-ironic | 12:49 | |
| *** anniec has joined #openstack-ironic | 12:51 | |
| *** kpavel_ has joined #openstack-ironic | 12:52 | |
| *** kpavel has quit IRC | 12:53 | |
| *** kpavel_ is now known as kpavel | 12:53 | |
| *** anniec_ has joined #openstack-ironic | 12:54 | |
| *** anniec has quit IRC | 12:55 | |
| *** anniec_ is now known as anniec | 12:55 | |
| openstackgerrit | Yuriy Zveryanskyy proposed a change to openstack/ironic: Improve error handling in validate_vendor_action https://review.openstack.org/63106 | 12:55 |
| *** jcooley_ has joined #openstack-ironic | 12:57 | |
| *** linggao has joined #openstack-ironic | 13:04 | |
| *** linggao_ has joined #openstack-ironic | 13:05 | |
| *** linggao has quit IRC | 13:05 | |
| openstackgerrit | Yuriy Zveryanskyy proposed a change to openstack/ironic: Improve error handling in validate_vendor_action https://review.openstack.org/63106 | 13:20 |
| *** jcooley_ has quit IRC | 13:27 | |
| *** prekarat has left #openstack-ironic | 13:32 | |
| *** rloo has joined #openstack-ironic | 13:33 | |
| *** romcheg has joined #openstack-ironic | 13:35 | |
| *** urulama has joined #openstack-ironic | 13:55 | |
| *** darrylms has quit IRC | 13:56 | |
| *** darrylms has joined #openstack-ironic | 13:56 | |
| *** urulama_ has joined #openstack-ironic | 13:56 | |
| *** derekh has quit IRC | 14:00 | |
| *** urulama has quit IRC | 14:00 | |
| *** lucasagomes has quit IRC | 14:06 | |
| *** lucasagomes has joined #openstack-ironic | 14:06 | |
| *** lucasagomes has quit IRC | 14:07 | |
| *** jcooley_ has joined #openstack-ironic | 14:18 | |
| *** jimjiang has quit IRC | 14:18 | |
| *** jimjiang has joined #openstack-ironic | 14:19 | |
| *** jimjiang has quit IRC | 14:24 | |
| *** jimjiang has joined #openstack-ironic | 14:25 | |
| *** mattymo has quit IRC | 14:32 | |
| *** darrylms has quit IRC | 14:32 | |
| *** mattymo has joined #openstack-ironic | 14:32 | |
| *** darrylms has joined #openstack-ironic | 14:32 | |
| *** viktors has joined #openstack-ironic | 14:35 | |
| GheRivero | hi all | 14:46 |
| *** openstack has joined #openstack-ironic | 14:49 | |
| *** jcooley_ has quit IRC | 14:52 | |
| *** matty_dubs|gone is now known as matty_dubs | 14:59 | |
| *** blamar has joined #openstack-ironic | 15:08 | |
| NobodyCam | Good Morning Ironic | 15:11 |
| GheRivero | morning | 15:12 |
| yuriyz | Morning Ironic | 15:13 |
| max_lobur | morning Everyone | 15:14 |
| rloo | hi all! | 15:14 |
| NobodyCam | morning yuriyz GheRivero and max_lobur | 15:14 |
| *** blamar has quit IRC | 15:16 | |
| *** blamar has joined #openstack-ironic | 15:17 | |
| *** kpavel has quit IRC | 15:24 | |
| *** vkozhukalov has quit IRC | 15:40 | |
| *** viktors has left #openstack-ironic | 15:42 | |
| *** rongze has joined #openstack-ironic | 15:42 | |
| *** jcooley_ has joined #openstack-ironic | 15:44 | |
| *** rongze_ has quit IRC | 15:45 | |
| *** anniec has quit IRC | 16:13 | |
| NobodyCam | oh sorry rloo I missed your hello | 16:14 |
| NobodyCam | :) | 16:14 |
| rloo | no worries NobodyCam! (sort of like the John boy thing...) | 16:14 |
| NobodyCam | :) | 16:14 |
| *** rongze has quit IRC | 16:15 | |
| *** jcooley_ has quit IRC | 16:16 | |
| *** rongze has joined #openstack-ironic | 16:16 | |
| *** jimjiang has quit IRC | 16:22 | |
| *** rongze has quit IRC | 16:22 | |
| *** jimjiang has joined #openstack-ironic | 16:22 | |
| *** rongze has joined #openstack-ironic | 16:22 | |
| *** urulama_ has quit IRC | 16:22 | |
| *** jcooley_ has joined #openstack-ironic | 16:33 | |
| *** jcooley_ has quit IRC | 16:52 | |
| *** mattymo has quit IRC | 16:52 | |
| *** ndipanov has quit IRC | 16:52 | |
| *** jcooley_ has joined #openstack-ironic | 16:55 | |
| *** mattymo has joined #openstack-ironic | 16:55 | |
| *** ndipanov has joined #openstack-ironic | 16:55 | |
| *** Maha has joined #openstack-ironic | 16:57 | |
