| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-04-02 | |||
| 17:07:22 | mriedem | sean-k-mooney: that's shade or oaktree or whatever mordred had before | |
| 17:07:25 | sean-k-mooney | but im not sure the bevhior of nova cli always using latest is actully good | |
| 17:07:50 | mriedem | "the nova cli is not a bash sdk" | |
| 17:08:07 | dansmith | you can run fooclient --help and bumble the way through a thing, but you can't do that with osc help because you don't have the magic version decoder ring | |
| 17:08:31 | mriedem | the osc help should tell you what version is needed to do things, | |
| 17:08:41 | sean-k-mooney | dansmith: ya its both better and worse that it show you the help for the microversion you specify | |
| 17:08:41 | mriedem | but that's been a gap in people reviewing and enforcing those osc changes | |
| 17:08:56 | sean-k-mooney | if you dont know the microverion exits your screwed | |
| 17:09:24 | dansmith | mriedem: okay I've literally never seen that on a help, but maybe because the things I needed didn't have it | |
| 17:09:37 | mriedem | e.g. https://docs.openstack.org/python-openstackclient/latest/cli/command-objects/server-event.html#server-event-show | |
| 17:09:49 | mriedem | part of the gaps closure in that etherpad is just adding stuff like that to the help | |
| 17:09:54 | dansmith | mriedem: oh that's in the docs not in the --help? | |
| 17:09:58 | mriedem | same | |
| 17:10:03 | mriedem | docs are generated from the help | |
| 17:10:07 | mriedem | like osc-placement | |
| 17:10:32 | dansmith | okay well, I've never seen it in the help for things I was looking at, but okay | |
| 17:10:52 | mriedem | another example for 2.73 that tssurya added https://docs.openstack.org/python-openstackclient/latest/cli/command-objects/server.html#server-lock | |
| 17:11:02 | mriedem | yeah it's a constant fight to make sure people remember to add those, | |
| 17:11:05 | mriedem | and get people to backfill them | |
| 17:11:08 | mriedem | hence the etherpad | |
| 17:11:33 | dansmith | in that one page, one says "use version X" and another says "use version X by passing --this-complicated-arg-thing" | |
| 17:11:49 | dansmith | but yeah, it would definitely be better if that was all consistent enough to rely on | |
| 17:12:09 | rambo_li | https://bugs.launchpad.net/neutron/+bug/1815989 And what is our plan to solve this bug? | |
| 17:12:09 | rambo_li | Excuse me, sean-k-mooney ralonsoh Please Let’s talk about the bug here: | |
| 17:12:09 | rambo_li | Excuse me, sean-k-mooney ralonsoh Please Let’s talk about the bug here: | |
| 17:12:11 | openstack | Launchpad bug 1815989 in OpenStack Compute (nova) "OVS drops RARP packets by QEMU upon live-migration causes up to 40s ping pause in Rocky" [Medium,In progress] - Assigned to sean mooney (sean-k-mooney) | |
| 17:12:13 | dansmith | I'm not helping so I know I can't complain too much, but in the current state of the world, I don't see osc as better | |
| 17:13:04 | mordred | mriedem: fwiw - I want to completely change what the version story is in osc - and I agree with sean-k-mooney - I do not think you should have to specify a min version on the cli | |
| 17:13:09 | sean-k-mooney | rambo_li: at the moment im not working on it, the patch i submited shoudl fix it but they dont work anymore due to rebases and refacoting | |
| 17:13:25 | mordred | it is that way for hysterical raisins and I think those raisins are no longer useful to us | |
| 17:14:04 | sean-k-mooney | mordred: if we default to oldest that support feature ex it keeps the stablity and if you want the new behavior you ask for it | |
| 17:14:07 | mordred | I am planning on writing something up so we can discuss it around PTG time to make sure it meets what people are wanting | |
| 17:14:26 | sean-k-mooney | but we dont do that today in many cases we say to use option x you need to pass version y | |
