| yoctozepto | hrw: ubuntu boots, cirros does not, both use the same kernel -> likely the bootloader is the difference | 07:00 |
|---|---|---|
| hrw | yoctozepto: nope | 08:16 |
| yoctozepto | please elaborate | 08:17 |
| hrw | yoctozepto: cirros uses 18.04 kernels | 08:17 |
| yoctozepto | oh! roger that | 08:17 |
| yoctozepto | so ubuntu 18.04 does not boot either then? | 08:17 |
| hrw | yoctozepto: and the problem is our use of 'max' cpu model which changes from qemu version to version | 08:17 |
| yoctozepto | ack, I see | 08:18 |
| hrw | so vote for https://review.opendev.org/c/openstack/kolla-ansible/+/844321 and it will work again | 08:18 |
| yoctozepto | approved | 08:19 |
| hrw | cool | 08:19 |
| yoctozepto | I did not see it because it was marked WIP | 08:19 |
| frickler | hrw: fwiw I tried with a cirros image built with jammy kernel, that didn't boot either | 08:19 |
| yoctozepto | hmm | 08:19 |
| yoctozepto | so the bootloader | 08:20 |
| hrw | yoctozepto: no. | 08:20 |
| yoctozepto | :D | 08:20 |
| yoctozepto | I'm not following | 08:20 |
| hrw | frickler: Debian 11 does not boot. Debian 11 + backports kernel does | 08:20 |
| hrw | the problem is cpu model not kernel/bootloader/etc | 08:20 |
| hrw | QEMU 7.0 got support for ARM Cortex-A76 and Neoverse N1 cpu cores. which are more advanced than Cortex-A72 qemu knew before | 08:21 |
| frickler | but if cpu model is broken, why do some images still boot? I'd like to build a cirros image that works with max | 08:21 |
| guesswhat | Guys? Why is octavia_network_type: "tenant" ( https://docs.openstack.org/kolla-ansible/latest/reference/networking/octavia.html ) not reliable for production ? | 08:22 |
| hrw | so amount of features of 'max' cpu increased. still it is virtual cpu which has all qemu supported arm cpu features but does not look like any existing one | 08:22 |
| hrw | I prefer to not spend time getting max working again as this is not something you use in real environments | 08:23 |
| hrw | I checked the issue, reported it upstream and provided fix for our project. | 08:24 |
| frickler | hrw: hmm, o.k., I guess that that's a valid argument. so you would consider using a72 a permanent solution? until maybe upstream fixes something? | 08:24 |
| hrw | and have some patches to k-a in a queue | 08:24 |
| hrw | frickler: a72 ftw. it is covering probably 90% of arm market, nearly all arm servers and most of cloud offering | 08:25 |
| hrw | especially when we use distro provided firmware/kernels which may not cover latest-greatest | 08:26 |
| yoctozepto | also, it does not really matter for k-a - we are not responsible for fixing qemu on aarch64; our job is done as it is | 08:26 |
| hrw | yep | 08:26 |
| hrw | sorting some of qemu things is my part of my other hat | 08:27 |
| guesswhat | mgoddard Question, Why is ctavia_network_type: "tenant" not reliable for production ? ( https://github.com/openstack/kolla-ansible/blame/aad331069def7faf09b046c95d9dca827a7bede2/doc/source/reference/networking/octavia.rst#L400 ) Thanks | 08:32 |
| mnasiadka | guesswhat: because it's relying on systemd units that configure the interface, which might fail - use on your own risk | 08:38 |
| mnasiadka | (or something else in OVS might fail - because these are basically OVSPorts) | 08:39 |
| guesswhat | mnasiadka: It fails for me, at least LBs are in Pending Upgrade or Error state after reboot, I tried to add Restart=on-failure to octavia-interface.service, but it randomly fails ofc... | 08:40 |
| guesswhat | Whats the other solution ? Everything includes OVSPorts, isnt? | 08:40 |
| mnasiadka | well, I just meant this solution does OVS components outside of Neutron | 08:41 |
| kevko | \o/ | 09:11 |
| kevko | morning | 09:11 |
| hrw | mnasiadka, kevko: https://review.opendev.org/c/openstack/kolla/+/844311 - please review | 09:26 |
| hrw | I am open for suggestions - especially for better argument name | 09:26 |
| kevko | going to | 09:27 |
| guesswhat | mnasiadka: and does exist some other solution for kolla? or its octavia_network_type: "tenant" or nothing, cuz it can not work with default install octavia_network_type: "provider" by default | 09:32 |
| mnasiadka | guesswhat: and why it cannot? | 09:33 |
| guesswhat | by default kolla creates vxlan lb-mgmt-net, octavia worker and healthcheck can not reach this network.. | 09:34 |
| opendevreview | Merged openstack/kolla-ansible master: Switch to Cortex-A72 cpu on AArch64 https://review.opendev.org/c/openstack/kolla-ansible/+/844321 | 09:36 |
| kevko | guesswhat: are u sure ? | 09:39 |
| yoctozepto | yay, aarch unbroken | 09:40 |
| yoctozepto | now time to backport | 09:40 |
