| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-08-10 | |||
| 13:32:41 | bauzas | so, we need to remove Ram *and* Diks | |
| 13:32:56 | bauzas | I was just remembering we still run 2 legacy filters | |
| 13:32:59 | mriedem | http://git.openstack.org/cgit/openstack-dev/devstack/tree/lib/nova#n104 | |
| 13:33:02 | mriedem | ok | |
| 13:33:16 | bauzas | anyway, patch is on its way | |
| 13:34:07 | mriedem | sdague: oh right, that docs bug was because cfriesen was asking about this, see the last paragraph or two in that bug report, | |
| 13:34:30 | mriedem | someone had forced the service down, upgraded everything else, and then when they tried to set forced_down=False, it failed with ServiceTooOld | |
| 13:34:50 | mriedem | which i think just caused some confusion in how that all works | |
| 13:35:01 | openstackgerrit | Sean Dague proposed openstack/nova master: Update api-guide and api-ref to be clear about forced-down https://review.openstack.org/492533 | |
| 13:35:07 | mriedem | the ServiceTooOld is i'm pretty sure a 500 out of the PUT /os-services API too | |
| 13:35:40 | sdague | mriedem: yeh, I don't know what the recovery path is on that all, but at least we can be clear on what the state transition into it should look like | |
| 13:35:46 | cdent | does anyone have anything specific for me to do, or shall I continue chasing reviews and randomly testing things by hand? | |
| 13:37:22 | edleafe | cdent: sudo make me a sandwich | |
| 13:38:50 | bauzas | sdague: mriedem: https://review.openstack.org/#/c/492537/ | |
| 13:39:04 | bauzas | ^ devstack change removing legacy filters FTW | |
| 13:39:21 | sdague | bauzas: ok, cool | |
| 13:39:28 | sdague | bauzas: is there a reason we override those in the first place? | |
| 13:39:39 | bauzas | sdague: correct me if I'm wrong but we don't need to modify grenade since it uses default config optoions ? | |
| 13:39:41 | sdague | would we just use defaults? | |
| 13:39:44 | mriedem | sdague: because devstack had the default + SameHost + DifferentHost | |
| 13:39:49 | bauzas | what mriedem said | |
| 13:39:57 | mriedem | tempest has tests for SameHost/DifferentHost filters | |
| 13:39:59 | sdague | ah, cool, good reason | |
| 13:40:00 | mriedem | which aren't defaults | |
| 13:40:02 | bauzas | we could tho make += "SameHost" | |
| 13:40:09 | bauzas | since it's a listopt | |
| 13:40:15 | mriedem | in bash? | |
| 13:40:23 | bauzas | ah, right | |
| 13:40:23 | sdague | bauzas: that doesn't work in setting nova.conf | |
| 13:40:35 | bauzas | I usually do this directly :) | |
| 13:41:01 | bauzas | sdague: anyway, my question still remains wrt grenade | |
| 13:41:13 | bauzas | sdague: do we need to s// something when we upgrade the node ? | |
| 13:41:23 | sdague | bauzas: grenade should be fine as long as those things didn't get deleted | |
| 13:41:25 | bauzas | in terms of nova.conf -ism | |
| 13:41:36 | sdague | it will use the ocata config in pike | |
| 13:41:46 | bauzas | yeah, my thoughts | |
| 13:41:48 | bauzas | not super then | |
| 13:42:13 | bauzas | we don't remove those filters because CachingScheduler still requires them | |
| 13:42:28 | bauzas | but in theory, you don't need to run those filters since Ocata | |
| 13:42:40 | sdague | right as long as there is normal deprecation cycles you don't need to do anything in grenade | |
| 13:42:51 | bauzas | that's just we wanted to be gentle in Ocata and not ask operators to change their conf when upgrading from Newton | |
| 13:42:59 | sdague | as long as you remove the deprecated setting of things in devstack in the release where you deprecate them | |
| 13:43:08 | bauzas | okay, I think it's reasonable then | |
| 13:43:10 | sdague | it just naturally rolls through it all | |
| 13:43:17 | bauzas | k | |
| 13:44:33 | bauzas | mriedem: looping over https://etherpad.openstack.org/p/nova-pike-release-candidate-todo | |
| 13:44:48 | sdague | mriedem: I think putting the entire recovery path inside the API ref for forced-down is not idea - https://review.openstack.org/#/c/492533/1/api-ref/source/os-services.inc | |
| 13:45:02 | bauzas | mriedem: AFAICT, the top prio for reviewing is then https://review.openstack.org/#/c/491012/ ? | |