| 17:15:18 | sean-k-mooney | i have noticed that lately with osc-placment when trying ot use some operation but im sure we have the same behviaor in the nova supprot in osc | |
| 17:15:22 | mordred | right - I think there's actually a third option which can be more aggressive about new things while not breaking people - but I need to write it up | |
| 17:16:04 | mordred | from my end-user pov - if I *ever* have to give osc a version for a reason other than "I am a developer debugging the rest api" something has completely failed | |
| 17:16:23 | dansmith | mordred: ++ | |
| 17:16:26 | mordred | which is not to say that this failure isn't happening consistently today :) | |
| 17:16:37 | mordred | but - it's a failure from an UX perspective and I want to fix it | |
| 17:16:51 | mordred | s/want/am going to/ | |
| 17:17:12 | mordred | I will know I'm successful when everyone here thinks it's behaving properly ;) | |
| 17:17:14 | dansmith | "if you've ever had to pass a version to OSC in the last six months, you get an ATC pass to the summit" | |
| 17:17:22 | mordred | dansmith: ++ | |
| 17:18:09 | sean-k-mooney | mordred: you will have to get the ironic folks on board too | |
| 17:18:36 | sean-k-mooney | mordred: sicne they went ahead and impleemnted the same version negoction logic in there plugin they had in the legacy client | |
| 17:19:10 | mordred | sean-k-mooney: yeah - I've already got good ironic support done by the ironic team in sdk though - so I think I'm closer with them | |
| 17:19:34 | sean-k-mooney | mordred: i think one thing to condiser is moveing everything after nova v2.0 out of osc and into a nova plugin | |
| 17:19:44 | mordred | I want to get rid of all plugins | |
| 17:20:02 | sean-k-mooney | mordred: well part of the issue currently is geting things landed in osc | |
| 17:20:17 | mordred | right. that's based on the past - we're making some massive changes | |
| 17:20:25 | sean-k-mooney | so project wich plugisn could make progress but "core" project had an issue | |
| 17:20:26 | mordred | so - gimme a little bit to let those changes percolate | |
| 17:20:52 | sean-k-mooney | sure im glad the sdk and ocs team are combined/combinging | |
| 17:31:45 | sean-k-mooney | mordred: honestly if we could deprecate the nova clinet and deliver changes as easily as we can today to osc for V that would be awsome but there are still a long list of gaps to close before that is realistic | |
| 17:33:02 | artom | sean-k-mooney, I don't think anyone disagrees - but to paraphrase mriedem, and directly quote Nike, "just do it" :) | |
| 17:37:39 | mordred | sean-k-mooney: I agree ... but I'm going to make a concerted push to see how many of those gaps I can close and how much I can ease the situation | |
| 17:38:37 | mordred | might not work, obviously- but otoh - getting novaclient parity and being able to deprecate it would be a _huge_ step forward in terms of gravity, so I think it's worth extra attention | |
| 17:39:35 | openstackgerrit | Merged openstack/nova master: Add test coverage of existing server diagnostics policies https://review.opendev.org/716800 | |
| 17:40:19 | sean-k-mooney | well i almost never need to use nova client for day to day things but it does happen. | |
| 17:40:31 | sean-k-mooney | i should really just chagne https://review.opendev.org/#/c/643578/ back to evacuate too | |
| 17:44:12 | sean-k-mooney | mordred: should i just abandon ^ by the way or if i rebase it and change the command back to evacuate will it get reviews | |
| 17:44:58 | sean-k-mooney | i kindof gave up on this since i was hoping it would be a non contoversionl quick gap to close and it ended up being the opisite | |
| 17:51:45 | mordred | sean-k-mooney: I will totally review it | |
| 17:51:52 | mordred | actually. lemme read the scrollback on it real quick | |
| 17:53:12 | sean-k-mooney | mordred: context was at the fourm sesseion we express frustration that evacuate did not match with peopel expcations of keeping data at least not unless you are on shared storage | |
