| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2022-02-09 | |||
| 11:07:55 | opendevreview | Felix Huettner proposed openstack/nova master: Gracefull recovery when attaching volume fails https://review.opendev.org/c/openstack/nova/+/828516 | |
| 11:22:06 | tobias-urdin | thx, you guys ok with backporting with the rename of the internal function in libvirt driver? | |
| 11:22:36 | sean-k-mooney | its not part of the driver api so it should be fine | |
| 11:22:58 | tobias-urdin | ack | |
| 11:24:05 | gibi | yepp, feel free to propose the backports | |
| 11:24:54 | opendevreview | Tobias Urdin proposed openstack/nova stable/wallaby: libvirt: Add announce-self post live-migration workaround https://review.opendev.org/c/openstack/nova/+/825178 | |
| 11:56:38 | sean-k-mooney | gibi: dmitriis im +2 on all the off-path backend series bar the docs patch at the end | |
| 11:57:29 | dmitriis | sean-k-mooney: ack, I am finishing up some changes to the doc patch | |
| 11:57:56 | dmitriis | interestingly, the mlx5 driver doesn't allow "trusted" to be set on a VF so I'll document that as a limitation | |
| 11:58:37 | dmitriis | i.e. some drivers may not allow you to do that which will limit the use of the promisc mode presumably - I'll explore more | |
| 11:59:13 | dmitriis | RTNETLINK answers: Operation not permitted | |
| 11:59:13 | dmitriis | $ sudo ip link set enp130s0f0 vf 0 trust off ; echo $? | |
| 11:59:13 | dmitriis | 2 | |
| 11:59:13 | dmitriis | $ sudo ip link set enp130s0f0 vf 0 trust on ; echo $? | |
| 11:59:15 | dmitriis | 2 | |
| 11:59:30 | sean-k-mooney | dmitriis: ya it depend on the driver | |
| 11:59:40 | sean-k-mooney | but honestly i hate that feature so meh | |
| 12:00:14 | dmitriis | sean-k-mooney: yeah, just trying to go through various combinations that may break | |
| 12:00:14 | sean-k-mooney | promisc mode is a security threat and it degrades performance | |
| 12:00:19 | dmitriis | agreed | |
| 12:00:56 | sean-k-mooney | at some point i woudl liek to stop using the binding_profile to enabel that too and actully have a api extension instead | |
| 12:01:37 | sean-k-mooney | this is currently the only value that can be set in the binding_profile that is not set by nova | |
| 12:01:51 | sean-k-mooney | and as a result we have to filter what we store in it | |
| 12:01:59 | sean-k-mooney | when clearing it as you saw | |
| 12:02:08 | dmitriis | yes, it feels very unnatural | |
| 12:04:21 | sean-k-mooney | ok i have an apointment in 40 mins so i better go. ill be back in a few hours | |
| 12:04:27 | dmitriis | sean-k-mooney: thanks a lot for all the support on this patch series - it's been quite a challenge to get it to this point | |
| 12:04:42 | sean-k-mooney | ack your almost there | |
| 12:29:10 | gibi | sean-k-mooney: I will check it a bit later | |
| 12:57:32 | opendevreview | Felix Huettner proposed openstack/nova master: Gracefull recovery when attaching volume fails https://review.opendev.org/c/openstack/nova/+/828516 | |
| 13:03:30 | artom | gdi test_tagged_attachment is failing in nova-next on xena as well | |
| 13:03:43 | opendevreview | Felix Huettner proposed openstack/nova master: Gracefull recovery when attaching volume fails https://review.opendev.org/c/openstack/nova/+/828516 | |
| 13:10:38 | opendevreview | Artom Lifshitz proposed openstack/nova stable/wallaby: Add nova-ovs-hybrid-plug job https://review.opendev.org/c/openstack/nova/+/828418 | |