| *** jcooley_ has quit IRC | 16:58 | |
| *** mattymo has quit IRC | 16:58 | |
| *** ndipanov has quit IRC | 16:58 | |
| *** martyntaylor has quit IRC | 17:00 | |
| devananda | morning, all! | 17:01 |
| NobodyCam | good morning devananda | 17:02 |
| GheRivero | morning devananda | 17:02 |
| rloo | mornin' devananda | 17:02 |
| *** jimjiang has quit IRC | 17:03 | |
| *** jimjiang has joined #openstack-ironic | 17:04 | |
| *** jcooley_ has joined #openstack-ironic | 17:04 | |
| *** mattymo has joined #openstack-ironic | 17:04 | |
| *** ndipanov has joined #openstack-ironic | 17:04 | |
| yuriyz | morning devananda | 17:05 |
| *** jcooley_ has quit IRC | 17:09 | |
| *** mattymo has quit IRC | 17:09 | |
| *** ndipanov has quit IRC | 17:09 | |
| yuriyz | devananda, about node's reservation and dead conductors, I think 'reservation' field should contains not only hostname but also process id (pid) of conductor | 17:10 |
| *** jimjiang has quit IRC | 17:10 | |
| devananda | yuriyz: interesting. why? | 17:10 |
| yuriyz | for cases if conductor restarted | 17:10 |
| yuriyz | and if we only have one conductor | 17:11 |
| *** jimjiang has joined #openstack-ironic | 17:11 | |
| yuriyz | and new field 'pid' in conductors table | 17:11 |
| yuriyz | for easy checks if reservation field is from dead conductor | 17:13 |
| devananda | yuriyz: are you suggesting that after restart, new conductor process would not be able to unlock old conductor's locks without using the lock-breaking mechanism? | 17:13 |
| yuriyz | yes | 17:13 |
| yuriyz | because new already registered | 17:13 |
| yuriyz | if restarted | 17:14 |
| devananda | yuriyz: in this approach, drivers would need to "take over" resources controlled by the previous instance of conductor process on that host | 17:14 |
| *** jcooley_ has joined #openstack-ironic | 17:15 | |
| *** mattymo has joined #openstack-ironic | 17:15 | |
| *** ndipanov has joined #openstack-ironic | 17:15 | |
| devananda | yuriyz: i think it is sufficient for all delete/tear_down/clean_up/etc methods to be idempotent. if conductor is restarted and an in-progress task failed, user can issue a delete on that resource, new conductor proces will still be able to get lock, and delete will happen | 17:15 |
| devananda | yuriyz: in your approach, if we require that lock-breaking can only happen for locks held by dead conductors, then heartbeat must also be based on PID | 17:16 |
| NobodyCam | bbt and more coffee time... brb | 17:17 |
| *** hemna has joined #openstack-ironic | 17:19 | |
| yuriyz | I will try todo a patch tomorrow only for manual unlocking w/o break compatibility | 17:21 |
| *** max_lobur is now known as max_lobur_afk | 17:21 | |
| *** jcooley_ has quit IRC | 17:22 | |
| devananda | yuriyz: i'm not convinced that we need pid-based locking (let alone that it's a priority for Icehouse-2 timeframe) | 17:26 |
| *** GheRivero has quit IRC | 17:28 | |
| yuriyz | ok, let it be proof-of-concept | 17:30 |
| devananda | yuriyz: I think there are more important things that we need to get working before we can apply for graduation / integration / first release | 17:33 |
| yuriyz | ok | 17:33 |
| * devananda steps away for a phone call | 17:33 | |
| *** rongze has quit IRC | 17:39 | |
| *** rongze has joined #openstack-ironic | 17:40 | |
| *** GheRivero has joined #openstack-ironic | 17:41 | |
| *** rongze has quit IRC | 17:44 | |
| *** vkozhukalov has joined #openstack-ironic | 17:46 | |
| *** mdurnosvistov_ has quit IRC | 17:50 | |
| NobodyCam | wost breakfast walkies | 17:50 |
| NobodyCam | bbiaf | 17:50 |
| *** Maha has quit IRC | 17:55 | |
| *** harlowja_away is now known as harlowja | 18:04 | |
| *** urulama has joined #openstack-ironic | 18:23 | |
| *** urulama has quit IRC | 18:27 | |
| *** ifarkas has quit IRC | 18:34 | |
| NobodyCam | makes a bagel | 18:36 |
| *** rongze has joined #openstack-ironic | 18:40 | |
