| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-04-17 | |||
| 10:19:55 | hrw | I just finished second | |
| 10:26:33 | lyarwood | just starting mine, tgif++ | |
| 10:26:49 | lyarwood | https://review.opendev.org/#/c/669674/ - could use a non-RH core on this given it has an impact across virt drivers. | |
| 11:52:37 | averi | hi, anyone has an idea on why code changes are not picked up by nova_compute containers (OSP 13)? | |
| 11:53:06 | averi | even when spinning up a new image based on the nova-compute one, it doesn't work, code changes are there but they're like ignored all together | |
| 11:53:32 | averi | even killing one of the library files results in the binary not to fail like the file was still around | |
| 12:47:55 | gibi | dansmith, sean-k-mooney, brinzhang: I need a second opinion in https://review.opendev.org/#/c/720670 | |
| 12:51:26 | sean-k-mooney | gibi: well my view is we should have been storing the acell profile in the request spec or in the instnace like we cache the flavor so we dont have this proble but ill go read what the patch acatully does now | |
| 12:52:27 | gibi | sean-k-mooney: it is like create an instance with accel, then remove cyborg service from deployment, then try to delete the instance which will fail | |
| 12:52:47 | sean-k-mooney | gibi: im kind of felling like we should do what ever we would do if cinder or neutron went away | |
| 12:53:01 | sean-k-mooney | i would assume we would fail to delete but i confess i have never tried it | |
| 12:53:20 | gibi | I can try that | |
| 12:54:27 | sean-k-mooney | cool you should jsut need to delete the entry in the service in keystone to emulate that right | |
| 12:55:19 | gibi | I'm not sure that a) this is a valid scenarios that cyborg service is removed from the deployment b) if yes then which is more important in general keeping cyborg - nova in sync or allowing to delete an instance during a temporary failure of cyborg | |
| 12:56:15 | sean-k-mooney | gibi: i am leaning towords this si operator error for removing a service before its consumer are removed | |
| 12:56:33 | sean-k-mooney | that said i dont think they should have to do db surgury to remove the instance | |
| 13:01:59 | gibi | sean-k-mooney: without neutron endpoint in keystone the instance delete fails with HTTP 500 | |
| 13:02:02 | gibi | http://paste.openstack.org/show/792326/ | |
| 13:02:18 | sean-k-mooney | ack | |
| 13:02:27 | sean-k-mooney | ya that is baskcially what i expected | |
| 13:02:43 | sean-k-mooney | and i would expect the same for cyborg | |
| 13:02:45 | gibi | so this also points towards that the scenario of removing cyborg is invalid | |
| 13:03:08 | sean-k-mooney | yes i was going to comment with a proceedure to do that in the bug | |
| 13:03:15 | sean-k-mooney | but just removing it i think is invalid | |
| 13:03:20 | sean-k-mooney | *patch | |
| 13:05:07 | openstackgerrit | Lee Yarwood proposed openstack/nova master: DNM - Test stable device rescue tests with BFV instances https://review.opendev.org/710050 | |
| 13:06:17 | sean-k-mooney | gibi: left a comment with my toughts | |
| 13:06:21 | gibi | sean-k-mooney: thanks | |
| 13:38:25 | gibi | sean-k-mooney, dansmith thanks for the reviews on the https://review.opendev.org/#/c/720670 I marked the bug invalid | |
| 13:38:50 | dansmith | cool | |
| 13:48:21 | sean-k-mooney | dansmith: gibi cross cell migrate was complete this cycle right it was not supported in train | |
| 13:48:42 | sean-k-mooney | well cross cell resize. same thing | |
| 13:48:46 | averi | dansmith, sean-k-mooney: sorry for the direct poke, got any pointer on the question above? | |
| 13:49:15 | sean-k-mooney | i did not see the quest can you repeat it. i can check scollback | |
| 13:49:20 | averi | ah sure, sorry | |
| 13:49:26 | averi | <averi> even killing one of the library files results in the binary not to fail like the file was still around | |
| 13:49:26 | averi | <averi> even when spinning up a new image based on the nova-compute one, it doesn't work, code changes are there but they're like ignored all together | |
| 13:49:26 | averi | <averi> hi, anyone has an idea on why code changes are not picked up by nova_compute containers (OSP 13)? | |
| 13:49:51 | averi | I cannot troubleshooting anything as soon as oslo service comes into play | |
| 13:50:35 | sean-k-mooney | how are you modifying the code and restarting thet container | |
| 13:50:40 | sean-k-mooney | are you using docker directly | |
| 13:51:17 | sean-k-mooney | if you exec into the contienr with doeck an modify the code then do a docker restart for the container it will pick up the changes | |
| 13:51:37 | dansmith | sean-k-mooney: I haven't tried it, but yeah it was U not T | |
| 13:51:51 | averi | I spin up a new container based on the upstream image, then exec within it, modify the library (interested in pymysql debugging), then shut down the container and start it again | |
| 13:52:02 | sean-k-mooney | if you are using paunch it pull a clean copy of the image every time you get it to restat the container so your changes will be lost | |
| 13:52:11 | averi | sean-k-mooney, the changes are there in the code, they persist, but they are completely ignored | |
| 13:52:16 | sean-k-mooney | dansmith: ok just traiging a donwstream bug | |
| 13:52:24 | dansmith | averi: this is probably not the best channel for that question | |
