Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-04-29
16:03:01 mriedem i didn't take that literally
16:04:53 openstackgerrit Matt Riedemann proposed openstack/nova master: Add nova-multi-cell job https://review.opendev.org/655222
16:04:54 openstackgerrit Matt Riedemann proposed openstack/nova master: Enable n-novnc in nova-multi-cell job https://review.opendev.org/655711
16:14:52 larsks Dumb nova question: is specifying "--live ''" to the 'openstack server migrate' command the same as "setting the host parameter to null" in the corresponding api request? The docs suggest using the "nova live-migration" command in this case, rather than "openstack server migrate", but that latter *seems* to work and I want to make sure that's not an accident...
16:15:38 sean-k-mooney that is a good question
16:15:55 sean-k-mooney we have unfortunetly overloaded --live to do two things
16:16:11 sean-k-mooney enable live migratin and specify the host
16:16:20 sean-k-mooney really we should have a --host option for that
16:16:57 sean-k-mooney i would have to check the osc code too see if we have special cased ''
16:16:59 larsks That sounds like a really good idea.
16:17:06 larsks (having --host separate)
16:18:06 larsks I was wondering more if an empty string would effectively evaluate the same as null on the server side.
16:19:03 sean-k-mooney initally at least it looks like we pass the '' through directly https://github.com/openstack/python-openstackclient/blob/master/openstackclient/compute/v2/server.py#L1468
16:20:56 sean-k-mooney the nova client also intiall does not special case '' https://github.com/openstack/python-novaclient/blob/master/novaclient/v2/servers.py#L436
16:23:18 larsks Okay. I'm not familiar enough with the code to see where that ends up on the server side. I'm hoping that we'll find that 'host' is being evaluated as a boolean to determine whether its set or not, in which case it would Just Work.
16:24:11 larsks Oh, I think I found it.
16:25:00 larsks https://github.com/openstack/nova/blob/master/nova/compute/api.py#L4502
16:25:08 larsks So I think that works out.
16:25:50 mtreinish mriedem: oh, that's confusing. I wouldn't have read that correctly either
16:26:07 mriedem i thought it was an example
16:26:20 sean-k-mooney right so the api say the seculer will select the host if the value is none https://developer.openstack.org/api-ref/compute/?expanded=live-migrate-server-os-migratelive-action-detail
16:26:47 mriedem larsks: the osc live migrate command *requires* a host which is bad if you're not using a newer microversion to run that host through the scheduler,
16:26:50 sean-k-mooney but looking at that coe an empty strig would also be ignored by that if
16:26:50 mriedem because it forces the host
16:27:16 larsks Right.
16:27:28 mriedem larsks: see ~L16 https://etherpad.openstack.org/p/DEN-osc-compute-api-gaps
16:27:34 mriedem going to talk about fixing that this week
16:27:37 mriedem there have been several attempts
16:28:01 mriedem because since queens you can also specify a target host for cold migrate but not yet with osc
16:28:28 sean-k-mooney i assume we are not allowed to change existing args in osc
16:28:55 sean-k-mooney but we caould add a --live-migrate flag and --host <host> args
16:28:55 larsks mriedem: ...and just to make sure I'm not missing something, there continues to be no "live migrate *all* instances off the specified host" command at the api level, so e.g. setting up for system maintenance still requires individually migrating each instance, right?
16:29:18 sean-k-mooney larsks: correct
16:29:21 larsks Thanks.
16:29:24 mriedem larsks: not in the api or osc
16:29:28 sean-k-mooney and there likely should not be one at teh api
16:29:33 mriedem there is a nova cli command that does that
16:29:44 sean-k-mooney larsks: the nova host-evecuate-live command is impleeted in the nova client
16:29:50 mriedem larsks: http://www.danplanet.com/blog/2016/03/03/evacuate-in-nova-one-command-to-confuse-us-all/
16:30:16 larsks Is there an osc command? Because the nova cli doesn't support clouds.yaml, which is a pain.
16:30:38 larsks Meh, I can work around that. In any case, thank you for the pointer.
16:30:38 mriedem no
16:30:44 sean-k-mooney larsks: by the way the reason i said its not a good idea to have this at the api level is starting multple live migration on the same host concurrently increase the risk that they will fail or not complete
16:31:00 mriedem if you do live migrate with osc, i'd specify --os-compute-api-version 2.30 to avoid that force behavior https://docs.openstack.org/nova/latest/reference/api-microversion-history.html#id27
16:31:16 mriedem sean-k-mooney: by default the compute throws those into a blocking queue
16:31:19 sean-k-mooney larsks: the best way to work around it is to serialise the migrations
16:31:19 mriedem one at a time
16:31:37 sean-k-mooney really i was not aware of that
16:31:44 sean-k-mooney is that contoled via a config?