| *** rloo has quit IRC | 18:42 | |
| *** rloo has joined #openstack-ironic | 18:43 | |
| *** rloo has quit IRC | 18:44 | |
| *** rloo has joined #openstack-ironic | 18:44 | |
| *** mdurnosvistov has joined #openstack-ironic | 18:45 | |
| *** rongze has quit IRC | 18:45 | |
| *** jimjiang has quit IRC | 18:51 | |
| *** jimjiang has joined #openstack-ironic | 18:51 | |
| *** jimjiang has quit IRC | 18:58 | |
| *** jimjiang has joined #openstack-ironic | 18:59 | |
| *** jimjiang has quit IRC | 19:03 | |
| *** jimjiang has joined #openstack-ironic | 19:04 | |
| *** jimjiang has quit IRC | 19:09 | |
| *** jimjiang has joined #openstack-ironic | 19:09 | |
| *** jimjiang has quit IRC | 19:15 | |
| *** jimjiang has joined #openstack-ironic | 19:16 | |
| *** jdob_ has joined #openstack-ironic | 19:21 | |
| * NobodyCam reboots his seed vm | 19:21 | |
| *** ndipanov has quit IRC | 19:31 | |
| *** jdob_ has quit IRC | 19:40 | |
| *** jdob has quit IRC | 19:40 | |
| *** jdob has joined #openstack-ironic | 19:40 | |
| NobodyCam | devananda: your neturn isse prevents you from deploying? | 19:49 |
| NobodyCam | or pxe booting I should say? | 19:50 |
| devananda | eh? | 19:50 |
| NobodyCam | the neutron issue your having | 19:50 |
| NobodyCam | for me heat stack-create is saying "create_complete" yet the node is saying no network devices | 19:51 |
| NobodyCam | just checking if that is the issue your see | 19:52 |
| NobodyCam | *seeing | 19:52 |
| * NobodyCam tries again.. from the beginning | 19:55 | |
| * NobodyCam makes more coffee while booting seed | 20:04 | |
| devananda | NobodyCam: do "neutron agent-list" -- if you see duplicated agents, you've hit the same bug | 20:06 |
| devananda | basically, if you reboot seed VM, neutron breaks by duplicating agents and mangling its net config. | 20:06 |
| devananda | fix is to a) rebuild seed, or b) just reconfigure neutron | 20:07 |
| NobodyCam | ack. will check that if I run in to it again | 20:08 |
| *** Haomeng has quit IRC | 20:17 | |
| *** Haomeng has joined #openstack-ironic | 20:17 | |
| *** urulama has joined #openstack-ironic | 20:24 | |
| *** urulama has quit IRC | 20:29 | |
| * devananda lunches | 20:38 | |
| NobodyCam | :) | 20:38 |
| *** rloo has quit IRC | 20:48 | |
| *** rloo has joined #openstack-ironic | 20:48 | |
| *** rloo has quit IRC | 20:49 | |
| *** rloo has joined #openstack-ironic | 20:49 | |
| *** vkozhukalov has quit IRC | 20:51 | |
| *** epim has joined #openstack-ironic | 20:53 | |
| * NobodyCam looks for some food stuffs.. back in a bit | 20:59 | |
| *** michchap_ has joined #openstack-ironic | 21:04 | |
| *** michchap has quit IRC | 21:06 | |
| *** mdurnosvistov has quit IRC | 21:34 | |
| *** jimjiang has quit IRC | 21:42 | |
| *** jimjiang has joined #openstack-ironic | 21:43 | |
| *** jdob has quit IRC | 21:46 | |
| *** linggao_ has quit IRC | 21:47 | |
| *** rloo has quit IRC | 21:53 | |
| *** rloo has joined #openstack-ironic | 21:54 | |
| * devananda is back | 21:56 | |
| lifeless | devananda: http://allthingsd.com/20131219/calxeda-chipmaker-who-sought-to-bring-arm-to-servers-has-shut-down/ | 22:02 |
| devananda | lifeless: erm..... wtf | 22:03 |
| *** epim has quit IRC | 22:04 | |
| devananda | lifeless: thanks. quite interesting | 22:08 |
| lifeless | indeed | 22:11 |
| *** jimjiang has quit IRC | 22:12 | |
| *** jimjiang has joined #openstack-ironic | 22:13 | |
| *** jcooley_ has joined #openstack-ironic | 22:19 | |
| hemna | devananda, do you have the url for the nova ironic patch on review.openstack.org ? | 22:19 |
| devananda | hemna: https://review.openstack.org/#/c/51328/7 | 22:19 |
| hemna | thanks | 22:19 |
| NobodyCam | :) | 22:20 |
| ekarlso | Hmms, moonshot killed calxeda ? | 22:21 |
| lifeless | no | 22:22 |
| lifeless | VC killed calxeda | 22:22 |
| ekarlso | VC ? | 22:22 |
| lifeless | venture capital | 22:22 |
