| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-07-10 | |||
| 18:13:00 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: Deprecate CONF.workarounds.enable_numa_live_migration https://review.opendev.org/640021 | |
| 18:13:00 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: NUMA live migration support https://review.opendev.org/634606 | |
| 18:26:11 | sean-k-mooney | someday we will have v4 api after the go rewite and then we can drop 2.1 api support | |
| 18:46:26 | mriedem | efried: i'm guessing you want to backport this right? https://review.opendev.org/#/c/637225/ | |
| 18:46:28 | mriedem | to at least rocky | |
| 18:48:51 | openstack | Launchpad bug 1835627 in devstack-plugin-ceph "test_shelve_unshelve_server failing on stable/pike (and probably ocata)" [Undecided,Confirmed] | |
| 18:48:51 | mriedem | lyarwood: fyi in case you never saw this https://bugs.launchpad.net/devstack-plugin-ceph/+bug/1835627 | |
| 19:26:18 | dansmith | efried: normally I'd like to see the top of a series be nearly ready to go before we merge a db migration for it: https://review.opendev.org/#/c/662697/ | |
| 19:26:26 | dansmith | I know you didn't +W, but, just my opinion | |
| 19:29:23 | efried | dansmith: I saw that go by and thought of you. | |
| 19:29:37 | efried | I considered -2ing for that reason, but then... didn't. | |
| 19:30:00 | efried | mriedem: yes, I would like to backport https://review.opendev.org/#/c/637225/ | |
| 19:30:19 | dansmith | I would have liked a different column name at least, but alas I missed my slot | |
| 19:30:44 | efried | sean-k-mooney: ack. What would be super awesome is if those efforts were sunk into getting us closer to NUMA modeling in placement. The nested magic stuff has merged. | |
| 19:31:16 | efried | dansmith: is it too late to correct it? We don't actually need to support mid-stream commit levels, do we? | |
| 19:31:41 | dansmith | uh, we do, and it's too late really | |
| 19:32:01 | dansmith | we could migrate the migration, but it's not worth it for that | |
| 19:32:27 | dansmith | but yes, we've always supported deploying any commit.. the sins that come from dropping that expectation are really bad | |
| 19:32:42 | efried | okay | |
| 19:42:31 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/stein: Fix GET /servers/detail host_status performance regression https://review.opendev.org/669958 | |
| 19:43:40 | efried | dansmith: I appreciate you looking at that vpmem series though. I look at the object stuff and all I can do is compare it to existing things - I don't really understand it. | |
| 19:44:27 | efried | btw, the conf thing isn't there because I asked them to put that last, as it's the "master switch" that makes the admin think something should work. Probably some additional refactoring would be appropriate so we're not referencing the conf option before we introduce it. | |
| 19:44:59 | mriedem | is that the thing that starts exposing the inventory? | |
| 19:45:41 | mriedem | hopefully they have pretty robust functional testing before that - which you should be able to do with stubs or something w/o the config option | |
| 19:45:50 | dansmith | efried: putting the config option after code that uses it is backwards | |
| 19:46:32 | mriedem | we've done something like that before to keep a series moving before exposing it in the api, | |
| 19:46:46 | dansmith | but not actually referencing a non-existent config option | |
| 19:46:51 | mriedem | gibi's bw stuff in stein did something like that | |
| 19:46:52 | mriedem | yeah | |
| 19:47:17 | mriedem | my cross-cell resize stuff does something like that with the new policy rule - stubs a compute api method way earlier in the series to make functional testing possible before the api actually supports cross-cell resize | |
| 19:47:39 | dansmith | wrap the check in a is_thing_enabled() method and patch that in your tests or something | |
| 19:47:49 | mriedem | yup that's what both gibi and i did | |
| 19:47:52 | dansmith | yeah | |
| 19:48:19 | dansmith | i this case, | |
| 19:48:21 | dansmith | however, | |
| 19:48:27 | mriedem | https://review.opendev.org/#/c/636253/36/nova/tests/functional/test_cross_cell_migrate.py@65 | |
| 19:48:31 | dansmith | the config thing is a complex encoded structure in a string list, | |
| 19:48:38 | dansmith | which doesn't lend itself to that pattern quite as well | |
| 19:49:05 | mriedem | mmm delicious pci whitelist offspring | |
| 19:49:09 | dansmith | maybe have the helper return the contents, and then later have it get the value from config and parse it or something | |
| 19:49:40 | dansmith | but putting the parsing in a part of the code that complains about the format not being adhered to, well in advance of the patch that introduces the config option and doc is kinda silly | |
| 19:49:40 | mriedem | that would lend itself to easier functional testing as well - just stub the helper to return whatever config you are testing | |
| 19:49:45 | dansmith | right | |
| 19:50:13 | efried | yup, agreed | |
| 19:50:54 | dansmith | to be clear, that thing should not just return the string(s), but do the god-awful parsing of the : and | tokens and return a structure | |
| 19:51:01 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add Instance.hidden field https://review.opendev.org/631123 | |
| 19:51:01 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add InstanceAction/Event create() method https://review.opendev.org/614036 | |