| opendevreview | Radosław Piliszek proposed openstack/kolla-ansible stable/yoga: Switch to Cortex-A72 cpu on AArch64 https://review.opendev.org/c/openstack/kolla-ansible/+/844384 | 09:40 |
| opendevreview | Marcin Juszkiewicz proposed openstack/kolla-ansible stable/yoga: Switch to Cortex-A72 cpu on AArch64 https://review.opendev.org/c/openstack/kolla-ansible/+/844384 | 09:40 |
| opendevreview | Marcin Juszkiewicz proposed openstack/kolla-ansible stable/wallaby: Switch to Cortex-A72 cpu on AArch64 https://review.opendev.org/c/openstack/kolla-ansible/+/844385 | 09:41 |
| hrw | yoctozepto: GMTA ;D | 09:41 |
| yoctozepto | oh well | 09:41 |
| yoctozepto | though we don't need it before yoga | 09:41 |
| guesswhat | kevko: https://github.com/openstack/kolla-ansible/blob/master/ansible/roles/octavia/tasks/prepare.yml#L105-L107 its ommited if not specified, that means fallback to vxlan thus there is no chance that worker and healtcheck running on the management network could reach the lb-mgmt-net , correct? | 09:42 |
| guesswhat | mnasiadka ^ | 09:42 |
| yoctozepto | guessing we can merge for consistency | 09:42 |
| opendevreview | Radosław Piliszek proposed openstack/kolla-ansible stable/xena: Switch to Cortex-A72 cpu on AArch64 https://review.opendev.org/c/openstack/kolla-ansible/+/844386 | 09:42 |
| hrw | yoctozepto: need or not it is better to not use max | 09:43 |
| * hrw -> updating patch | 09:43 | |
| mnasiadka | guesswhat: so just set it properly in octavia_amp_network dict? | 09:46 |
| opendevreview | Marcin Juszkiewicz proposed openstack/kolla master: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844311 | 09:47 |
| guesswhat | mnasiadka: what should I set there? it will work only for another physnet ( not for management network ) and it has to be a flat, external network ... , otherwise one must use ports, veth pairs + bridge or routing with router... Octavia is quite confusing in kolla, I would like to see working example, as docs is not helpful in this case.. | 09:51 |
| yoctozepto | hrw: agreed | 09:51 |
| hrw | long time since my simple patch got so many comments | 09:53 |
| opendevreview | Marcin Juszkiewicz proposed openstack/kolla master: doc: get rid of remainders of binary images https://review.opendev.org/c/openstack/kolla/+/844411 | 10:04 |
| opendevreview | Merged openstack/kolla-ansible master: Add support for custom alert notification templates https://review.opendev.org/c/openstack/kolla-ansible/+/842100 | 10:05 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Remove configuration related to api_workers https://review.opendev.org/c/openstack/kolla-ansible/+/843130 | 10:09 |
| opendevreview | Bartosz Bezak proposed openstack/kolla-ansible stable/xena: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844416 | 11:03 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Remove nova related config from placement https://review.opendev.org/c/openstack/kolla-ansible/+/844420 | 11:26 |
| opendevreview | Bartosz Bezak proposed openstack/kolla-ansible stable/xena: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844416 | 11:41 |
| opendevreview | Merged openstack/kolla-ansible master: Split haproxy and keepalived restarts https://review.opendev.org/c/openstack/kolla-ansible/+/838930 | 11:45 |
| opendevreview | Bartosz Bezak proposed openstack/kolla-ansible stable/wallaby: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844425 | 12:00 |
| opendevreview | Bartosz Bezak proposed openstack/kolla-ansible stable/wallaby: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844425 | 12:02 |
| bhreddy | Hi All, I am facing issue while using merge_configs module. I am merging api-paste.ini in nova, neutron and other services, the issue is the colon ":" is getting replaced with equals to "=" in target container | 12:10 |
| bhreddy | for example, In the nova api-paste.ini template "/: meta" changed as "/ = meta" in container | 12:12 |
| kevko | bhreddy: did you fill a bug somewhere ? | 12:23 |
| frickler | mnasiadka: yoctozepto: one question you didn't answer regarding stable releases: should I include kayobe and/or ansible-collection-kolla? | 12:24 |
| frickler | o.k., ansible-collection-kolla doesn't have updates yet, so will skip that anyway for now | 12:26 |
| bhreddy | no I didn't file any bug | 12:28 |
| bhreddy | I thought there will be extra quotes or slashes we can use to retain colon | 12:29 |
| bhreddy | I tried like {{ ":" }} and {{ ':'}} and backslash followed by colon...nothing helped me | 12:29 |
| frickler | included kayobe now, let me know if I should drop it. https://review.opendev.org/c/openstack/releases/+/844438 Tag monthly kolla stable releases | 12:32 |