16:36:04 mriedem sean-k-mooney: yes
16:36:26 mriedem sean-k-mooney: https://docs.openstack.org/nova/latest/configuration/config.html#DEFAULT.max_concurrent_live_migrations
16:36:46 mriedem dansmith added that i believe because of these evacuate all clis
16:36:52 mriedem which was about the time he wrote that blog post
16:37:54 dansmith yes, on advice from libvirt/qemu people that they basically should be serialized anyway
16:38:01 dansmith like, that default should really be =1
16:38:04 sean-k-mooney ah ok im not sure if that is the default we ship in OSP. but ya from talking to danpb about this in the past he strongly recommended serailising it
16:38:16 dansmith but we made it =10 to kinda keep existing behavior, but in reality it should be 1
16:38:20 dansmith yeah
16:38:39 sean-k-mooney ah ok so it is an osp config change
16:38:43 sean-k-mooney cool
16:40:07 mriedem the default is 1
16:40:15 sean-k-mooney yes upstream
16:40:23 sean-k-mooney which is what it should be donwstream but its not
16:40:44 sean-k-mooney that was what was confusing me as this came up internally about a month ago
16:50:07 openstackgerrit Matt Riedemann proposed openstack/nova master: Add functional test for AggregateMultiTenancyIsolation + migrate https://review.opendev.org/571265
16:55:08 artom For those of us not (yet) in Denver, is there a way to recheck a single job in the gate?
16:55:54 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_info_cache_update_at_top' https://review.opendev.org/651299
16:55:54 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_fault_create_at_top' https://review.opendev.org/651298
16:55:55 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_update_at_top', 'instance_destroy_at_top' https://review.opendev.org/651301
16:55:55 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'get_keypair_at_top' https://review.opendev.org/651300
16:55:56 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling 'update_cells' on 'BandwidthUsage.create' https://review.opendev.org/651303
16:55:56 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_update_from_api' https://review.opendev.org/651302
16:55:57 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove cells code https://review.opendev.org/651306
16:55:57 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling cells v1 for instance naming https://review.opendev.org/651304
16:55:58 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove unnecessary wrapper https://review.opendev.org/651308
16:55:58 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling 'InstanceUnknownCell' exception https://review.opendev.org/651307
16:55:59 openstackgerrit Stephen Finucane proposed openstack/nova master: conf: Remove cells v1 options, group https://review.opendev.org/651310
16:55:59 openstackgerrit Stephen Finucane proposed openstack/nova master: db: Remove cell APIs https://review.opendev.org/651309
17:15:34 aspiers[m] artom: I couldn't find one
17:16:07 artom aspiers[m], thanks for looking!
17:16:12 artom I think I have other problems, anyways
17:16:25 artom Downstream CI Neutron and upstream CI Neutron are apparently doing different things :/
19:01:22 mriedem dansmith: you might want to take a look at this bug fix https://review.opendev.org/#/c/654584/
19:05:06 openstackgerrit Matt Riedemann proposed openstack/nova stable/stein: AZ list performance optimization: avoid double service list DB fetch https://review.opendev.org/656382
19:08:15 mriedem need a stable core to hit this https://review.opendev.org/#/c/656176/
19:58:54 openstackgerrit Merged openstack/nova master: Log notifications if assertion in _test_live_migration_force_complete fails https://review.opendev.org/650155
19:59:03 openstackgerrit Merged openstack/nova master: Add functional regression recreate test for bug 1825020 https://review.opendev.org/653268
19:59:04 openstack bug 1825020 in OpenStack Compute (nova) "resize of instance, that was booted from volume, trigger image size check then fails resize" [High,In progress] https://launchpad.net/bugs/1825020 - Assigned to Matt Riedemann (mriedem)
20:02:35 HW_Peter hey quick question
20:02:56 HW_Peter are network events
20:03:02 HW_Peter such as vif plugging and unplugging
20:03:09 HW_Peter supported in a multi cellsv2 environmnet
20:24:39 artom HW_Peter, Summit/PTG are happening, so you're unlikely to get too much attention here (sorry!), but yeah, since events are necessary for basic things like booting instances (with libvirt at least), I imagine we made sure they still worked in multicell
21:03:58 mriedem mtreinish: failed again so i guess the tempest docs are busted or i'm busted, but i'm going to drop the new lines and comments
21:04:38 mriedem gmann: fyi i pulled the tempest_test_regex and tempest_black_regex examples from the tempest readme and they don't seem to be working https://review.opendev.org/#/c/655222/7/.zuul.yaml
21:07:44 mtreinish mriedem: yeah I'm not sure if the comments will work (the new lines should I think, not 100% sure). I think with '|' the yaml parser passes comments as part of the string which probably is breaking the regex
21:08:30 mriedem i'm just going to use tempest_test_blacklist
21:08:33 mtreinish mriedem: this type of use case is why stestr has the whitelist and blacklist files so you don't need to try and use a giant complicated regex as a cli arg
21:09:39 openstackgerrit Matt Riedemann proposed openstack/nova master: Enable n-novnc in nova-multi-cell job https://review.opendev.org/655711
21:09:39 openstackgerrit Matt Riedemann proposed openstack/nova master: Add nova-multi-cell job https://review.opendev.org/655222

Earlier   Later