| ekarlso | ;) | 22:25 |
| *** urulama has joined #openstack-ironic | 22:25 | |
| *** urulama has quit IRC | 22:30 | |
| *** darrylms has quit IRC | 22:30 | |
| *** darrylms has joined #openstack-ironic | 22:31 | |
| hemna | so, it looks like I need to update my BP on the cinder integration. | 22:37 |
| hemna | if the nova patch lands for ironic, then nova will still be orchestrating the whole process | 22:37 |
| devananda | hemna: yes. that's the plan | 22:38 |
| NobodyCam | hemna: its a bit away from that | 22:38 |
| hemna | in which case ironic doesn't need to do anything with Cinder from what I can tell | 22:38 |
| NobodyCam | landing that is | 22:38 |
| devananda | hemna: ironic still needs to do the hw-bios config | 22:38 |
| hemna | Nova already knows how to talk to Cinder to call cinder's drivers to do the volume attach, and collect the initiator/target info | 22:38 |
| devananda | hemna: which means it needs to receive the cinder volume info from nova and pass it to the bios | 22:38 |
| hemna | ironic will just need the hw-bios layer | 22:38 |
| devananda | hemna: right. and nova/cinder should do that stuff still | 22:38 |
| hemna | yah | 22:38 |
| hemna | nova/cinder already do that | 22:39 |
| devananda | hemna: ironic needs the bios layer. and some API / glue to receive that info from nova/cinder | 22:39 |
| hemna | we'll just have to update the ironic driver | 22:39 |
| hemna | to call the ironic API to do 2 things | 22:39 |
| hemna | 1) to get the connector info (the initiator iqn / wwns) | 22:39 |
| *** matty_dubs is now known as matty_dubs|gone | 22:39 | |
| hemna | 2) pass the cinder volume target info into ironic, which calls the hw-bios layer to configure the bios network cards to boot from san. | 22:40 |
| hemna | easy! | 22:40 |
| hemna | :P | 22:40 |
| devananda | hemna: sounds about right | 22:40 |
| hemna | the hw-bios layer needs to get defined and drivers written for it. | 22:41 |
| hemna | so, I'll update the BP I wrote to reflect this new understanding (of mine) | 22:41 |
| devananda | :) | 22:42 |
| hemna | what we could do for Icehouse is try and define the API for Part 1 | 22:42 |
| devananda | hemna: will ironic need to do any proxying of traffic? -- I doubt it, but some of the nova-baremetal volume code did that at one point in time | 22:43 |
| hemna | and generate an IQN for the BM node on the ironic host | 22:43 |
| hemna | since the BM node isn't up, we'll have to generate an iSCSI iqn for it. | 22:43 |
| devananda | hmm | 22:43 |
| hemna | once the hw-bios layer is in place we'll be able to query for the FC wwns | 22:44 |
| devananda | in my naivete, that seems to suggest ironic will need to do something to forward/proxy the iSCSI traffic ?? | 22:44 |
| hemna | nah | 22:44 |
| devananda | k | 22:44 |
| hemna | when open-iscsi is installed on linux, it generates an initiator.iqn and puts it in /etc/iscsi/iniaitorname.iscsi | 22:45 |
| hemna | since the host os isn't up on the BM node | 22:45 |
| hemna | that's not there | 22:45 |
| devananda | ah | 22:45 |
| hemna | but the ironic node can call iscsi-iname and generate one for the BM node's bios | 22:46 |
| devananda | so ironic-conductor will generate that and then need to supply it to the host OS via cloud-init | 22:46 |
| devananda | ... or stick it in the BIOS | 22:46 |
| hemna | and that gets passed back to nova at ironic/driver.py:get_volume_connector() time | 22:46 |
| hemna | yah | 22:46 |
| devananda | cool | 22:46 |
| hemna | I think a host can have more than one IQN | 22:46 |
| devananda | NobodyCam: following that ^ ? :) | 22:46 |
| hemna | so we could have an IQN just for the BIOS at boot from SAN config time | 22:47 |
| hemna | we could do this for Icehouse | 22:47 |
| hemna | regardless of the hw-bios layer existing | 22:47 |
| devananda | without hw-bios layer yet, it seems like we should be able to pass that to a node via cloud-init | 22:49 |