| SvenKieske | I got a question regarding https://review.opendev.org/c/openstack/kolla-ansible/+/802304 (allow designate_ns_record to be a list): is it already possible to define multiple different dns backend pools via kolla-ansible? | 12:41 |
| opendevreview | Merged openstack/kolla master: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844311 | 12:45 |
| opendevreview | Merged openstack/kolla master: doc: get rid of remainders of binary images https://review.opendev.org/c/openstack/kolla/+/844411 | 12:45 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Add proxysql support for database https://review.opendev.org/c/openstack/kolla-ansible/+/770215 | 12:52 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Edit services roles to support database sharding https://review.opendev.org/c/openstack/kolla-ansible/+/770216 | 12:52 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: [CI] Test ProxySQL with shards in the nova cells scenario https://review.opendev.org/c/openstack/kolla-ansible/+/770621 | 12:52 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: [DNM] Trigger cells job https://review.opendev.org/c/openstack/kolla-ansible/+/838916 | 12:52 |
| opendevreview | Pierre Riteau proposed openstack/kayobe stable/wallaby: Use net_mask filter instead of ansible's ipaddr https://review.opendev.org/c/openstack/kayobe/+/844387 | 13:12 |
| opendevreview | Pierre Riteau proposed openstack/kayobe stable/victoria: Use net_mask filter instead of ansible's ipaddr https://review.opendev.org/c/openstack/kayobe/+/844388 | 13:12 |
| opendevreview | Maksim Malchuk proposed openstack/kolla stable/yoga: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844455 | 13:29 |
| opendevreview | Maksim Malchuk proposed openstack/kolla stable/xena: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844456 | 13:30 |
| opendevreview | Maksim Malchuk proposed openstack/kolla stable/wallaby: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844457 | 13:30 |
| opendevreview | Merged openstack/kolla-ansible master: Remove configuration related to api_workers https://review.opendev.org/c/openstack/kolla-ansible/+/843130 | 13:38 |
| opendevreview | Verification of a change to openstack/kolla-ansible master failed: Remove nova related config from placement https://review.opendev.org/c/openstack/kolla-ansible/+/844420 | 13:38 |
| opendevreview | Pierre Riteau proposed openstack/kolla-ansible master: Increase openstack-exporter timeout to 45 seconds https://review.opendev.org/c/openstack/kolla-ansible/+/844459 | 13:55 |
| opendevreview | Pierre Riteau proposed openstack/kolla-ansible master: Increase openstack-exporter timeout to 45 seconds https://review.opendev.org/c/openstack/kolla-ansible/+/844459 | 13:57 |
| mnasiadka | frickler: yes, include kayobe for now :) | 14:05 |
| opendevreview | Maksim Malchuk proposed openstack/kolla stable/yoga: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844455 | 14:16 |
| opendevreview | Maksim Malchuk proposed openstack/kolla stable/wallaby: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844457 | 14:17 |
| opendevreview | Maksim Malchuk proposed openstack/kolla stable/xena: Allow to provide own repos.yaml file https://review.opendev.org/c/openstack/kolla/+/844456 | 14:18 |
| SvenKieske | I got an architectural question, because I'm currently revisiting some internal work on kolla-ansible: Would it be an interesting usecase to provide a facility to further split networking? We are entertaining the idea of a separate logging/monitoring network (currently the internal api network is used for that) | 14:34 |
| SvenKieske | we'd rather not only use this downstream and have to patch kolla-ansible into eternity, so I'd like to ask if there would be interest in providing templates to use a dedicated (virtual) logging network? | 14:35 |
| opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Remove nova related config from placement https://review.opendev.org/c/openstack/kolla-ansible/+/844420 | 14:36 |
| hrw | SvenKieske: I think that if it has a use and would come with documentation then should be fine | 15:42 |
| opendevreview | Verification of a change to openstack/kolla-ansible stable/xena failed: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844416 | 15:53 |
| opendevreview | Merged openstack/kolla-ansible stable/wallaby: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844425 | 18:10 |
| guesswhat | Can I add interface to br-int somehow automatically via ansible parameter? | 19:36 |
| guesswhat | kevko ping | 20:01 |
| guesswhat | @kevko | 20:26 |
| opendevreview | Merged openstack/kolla-ansible stable/xena: designate: allow designate_ns_record to be a list https://review.opendev.org/c/openstack/kolla-ansible/+/844416 | 20:34 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!