| 19:51:02 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add CrossCellMigrationTask https://review.opendev.org/631581 | |
| 19:51:02 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add TargetDBSetupTask https://review.opendev.org/627892 | |
| 19:51:03 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Execute TargetDBSetupTask https://review.opendev.org/633853 | |
| 19:51:04 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add PrepResizeAtDestTask https://review.opendev.org/627890 | |
| 19:51:04 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add prep_snapshot_based_resize_at_dest compute method https://review.opendev.org/633293 | |
| 19:51:05 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add PrepResizeAtSourceTask https://review.opendev.org/627891 | |
| 19:51:05 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add nova.compute.utils.delete_image https://review.opendev.org/637605 | |
| 19:51:05 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add prep_snapshot_based_resize_at_source compute method https://review.opendev.org/634832 | |
| 19:51:06 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Refactor ComputeManager.remove_volume_connection https://review.opendev.org/642183 | |
| 19:51:46 | dansmith | and I'd like to see that whole stack ready to go at merge time, | |
| 19:51:54 | dansmith | not merging dead code in the driver well in advance of the later patch | |
| 19:52:19 | mriedem | dansmith: i lost your +2 on https://review.opendev.org/#/c/631123/because of the db migration | |
| 19:52:22 | dansmith | I didn't walk up the set any further because it has a bunch of merged or abandoned patches in the middle | |
| 19:52:23 | dansmith | so probably needs a good restacking | |
| 19:53:11 | dansmith | mriedem: maybe you'll be more careful with this one | |
| 19:53:26 | mriedem | ? | |
| 19:53:37 | dansmith | you lost the last one I gave you | |
| 19:54:19 | mriedem | i'm pretty sure this is the last time i have to rebase this series | |
| 19:56:57 | efried | Do we cover revert_resize in any of our CI jobs? | |
| 19:57:20 | mriedem | yes | |
| 19:57:58 | mriedem | definitely single node, the trickier is multi-node coverage | |
| 19:58:13 | efried | I'm looking at https://review.opendev.org/#/c/665138/ and not finding out a way to tell whether it's doing... anything at all. | |
| 19:58:54 | mriedem | http://logs.openstack.org/39/669539/1/check/tempest-full-py3/1555a09/job-output.txt.gz#_2019-07-05_22_24_52_926869 | |
| 19:58:56 | artom | efried, you stay away from that stuff. Run, I tell you. RUN! | |
| 19:59:20 | artom | efried, you could try using https://review.opendev.org/#/c/664442/ somehow | |
| 19:59:34 | sean-k-mooney | efried: for what its worth i have poposed that we dont backfil the libvirt pos feature and instead use any bandwith it frees up to work on things like numa in placment or novas use of placement in general | |
| 19:59:51 | efried | sean-k-mooney: proposed internally at RH you mean? | |
| 19:59:58 | sean-k-mooney | ya | |
| 20:00:01 | efried | that's wonderful. | |
| 20:00:27 | efried | I would really like to see us focus on moving forward that way rather than adding to the burden of stuff that helps us stay where we are. | |
| 20:00:38 | sean-k-mooney | we will see if that helps or not but my intent at least is to finish off the work i have started and then see what time is left in the cycle | |
| 20:00:53 | mriedem | efried: the tempest-slow job has a cold migration test that is doing revert resize but it's skipped http://logs.openstack.org/39/669539/1/check/tempest-slow-py3/cd3cb0e/job-output.txt.gz#_2019-07-05_22_44_58_146668 until sean-k-mooney/artom's fix lands... | |
| 20:00:57 | mriedem | so a catch-22 | |
| 20:01:25 | mriedem | https://review.opendev.org/#/c/663405/ | |
| 20:02:09 | sean-k-mooney | mriedem: dansmith merged your change to add teh iptables job to the experimenatl pipline right so that can be used for testing revert for now | |
| 20:02:17 | mriedem | but yeah https://review.opendev.org/#/c/664442/ is the canary | |
| 20:02:24 | mriedem | sean-k-mooney: oh yeah that's right | |
| 20:02:39 | artom | sean-k-mooney, I thought that was single node? | |
| 20:02:41 | mriedem | see? there are only like 5 need-to-know manual on-demand things | |
| 20:03:03 | artom | (That being the reason it failed - we didn't test same host revert with OVS hybrid plug) | |
| 20:03:19 | mriedem | artom: it is | |
| 20:03:39 | mriedem | so neutron-tempest-iptables_hybrid tests ovs hybrid plug same-host resize, | |
| 20:03:52 | mriedem | tempest-slow (will eventually) test multi-node revert | |
| 20:03:56 | mriedem | but not ovs hybrid plug | |
| 20:04:04 | artom | We have same-host resize in tempest-full-py3 too, no? | |
| 20:04:07 | mriedem | and tempest-full tests non-hybrid plug same host resize revert | |
| 20:04:11 | efried | So I could update https://review.opendev.org/#/c/665138/ to depend-on https://review.opendev.org/#/c/664442/ and that would tell me... what? | |
| 20:04:12 | mriedem | yes, but not ovs hybrid plug | |
| 20:04:12 | artom | jynx | |
| 20:04:14 | mriedem | and single node | |
| 20:04:16 | sean-k-mooney | artom: the iptables one is singel node yes | |
| 20:04:33 | artom | My brain is leaking | |
| 20:04:37 | efried | I'm probably too dense today to understand it anyway; maybe I'll just push it to the back of the queue again until I'm feeling sharp. | |