| 13:10:39 | opendevreview | Artom Lifshitz proposed openstack/nova stable/wallaby: Revert "Revert resize: wait for events according to hybrid plug" https://review.opendev.org/c/openstack/nova/+/828419 | |
| 13:16:34 | opendevreview | Merged openstack/nova master: Cleanup old resize instances dir before resize https://review.opendev.org/c/openstack/nova/+/827865 | |
| 13:17:00 | opendevreview | Dmitrii Shcherbakov proposed openstack/nova master: Document remote-managed port usage considerations https://review.opendev.org/c/openstack/nova/+/827513 | |
| 13:30:17 | opendevreview | Alexey Stupnikov proposed openstack/nova master: Fix clean-up process for aborted queued live migrations https://review.opendev.org/c/openstack/nova/+/828374 | |
| 13:33:21 | opendevreview | Tobias Urdin proposed openstack/nova stable/xena: Cleanup old resize instances dir before resize https://review.opendev.org/c/openstack/nova/+/828407 | |
| 14:26:54 | gibi | gmann: can I fix this bug outside of an api microversion? or this also needs an api microversion to be fixed? https://review.opendev.org/c/openstack/placement/+/826491/7#message-af8324d5631836fe3ec8052192994f9a42e00492 | |
| 14:27:31 | gibi | sean-k-mooney, bauzas, melwitt ^^ | |
| 14:28:51 | sean-k-mooney | we have not done a release yet with any traits so if its a bug in the any traits impl i dont think we need a microversion | |
| 14:28:56 | sean-k-mooney | just reading now | |
| 14:29:41 | opendevreview | Artom Lifshitz proposed openstack/nova stable/xena: Add nova-ovs-hybrid-plug job https://review.opendev.org/c/openstack/nova/+/828413 | |
| 14:29:42 | opendevreview | Artom Lifshitz proposed openstack/nova stable/xena: Revert "Revert resize: wait for events according to hybrid plug" https://review.opendev.org/c/openstack/nova/+/828414 | |
| 14:29:43 | opendevreview | Artom Lifshitz proposed openstack/nova stable/xena: skip test_tagged_attachment in nova-next https://review.opendev.org/c/openstack/nova/+/828542 | |
| 14:30:14 | gibi | sean-k-mooney: this is a bug without any traits | |
| 14:30:42 | sean-k-mooney | oh sorry i was ment to read the comment not the commit message | |
| 14:30:45 | gibi | sean-k-mooney: today placement accepts !!!!!!FOO as a forbidden trait as it strips all the ! | |
| 14:30:53 | sean-k-mooney | i didnt scorll down | |
| 14:31:09 | sean-k-mooney | ah | |
| 14:31:30 | sean-k-mooney | well i thinkn that shoudl be fixable if you make it a 400 | |
| 14:31:46 | sean-k-mooney | given ! is not allowed in a trait other then to mark it forbiden | |
| 14:32:10 | sean-k-mooney | gibi: is there an error in our regex for validation or do we only validate that in nova | |
| 14:32:29 | gibi | placement does not validate that with a schema | |
| 14:32:43 | sean-k-mooney | ah ok https://github.com/openstack/nova/blob/master/nova/api/validation/extra_specs/traits.py | |
| 14:32:52 | sean-k-mooney | so we do if you enable the extra spec validation | |
| 14:33:12 | sean-k-mooney | but i guess images and direct request could still contian the invalid traits | |
| 14:34:34 | gibi | this is pretty similar to the repeate query arg bug, something that as HTTP200 would be HTTP400 if we fix the bug so we could break clients relying on the faulty behavior (e.g sending !!FOO) | |
| 14:35:09 | sean-k-mooney | right well at least im consitent i think we shoudl be fixing both without a microverion | |
| 14:35:15 | gibi | yeah :0 | |
| 14:35:16 | gibi | :) | |
| 14:35:51 | gibi | it is differnt in a way that now in 1.39 I can choose to a) fix the bug (as proposed) b) keep the faulty !!! accepted | |
