| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-09-28 | |||
| 13:48:47 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Add hide server address tests in test_serversV21.py https://review.openstack.org/605435 | |
| 13:53:29 | mdbooth | dansmith: lazy load of a field from a deleted object? https://review.openstack.org/#/c/604400/5 | |
| 13:54:11 | mriedem | mdbooth: that's been fixed | |
| 13:54:23 | mriedem | well, something related to that has been fixed | |
| 13:54:56 | mriedem | https://github.com/openstack/nova/blob/master/nova/objects/instance.py#L900 | |
| 13:55:13 | mriedem | https://github.com/openstack/nova/commit/6ba8a35825a7ec839b2d0aab7559351d573130ab | |
| 13:55:19 | mriedem | but that's generic attributes, which flavor is not | |
| 13:55:41 | mriedem | https://github.com/openstack/nova/blob/master/nova/objects/instance.py#L958 | |
| 13:56:51 | mriedem | ah yup http://logs.openstack.org/00/604400/5/check/nova-live-migration/6aa7a4b/logs/subnode-2/screen-n-cpu.txt.gz#_Sep_26_23_53_06_475005 | |
| 13:57:32 | mriedem | mdbooth: good call | |
| 13:57:35 | mriedem | do we have a bug for that? | |
| 13:58:23 | mdbooth | mriedem: Haven't checked. I wrote that comment up in a panic as I'm about to be late for a meeting ;) | |
| 14:00:45 | mriedem | i'll look | |
| 14:00:54 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/rocky: Null out instance.availability_zone on shelve offload https://review.openstack.org/606086 | |
| 14:03:56 | melwitt | . | |
| 14:06:16 | openstackgerrit | Merged openstack/nova master: Send soft_delete from context manager https://review.openstack.org/476459 | |
| 14:07:37 | mriedem | https://bugs.launchpad.net/nova/+bug/1794996 | |
| 14:07:37 | openstack | Launchpad bug 1794996 in OpenStack Compute (nova) "_destroy_evacuated_instances fails and kills n-cpu startup if lazy-loading flavor on a deleted instance" [High,Triaged] | |
| 14:20:05 | jungleboyj | melwitt: You around? | |
| 14:20:16 | melwitt | jungleboyj: yes | |
| 14:20:22 | jungleboyj | melwitt: Morning! | |
| 14:20:25 | jungleboyj | Quetion for you. | |
| 14:20:30 | melwitt | gmorning | |
| 14:21:04 | jungleboyj | So, talking to a customer asking about being able to replicate instances between multiple data centers. | |
| 14:21:20 | jungleboyj | So, for DR, where if one DC goes down the other can just pick up. | |
| 14:21:35 | jungleboyj | melwitt: Nova doesn't currently have anything like that. Right? | |
| 14:22:27 | mriedem | jungleboyj: no | |
| 14:22:38 | mriedem | the root volume attach/detach blueprint is semi related though | |
| 14:22:55 | mriedem | https://review.openstack.org/#/c/600628/ | |
| 14:23:09 | mriedem | detach root volume from DC1, attach new root volume from DC2 | |
| 14:23:27 | melwitt | jungleboyj: you mean like some sort of instance HA? we don't have that. downstream, we have some kind of scripts that do something to handle the use case but I'm not too familiar with them | |
| 14:23:42 | jungleboyj | mriedem: Ok. | |
| 14:23:57 | mriedem | jungleboyj: kingbird is supposed to be related to this i think https://wiki.openstack.org/wiki/Kingbird in some way | |
| 14:24:05 | jungleboyj | melwitt: Ah, so there are ways to make it happen but nothing that is community developed? | |
| 14:24:12 | jungleboyj | Or supported by Nova? | |
| 14:24:35 | mriedem | and freezer? https://docs.openstack.org/freezer/latest/ | |
| 14:24:49 | mriedem | nova does not natively orchestrate DR | |
| 14:24:59 | mriedem | masakari deals with instance HA | |
| 14:25:10 | mriedem | freezer and kingbird are about backups/DR/replication i think | |
| 14:25:52 | jungleboyj | So that is what all those other projects are about. :-) | |
| 14:27:48 | jungleboyj | So, it looks like the answer is that there are projects out there working on this kind of problem. | |
| 14:28:53 | melwitt | jungleboyj: yeah, I mean, we've talked about this before and DR/HA would be best orchestrated outside of nova by another service/entity. and I wasn't familiar with the projects mriedem has mentioned ^ | |
| 14:29:13 | jungleboyj | melwitt: We both learned something. :-) | |
| 14:30:43 | melwitt | yep :) | |
| 14:35:09 | jungleboyj | mriedem: Thank you very much. Very helpful. | |
| 14:36:07 | mdbooth | mriedem: That bug iss an immediate practical win for your new test :) | |
| 14:41:31 | mdbooth | mriedem: Side note, I'm not a fan of using temporary_mutation on a context object, because context objects are shared between threads in various places. That could produce some incredibly weird behaviour would would be almost impossible to debug. | |
| 14:42:42 | mdbooth | Better to use a copy | |
| 14:43:48 | mdbooth | Outside of tests it's only ever used to set read_deleted on a context object, so probably best to add a method to context which returns a copy with read deleted set | |
| 14:44:10 | mdbooth | -> no possibility of weird threading issues | |
| 14:46:08 | mriedem | mdbooth: i've got a recreate regression functional test and it's an easy fix on top, | |
| 14:46:13 | mriedem | just need to clean up these patches | |
| 14:46:50 | mriedem | and yes it's amazing what you find when you actually have integration testing for features... :) | |
| 15:00:45 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add functional regression test for bug 1794996 https://review.openstack.org/606106 | |
| 15:00:45 | openstack | bug 1794996 in OpenStack Compute (nova) rocky "_destroy_evacuated_instances fails and kills n-cpu startup if lazy-loading flavor on a deleted instance" [High,Triaged] https://launchpad.net/bugs/1794996 | |