| devananda | so baremetal nodes could have cinder volumes | 22:49 |
| devananda | even if they couldn't boot from cinder vols yet | 22:49 |
| hemna | that's prior to the OS booting ? | 22:49 |
| devananda | cloud-init is part of the init process of the OS | 22:49 |
| hemna | if the OS is up on the node, then (if it's linux), the iqn lives in /etc/iscsi/initatorname.iscsi | 22:50 |
| devananda | the parameterization of that cloud instance | 22:50 |
| devananda | eg, how a cloud instance gets its IP and hostname | 22:50 |
| hemna | I suppose cloud-init could then dump that iqn inside the /etc/iscsi/iniatorname.iscsi | 22:50 |
| devananda | yep | 22:50 |
| devananda | i suspect it could | 22:50 |
| devananda | along with any credentials we needed to inject | 22:50 |
| hemna | yah | 22:51 |
| hemna | I guess the other side of the cinder integration is how to attach a volume to the node post OS install | 22:51 |
| devananda | so, i dont know that ironic /can/ do that | 22:51 |
| hemna | that's a bit more difficult | 22:51 |
| hemna | heh | 22:51 |
| devananda | there's no agent under the OS | 22:51 |
| devananda | right? | 22:52 |
| hemna | especially if we don't have a process running on that BM node | 22:52 |
| devananda | wedont. period. | 22:52 |
| hemna | if we had an agent, we could do it. | 22:52 |
| devananda | we have the BMC and may be able to control it via IPMI or some other out-of-band mgmt | 22:52 |
| devananda | eg, an iLO driver to config the BIOS | 22:52 |
| devananda | which may or may not enable a real-time attachment of a network storage device without rebooting the host .... | 22:53 |
| hemna | yah, but we'd have to ask the kernel on the BM node if it found the new LUN showing up from the cinder backend | 22:53 |
| devananda | heh | 22:53 |
| devananda | right | 22:53 |
| devananda | AIUI, after init is done, we can provide the volume but not actually mount it | 22:54 |
| hemna | FC is a little easier in this respect | 22:55 |
| hemna | there is no session that needs to be maintained between the initiator and the target | 22:55 |
| hemna | so we could possibly do a blind attach | 22:55 |
| hemna | cinder exports the volume | 22:56 |
| hemna | if the initiator and target are zoned up, the LUN *should* just show up. | 22:56 |
| hemna | iscsi on the other hand requires a session to be initiated | 22:56 |
| hemna | and CHAP credentials to be exchanged | 22:56 |
| hemna | so we can't do an attach w/o being on the BM node | 22:57 |
| devananda | hemna: so we probably need some script(s) added to the disk image that will initiate the iscsi connection during first boot | 22:58 |
| devananda | hemna: those would need to work with diskimage-builder. i'm not sure this quite fits in the tripleo-image-elements repo, but something similar | 22:59 |
| hemna | that only works if cinder has only 1 backend and it never changes | 22:59 |
| devananda | hemna: i maen, the script(s) would pull the particular details from cloud-init during first boot | 22:59 |
| hemna | cinder can have N backends (arrays) that it manages, and if the BM node wants to do iSCSI attach to any of those, it has to create an iSCSI session with it. | 23:00 |
| devananda | hemna: nova // cinder // ironic would stash the IQN, CHAP credentials, or what ever else, in cloud metadata before the node powers on | 23:00 |
| devananda | then the script would pull that metadata and configure the node before initializing the iSCSI conenction | 23:00 |
| devananda | eg, write out /etc/iscsi/initiatorname.iscsi | 23:00 |
| devananda | however | 23:00 |
| hemna | yah that works at boot time | 23:01 |
| devananda | we wouldn't be able to change that info without re-deploying the node | 23:01 |
| hemna | yah that's ok. | 23:01 |
| devananda | k k | 23:01 |
| hemna | as long as it has an iqn | 23:01 |