| 13:45:06 | sdague | there probably needs to be a specific HA recovery document for the whole thing | |
| 13:45:15 | sdague | this is just "don't use this unless you really are sure" | |
| 13:45:34 | bauzas | :p | |
| 13:45:34 | bauzas | it's a force thing | |
| 13:45:51 | bauzas | 'force' is meaningful in Linux terminology :) | |
| 13:46:08 | mriedem | bauzas: yes | |
| 13:46:15 | bauzas | don't expect good things to happen magically if you use a force flag | |
| 13:46:33 | mriedem | sdague: i didn't say to put the entire recovery path in there... | |
| 13:46:41 | mriedem | just 'never come back' is not really correct, | |
| 13:46:50 | mriedem | i think maybe you mean 'never come back in it's current state' | |
| 13:46:51 | mriedem | or something | |
| 13:47:27 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Resource tracker compatibility with Ocata and Pike https://review.openstack.org/491012 | |
| 13:47:35 | mriedem | cdent: jaypipes: ^ fixed the functional test failure | |
| 13:47:38 | bauzas | you can "force enable" back | |
| 13:47:53 | cdent | mriedem: roger | |
| 13:47:59 | bauzas | but honestly, that forced-down thing is just a hack because of our SG API | |
| 13:48:15 | bauzas | I'm not sure we should be explicit here | |
| 13:51:15 | openstackgerrit | Ed Leafe proposed openstack/nova master: Handle addition of new nodes/instances in ironic flavor migration https://review.openstack.org/487954 | |
| 13:53:31 | openstackgerrit | Sean Dague proposed openstack/nova master: Update api-guide and api-ref to be clear about forced-down https://review.openstack.org/492533 | |
| 13:59:18 | mriedem | nova meeting in 1 minute | |
| 13:59:21 | bauzas | sdague: tried to help with defining why we have a force-down flag https://review.openstack.org/#/c/492533/2 | |
| 13:59:35 | bauzas | seeing that being used by operators scares me | |
| 13:59:42 | bauzas | it was never the intent | |
| 14:00:04 | bauzas | and like the bug report mentions, we have the disabling thing | |
| 14:00:10 | ioggstream | hi@all | |
| 14:00:14 | sdague | bauzas: right, so I think we have to be careful about putting too much into api-ref because it's supposed to be a reference | |
| 14:00:25 | sdague | we probably need an HA guide here that explains the whole thing | |
| 14:00:42 | ioggstream | about OS:: Nova::ServerGroup policy, how will anti-affinity (default) policy work if I have more vms than compute nodes? | |
| 14:00:52 | bauzas | sdague: well the problem is that our API can be consumed by both end-users and automation tools | |
| 14:01:13 | sdague | bauzas: I'll be honest, I'm not going to spend all day redrafting this docs patch. I think what I have up there is an improvement, if folks want to take it over and write it instead, I'm good with that | |
| 14:01:13 | mriedem | bauzas: the api-ref could link to a more detailed guide | |
| 14:01:14 | bauzas | sdague: to be frank, I do wonder if we should just cut that flag from the nova CLI | |
| 14:01:20 | mriedem | sdague: there is, or was, a maintenance guide | |
| 14:01:23 | bauzas | sdague: okay, I'll push a rev then | |
| 14:01:25 | mriedem | planned and unplanned | |
| 14:01:29 | sdague | mriedem: sure | |
| 14:01:40 | mriedem | i'd put it in there, but it might have been in the now defunct ops guide | |
| 14:01:46 | bauzas | mriedem: sdague: what are you thinking of just removing that method from our nova CLI ? | |
| 14:01:46 | mriedem | would have to dig | |
| 14:01:56 | bauzas | since only scripts should call it | |
| 14:02:00 | mriedem | meeting time | |
| 14:02:13 | sdague | bauzas: people might have manually downed their nodes as well | |
| 14:02:21 | sdague | bauzas: I think it's fine to be in the nova cli | |
| 14:02:30 | sdague | that's an admin tool | |
| 14:02:38 | sdague | they just need to realize what they are doing | |
| 14:02:43 | bauzas | agreed | |
| 14:02:58 | bauzas | sdague: I'll try a new rev | |
| 14:03:54 | sdague | bauzas: I think your assumption that no one should ever call this manually is wrong | |
| 14:03:58 | jianghuaw | jianghua | |
| 14:04:02 | sdague | the important point is they met preconditions | |
| 14:04:06 | sdague | the service is fenced | |
| 14:04:12 | sdague | we don't care how they met those | |