| 17:53:26 | sean-k-mooney | so it was suggested to change the name to resurect or recreate | |
| 17:53:46 | sean-k-mooney | so i did but there is also the consernt that using a different name | |
| 17:54:04 | sean-k-mooney | would confuse people that think its somehting new or different | |
| 17:54:38 | mordred | sean-k-mooney: hrm. well - I like the idea of the term ressurrect | |
| 17:54:47 | mordred | (for all the reasons you mention in the patch) | |
| 17:55:19 | sean-k-mooney | i kind of feel like we should really change how the api works | |
| 17:55:34 | sean-k-mooney | and then have two seperate commands | |
| 17:55:41 | mordred | sean-k-mooney: I kinda want it to be resurrect and then just have evacuate work as an alias for it - so that people with muscle memory will get what they're expecting, but new people will learn something new | |
| 17:55:45 | mordred | or maybe that | |
| 17:55:50 | sean-k-mooney | one that will always preserve your data and refuse to work if not on shared stroate | |
| 17:56:01 | sean-k-mooney | and one that always recretes the vm | |
| 17:56:09 | sean-k-mooney | basicaly just rebuild on a differnet host | |
| 17:56:17 | mordred | yeah - but with the same volumes and ports | |
| 17:56:41 | sean-k-mooney | ya so litrally rebuild on different host reimabing the root disk | |
| 17:56:55 | sean-k-mooney | *reimageing | |
| 17:58:01 | sean-k-mooney | see its tricky and once its in a cli we cant really remove it | |
| 18:30:27 | openstackgerrit | Merged openstack/nova master: Introduce scope_types in server diagnostics https://review.opendev.org/716803 | |
| 18:39:12 | openstackgerrit | Merged openstack/nova master: Add new default roles in server diagnostics policies https://review.opendev.org/716805 | |
| 18:39:19 | openstackgerrit | Merged openstack/nova master: func tests: move _run_periodics() into base class https://review.opendev.org/714228 | |
| 18:39:27 | openstackgerrit | Merged openstack/nova master: tests: work around malformed serial XML https://review.opendev.org/705546 | |
| 19:55:22 | _erlon_ | bauzas: ping | |
| 19:56:11 | _erlon_ | bauzas: so, you mentioned that there is a way to add a custom filter without having to change the egg files? | |
| 20:07:13 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Support for nova-manage placement heal_allocations --cell https://review.opendev.org/714459 | |
| 20:07:51 | mriedem | hello nova cores, ^ is pretty simple and would be nice to get in before that stuff is all refactored to nova-audit | |
| 20:36:02 | artom | _erlon_, since bauzas is in France is done for the day, maybe https://docs.openstack.org/nova/latest/user/filter-scheduler.html#writing-your-own-filter can help you? | |
| 20:39:46 | _erlon_ | artom: hey, thanks for helping, my question was more about the import error that I was getting the last time I tried to add a class that was not in the built in filters | |
| 20:42:08 | _erlon_ | artom: let's say I create a new file my_filter.py, with TheFilter, inside. If I put my_filter,py inside the nova/scheduler/filters, and change enabled_filters=+TheFilter, it use to give me an import error | |
| 20:42:50 | _erlon_ | artom: so, we needed to add an entry_point in the nova-egg | |
| 21:01:48 | mordred | _erlon_: don't put it in nova/scheduler/filters - check that link | |
| 21:07:10 | _erlon_ | available_filters = myfilter.MyFilter | |
| 21:07:10 | _erlon_ | available_filters = nova.scheduler.filters.all_filters | |
| 21:07:10 | _erlon_ | [filter_scheduler] | |
| 21:07:10 | _erlon_ | see that thereis 2 instances under the same stanza: | |
| 21:07:10 | _erlon_ | mordred: got it! As he puts in the config file, isnt that overwriting the default available filters? | |
| 21:09:31 | mriedem | no, it's multi-valued https://docs.openstack.org/nova/latest/configuration/config.html#filter_scheduler.available_filters | |