| 14:36:06 | sean-k-mooney | i really dont like input validation bugs since they can lead to sql injection issues | |
| 14:37:04 | sean-k-mooney | yes ya you could fix it as part of the feature | |
| 14:37:34 | sean-k-mooney | and nova coudl just always start using 1.39 if its supported by placment | |
| 14:37:57 | sean-k-mooney | stephenfin: didnt ye make osc-placement use the latest microversion recently | |
| 14:38:04 | sean-k-mooney | like in xena? | |
| 14:38:54 | gibi | I think it is last microversion up until the gap in support | |
| 14:39:42 | sean-k-mooney | we have 1.38 support | |
| 14:39:47 | sean-k-mooney | for consumer types | |
| 14:40:14 | sean-k-mooney | so assuming you add 1.39 support for any traits we shoudl be ok form that perspective | |
| 14:42:15 | gibi | I don't find the patch but my env defaults to 1.29 not 1.38 | |
| 14:42:28 | sean-k-mooney | https://github.com/openstack/osc-placement/commits/master | |
| 14:42:35 | sean-k-mooney | i was just looking at the recent patches | |
| 14:43:11 | sean-k-mooney | ah yes | |
| 14:43:13 | sean-k-mooney | MAX_VERSION_NO_GAP = '1.29' | |
| 14:43:17 | sean-k-mooney | https://github.com/openstack/osc-placement/blob/281ace20df8bd01f4cb8ae29c39d80e754a9d0a3/osc_placement/version.py#L57 | |
| 14:43:43 | gibi | yepp, so that gets picked when nothing is provided in the CLI | |
| 14:44:16 | sean-k-mooney | yes ok so we would need to close the gap to have a new micorvesion auto micked | |
| 14:44:22 | sean-k-mooney | so 30-36 | |
| 14:44:23 | gibi | yes | |
| 14:45:03 | sean-k-mooney | ok well i would hope we could fix it with out a microverion but if we do need one then i guess that can improve the ux | |
| 14:47:19 | sean-k-mooney | 1.30 - ‘/reshaper’, 1.31 - ‘in_tree’ 1.32 - Support forbidden aggregates 1.33 - Support string request group suffixes 1.34 - Request group mappings in allocation candidates¶ 1.35 - Support ‘root_required’ and 1.36 - Support ‘same_subtree’ | |
| 14:47:56 | sean-k-mooney | so ingnoring 1.30 which we proably shoudl add as unused | |
| 14:48:05 | sean-k-mooney | the rest seam to be allocation canidate related | |
| 14:48:18 | sean-k-mooney | do we want osc-placemetn to be able to get allcoation candiates | |
| 14:48:58 | sean-k-mooney | i guess it alreday can https://github.com/openstack/osc-placement/blob/master/osc_placement/resources/allocation_candidate.py | |
| 14:50:10 | gibi | it can today | |
| 14:50:19 | gibi | even with multiple requrest groups | |
| 14:50:38 | sean-k-mooney | ya i guess that is somehwat useful for debuging | |
| 14:50:52 | sean-k-mooney | you can get teh query form the sechduler log and try and recreate it i guess | |
| 14:51:22 | sean-k-mooney | well i guess its useful if you are using placment for things other then openstack or in a standalone mode | |
| 14:51:34 | sean-k-mooney | like the fidge/sandwich demos | |
| 14:54:52 | opendevreview | Balazs Gibizer proposed openstack/osc-placement master: Support microversion 1.39 https://review.opendev.org/c/openstack/osc-placement/+/828545 | |
| 14:55:59 | bauzas | gibi: sorry, just starting my day after ~4h of meetings | |
| 14:56:09 | bauzas | (lovely) | |
| 14:56:14 | gibi | bauzas: that sounds hard | |
| 14:56:18 | gibi | why soo many meetings? | |
| 14:58:40 | bauzas | internal needs (research tax credits discussion, documentation, mentoring...) | |
| 15:03:31 | gibi | all the fun stuff :) | |