| 13:52:59 | averi | dansmith, sorry, a contact internally mentioned this was a possible good channel for that type of question | |
| 13:53:08 | sean-k-mooney | averi: then i would guess its not using the files you are modifying and the kolla or ooo folks would be betere able to help | |
| 13:53:26 | averi | what would be the correct channel? | |
| 13:53:50 | sean-k-mooney | averi: #openstack-kolla or better #openstack-tripleo | |
| 13:53:56 | averi | thanks | |
| 13:54:12 | averi | mm.. 4 folks in #openstack-tripleo | |
| 13:54:28 | dansmith | averi: are you a redhatter? | |
| 13:54:32 | averi | yes | |
| 13:55:07 | sean-k-mooney | maybe that is not there normal channel | |
| 13:55:23 | dansmith | averi: this is the upstream nova channel.. people here don't know anything about OSP :) I'll pm | |
| 13:56:31 | sean-k-mooney | averi: fyi is just #tripleo | |
| 13:56:36 | averi | ack | |
| 14:00:04 | gibi | sean-k-mooney: cross cell resize merged in Ussuri not supported in Train | |
| 14:01:01 | sean-k-mooney | gibi: yep i just have a resize bug specificly in a multi cell enve and i wanted to triple check that cross cell resize is not supported | |
| 14:02:45 | bauzas | gibi: sorry, I was not around since this morning | |
| 14:02:54 | bauzas | gibi: I haven't yet looked at the bugs | |
| 14:03:02 | bauzas | thanks for working on it | |
| 14:03:26 | openstack | Launchpad bug 1841932 in OpenStack Compute (nova) "hide_hypervisor_id extra_specs in nova flavor cannot pass AggregateInstanceExtraSpecsFilter" [Undecided,New] | |
| 14:03:26 | rambo_li | sean-k-mooney: maybe their is an error, for the bug :https://bugs.launchpad.net/nova/+bug/1841932. when we set the metadata "hide_hypervisor_id=true"in aggreate,so we will go to the https://github.com/openstack/nova/blob/46a3bcd80b41e99ec4923c7cf3d0f8dd8505e97c/nova/scheduler/filters/aggregate_instance_extra_specs.py#L69-L71, but this should not break, because the aggregate_val doesn't match the req. | |
| 14:03:27 | gibi | bauzas: no worries there was one potential regression around cyborg but we just ruled it out as unsupported scenario | |
| 14:03:56 | gibi | bauzas: and honestly I did not triaged any bug, I'm just reviewing https://review.opendev.org/#/c/719163 | |
| 14:04:49 | rambo_li | the aggregate_vals is "true" | |
| 14:04:50 | gmann | melwitt: stephenfin can i get your review on this which cover remaining policy of server (not reviewed previously ) - https://review.opendev.org/#/c/720104/ and its scope and defaults change patch. | |
| 14:08:27 | melwitt | gmann: yeah can do. I didn't review it yet because earlier patch in the stack I wanted to wait for johnthetubaguy since he spotted issues before. but you are right, can still look at the others | |
| 14:08:53 | gmann | melwitt: yeah, thanks. | |
| 14:09:17 | bauzas | gibi: ack, will do as well later | |
| 14:09:18 | melwitt | gibi: hi, I have small follow ups here for the host_status:unknown-only policy rule here if you could please look https://review.opendev.org/713295 | |
| 14:09:41 | gibi | melwitt: sure. looking | |
| 14:09:46 | melwitt | thanks | |
| 14:11:02 | openstackgerrit | Andreas Jaeger proposed openstack/nova master: Remove Babel requirement https://review.opendev.org/720725 | |
| 14:23:07 | rambo_li | sean-k-mooney: if we set the metadata "hide_hypervisor_id = true "in aggregate, so the code https://github.com/openstack/nova/blob/master/nova/scheduler/filters/aggregate_instance_extra_specs.py#L62 , the aggregate_vals in code is "true". so last will also return False.so workaround this is to set hw:hide_hypervisor_id replace hide_hypervisor_id in flavor extra_spec and aggregate metadata. | |
| 14:39:13 | gibi | melwitt: +A on the followup | |
| 14:39:23 | melwitt | danke | |
| 14:51:10 | melwitt | gmann, bauzas, gibi, dansmith: does anyone know about babel to help review this? https://review.opendev.org/720725 it seems ok based on the ML post but I don't understand it well enough whether there could be any issue with removing it as a requirement | |
| 14:51:39 | melwitt | stephenfin too | |
| 14:51:52 | bauzas | weirdo | |
| 14:52:10 | gibi | melwitt: I have to pass on that now. I hope somebody else can take it | |
| 14:52:32 | bauzas | we don't use babel for translations | |
| 14:52:33 | melwitt | it's not urgent, just wanted to mention it while people are around | |
| 14:52:38 | bauzas | but I can git blame | |
| 14:53:11 | bauzas | melwitt: wow, this was old | |
| 14:53:32 | bauzas | melwitt: https://review.opendev.org/#/c/3870/ | |
| 14:53:52 | bauzas | but the translation team no longer uses it AFAIK | |
| 14:53:56 | bauzas | (for a while) | |
| 14:54:26 | bauzas | https://docs.openstack.org/i18n/latest/infra.html | |
| 14:54:37 | bauzas | this is now zanata | |
| 14:54:43 | bauzas | so I think we can drop it | |
| 14:54:52 | bauzas | but I leave the decision to other folks :p | |
| 14:55:11 | melwitt | yeah. hm, so maybe we would want to remove all references. ajaeger had said babel is used to build translations so I was not sure what's going on | |
| 14:55:12 | stephenfin | melwitt: That _seems_ correct. You need Babel to run the e.g. 'python setup.py extract_messages' distutils command but I think _that_ would require it already be present, like setuptools | |
| 14:55:16 | bauzas | I'm just afraid of breaking things so close to the HardStringFreeze so I'd defer this to later | |
| 14:55:29 | melwitt | ah, nevermind | |