| 15:12:10 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Ignore forcing of evacuation for nested instance https://review.openstack.org/606111 | |
| 15:12:38 | giblet | mriedem: FYI I applied the same ignore-the-force-flag tactic for the force evacuation with nested allocation case ^^ | |
| 15:22:38 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Run negative server moving tests with nested RPs https://review.openstack.org/604125 | |
| 15:36:41 | openstackgerrit | Elod Illes proposed openstack/nova master: Reject networks with QoS policy https://review.openstack.org/570079 | |
| 15:48:00 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add functional regression test for bug 1794996 https://review.openstack.org/606106 | |
| 15:48:00 | openstack | bug 1794996 in OpenStack Compute (nova) rocky "_destroy_evacuated_instances fails and kills n-cpu startup if lazy-loading flavor on a deleted instance" [High,Triaged] https://launchpad.net/bugs/1794996 | |
| 15:48:00 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Fix InstanceNotFound during _destroy_evacuated_instances https://review.openstack.org/606122 | |
| 15:53:56 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add functional regression test for bug 1794996 https://review.openstack.org/606106 | |
| 15:53:56 | openstack | bug 1794996 in OpenStack Compute (nova) "_destroy_evacuated_instances fails and kills n-cpu startup if lazy-loading flavor on a deleted instance" [High,In progress] https://launchpad.net/bugs/1794996 - Assigned to Matt Riedemann (mriedem) | |
| 15:53:56 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Fix InstanceNotFound during _destroy_evacuated_instances https://review.openstack.org/606122 | |
| 15:53:56 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Run evacuate tests with local/lvm and shared/rbd storage https://review.openstack.org/604400 | |
| 15:53:57 | openstackgerrit | Matt Riedemann proposed openstack/nova master: DNM: Test against mdbooth's evacuate patch https://review.openstack.org/604423 | |
| 15:55:58 | mriedem | melwitt: please to be approving backport to queens https://review.openstack.org/#/c/602409/ | |
| 15:56:08 | mriedem | oh and the change below it for testing | |
| 16:06:52 | mriedem | ocata backport https://review.openstack.org/#/c/570653/ is ready to go | |
| 16:08:33 | mriedem | and https://review.openstack.org/#/c/599885/1 | |
| 16:08:56 | cfriesen | when adding a new flavor extra-spec, what's the appropriate place to document it? embedded docs, reno, both? | |
| 16:09:02 | mriedem | and https://review.openstack.org/#/c/583422/ | |
| 16:09:19 | mriedem | cfriesen: https://docs.openstack.org/nova/latest/user/flavors.html#extra-specs | |
| 16:09:25 | mriedem | and reno | |
| 16:09:35 | cfriesen | thanks | |
| 16:10:46 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/ocata: Fix host validity check for live-migration https://review.openstack.org/590611 | |
| 16:12:31 | mriedem | stable core should just fast approve this https://review.openstack.org/#/q/I9435fe84c8dc8a6f35ffb195ddbf71ca34edb242+status:open | |
| 16:12:42 | mriedem | it avoids us running the ironic job on docs-only changes in stable | |
| 16:17:45 | mdbooth | cfriesen: Are you in the US, btw? | |
| 16:18:44 | mdbooth | cfriesen: The evacuate anti-affinity bug again: https://review.openstack.org/#/c/605436/ now with support for max_server_per_host | |
| 16:18:56 | mdbooth | That actually made it a bit simpler :) | |
| 16:19:17 | mdbooth | I feel like we're squashing quite a lot of evacuate bugs | |
| 16:20:25 | mdbooth | cfriesen: Apologies, Canada. Of course I knew that... | |
| 16:20:45 | mdbooth | I had it in my head you were in the UK for some bizarre reason. | |
| 16:21:43 | mriedem | ocata-em call to arms email sent | |
| 16:22:40 | cfriesen | mdbooth: no worries, we have the same queen. :) | |
| 16:22:48 | mdbooth | Incidentally, as a bonus the above patch will also correctly handle multiple builds from the same anti-affinity group sent to a single compute. Previously they could all be rejected, now it will always allow max_server_per_host to succeed | |
| 16:23:40 | mdbooth | That's builds in addition to rebuilds | |
| 16:31:26 | mdbooth | Is there some documentation somewhere of the query language used by logstash? | |
| 16:31:44 | mdbooth | This was what got me last time I tried to use this. | |
| 16:31:48 | cfriesen | mdbooth: do we still need the nested contexts given we're basically doing refcounting? I mean, it's an elegant way to ensure we call builders.remove(instance.uuid) but we could presumably also do that other ways. Just wondering if it'd end up being less code churn. | |
| 16:32:32 | mdbooth | cfriesen: I'd strongly prefer to keep the nested contexts | |
| 16:32:49 | mdbooth | Otherwise you'll end up having disjoint code, which is a maintenance nightmare | |
| 16:33:07 | mdbooth | and we have enough of it already | |
| 16:34:10 | mdbooth | cfriesen: Consider all the error cleanup edge cases you'd have to consider | |
| 16:34:17 | cfriesen | fair enough...just thought I'd mention it | |
| 16:35:46 | mdbooth | Incidentally, it also makes sense to do this test *before* taking a claim, as it's really scheduler take 2 | |
| 16:37:54 | cfriesen | arguably both the claim and the affinity validation are both scheduler take 2. | |
| 16:38:37 | cfriesen | but sure, it'd be less wasted work to check affinity first | |