| hemna | the problem is still volume attach post OS first boot. | 23:01 |
| hemna | we might be able to do a attach at first boot. | 23:01 |
| hemna | make it part of that same script | 23:01 |
| hemna | but, we still couldn't unattach | 23:02 |
| hemna | w/o being on the node later | 23:02 |
| NobodyCam | brb | 23:02 |
| hemna | or | 23:02 |
| hemna | just be ok with dangling LUNs | 23:02 |
| devananda | for iSCSI, it sounds like that's not feasible -- in a general case. it may be possible in special cases, eg. if we know the OS image contains something which periodically checks for new metadata | 23:02 |
| devananda | hemna: eg something like os-refresh-config / os-apply-config | 23:02 |
| devananda | lifeless: this conversation may be very interesting to you ^^ (long scroll back) | 23:02 |
| hemna | the other option is to write an agent, and let the BM node owner install it if they want to attach/detach volumes ? | 23:03 |
| hemna | that's kinda icky | 23:04 |
| devananda | hemna: agent is the only way to do this with baremetal. tripleo has a set of scripts that may be able to do this without really being "agents" | 23:04 |
| hemna | yah I agree | 23:05 |
| devananda | hemna: https://github.com/openstack/os-refresh-config | 23:05 |
| hemna | the good thing is, most of the agent code needed to handle the attach already exists. | 23:05 |
| hemna | I wrote a subproject in Cinder called Brick | 23:05 |
| hemna | it handles the LUN discovery on a host for iSCSI and FC | 23:05 |
| hemna | it also handles detaching (removing the LUN from the kernel) | 23:05 |
| hemna | we are pulling brick out of Cinder in Icehouse and making it it's own pypi library. | 23:06 |
| devananda | nice | 23:06 |
| hemna | at least that's the plan :) | 23:06 |
| devananda | hemna: also, instead of requireing BM node owner to install something, it could be baked into the disk image that was deployed onto the BM node | 23:07 |
| hemna | I think that'd be best | 23:07 |
| devananda | s/could/should/ :) | 23:07 |
| *** harlowja is now known as harlowja_away | 23:12 | |
| NobodyCam | devananda: fyi: line 139 / 147 http://paste.openstack.org/show/s3RpsMVCgoKhvcUUfa8L/ | 23:20 |
| devananda | new bug for anyone looking for low-hanging-fruit - https://bugs.launchpad.net/ironic/+bug/1262912 | 23:20 |
| devananda | NobodyCam: nice! nova is passing stuff in :) | 23:22 |
| NobodyCam | :) | 23:23 |
| devananda | NobodyCam: should the pxe_deploy_* bits be parsed out of glance by nova or ironic? | 23:23 |
| NobodyCam | devananda: idealy ironic. I added so I could validate | 23:24 |
| NobodyCam | well I pulled them from the flavor | 23:25 |
| NobodyCam | or at lest should have | 23:26 |
| NobodyCam | | extra_specs | {u'cpu_arch': u'i386', u'baremetal:deploy_kernel_id': u'22d8a929-1f23-4584-9d5c-f13773319ac5', | 23:26 |
| NobodyCam | ^ from nova flavor show | 23:26 |
| devananda | right | 23:26 |
| devananda | so nova needs to pass that info to ironic | 23:26 |
| devananda | in the driver_info | 23:26 |
| NobodyCam | yep | 23:27 |
| NobodyCam | ironic will get the user image K&R | 23:28 |
| NobodyCam | from glance | 23:29 |
| NobodyCam | unless we change and put that in chassis | 23:29 |
| NobodyCam | brb gota grap chairs from out side looks like rain soon... | 23:33 |
| NobodyCam | grab even | 23:33 |
| *** romcheg has quit IRC | 23:38 | |
| *** romcheg has joined #openstack-ironic | 23:39 | |
| *** romcheg has quit IRC | 23:40 | |
| devananda | odd. i thought we fixed the client-dumps-the-whole-traceback issue | 23:41 |
| NobodyCam | ??? | 23:43 |
| *** rongze has joined #openstack-ironic | 23:43 | |
| *** rongze has quit IRC | 23:48 | |
| *** harlowja_away is now known as harlowja | 23:57 | |
| devananda | https://bugs.launchpad.net/ironic/+bug/1244747 | 23:57 |
| devananda | max_lobur_afk: ^ | 23:57 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!