| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-04-14 | |||
| 14:26:18 | sean-k-mooney | gibi: yep agree with dansmith | |
| 14:26:24 | gibi | sean-k-mooney: cool | |
| 14:26:59 | sean-k-mooney | speaking of ill go file said bug before i get distracted | |
| 14:27:14 | gibi | sean-k-mooney: thanks | |
| 14:33:33 | sean-k-mooney | gibi: should i mark this as rc potential or shoudl we just treat it as a normal bugfix that we will backport after ussuri is released as normal. | |
| 14:33:58 | stephenfin | gmann: Thanks for the explanation. Would it be possible edit the commit message to include that information there? | |
| 14:34:57 | gibi | sean-k-mooney: what is the worst case issue by that bug? | |
| 14:35:42 | gmann | stephenfin: sure, doing. | |
| 14:36:36 | sean-k-mooney | if we create the ARQ and fail to clean them up becasue binding fail then we leak the arq. since its not bound i think its just leaking a db recored. i need to double check that | |
| 14:37:45 | sean-k-mooney | gibi: https://bugs.launchpad.net/nova/+bug/1872730 | |
| 14:37:45 | openstack | Launchpad bug 1872730 in OpenStack Compute (nova) "Delete ARQs for an instance when the instance is deleted only delete bound arqs" [High,Triaged] | |
| 14:38:10 | gibi | sean-k-mooney: then I think it is not an rc potential bug | |
| 14:38:13 | openstackgerrit | Merged openstack/nova master: Introduce scope_types in server topology https://review.opendev.org/717584 | |
| 14:38:19 | openstackgerrit | Merged openstack/nova master: Add new default roles in server topology policies https://review.opendev.org/717585 | |
| 14:38:27 | openstackgerrit | Merged openstack/nova master: workarounds: Add option to locally attach RBD volumes on compute hosts https://review.opendev.org/708029 | |
| 14:38:42 | gibi | sean-k-mooney: thanks for fileing it | |
| 14:40:02 | sean-k-mooney | gibi: ya i might ping thet cyborg folk to take a look too and confirm its not going to leak devices on there end. but my understanding is they dont allocate or program the device until we bind it as that is when we set the host and pass the allocation to cyborg | |
| 14:40:16 | sean-k-mooney | so if we failed to bind it then they should not need to clean anything up | |
| 14:40:30 | sean-k-mooney | we should be deleted and recreating the arqs when we reschdule | |
| 14:40:34 | gibi | I have a similar understanding, thanks for following it up | |
| 14:45:41 | openstackgerrit | Merged openstack/nova master: fup: Add removal TODOs for disable_native_luksv1 and rbd_volume_local_attach https://review.opendev.org/718099 | |
| 14:45:52 | openstackgerrit | Merged openstack/nova master: Add new default roles in os-evacuate policies https://review.opendev.org/708237 | |
| 14:52:26 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Add new default roles in server group policies https://review.opendev.org/717175 | |
| 14:52:47 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Fix servers policy for admin_or_owner https://review.opendev.org/718501 | |
| 14:55:16 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Fix servers policy for admin_or_owner https://review.opendev.org/718501 | |
| 14:55:50 | gmann | stephenfin: done. also this one too - https://review.opendev.org/#/c/718501/ | |
| 14:56:09 | gmann | stephenfin: once you are ok with those then i will rebase the series | |
| 14:56:25 | openstackgerrit | Merged openstack/nova master: Pass the target in os-services APIs policy https://review.opendev.org/676688 | |
| 14:56:32 | openstackgerrit | Merged openstack/nova master: Pass the actual target in server group policy https://review.opendev.org/717176 | |
| 14:56:40 | openstackgerrit | Merged openstack/nova master: Add new default roles in server external events policies https://review.opendev.org/717168 | |
| 14:57:55 | stephenfin | gmann: comment on https://review.opendev.org/#/c/717175/10 | |
| 14:58:21 | stephenfin | I could be wrong though | |
| 14:59:38 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Add new default roles in server group policies https://review.opendev.org/717175 | |
| 15:00:10 | gmann | stephenfin: i think yes. sorry i update that directly. | |
| 15:07:37 | openstackgerrit | Merged openstack/nova master: Pass the actual target in server external events policy https://review.opendev.org/717169 | |
| 15:41:07 | gmann | johnthetubaguy: one comment, same for extended attr also - https://review.opendev.org/#/c/719730/1/nova/policies/servers.py@133 | |
| 16:09:39 | openstackgerrit | Takashi Natsume proposed openstack/nova master: Update compute rpc version alias for ussuri https://review.opendev.org/719315 | |
| 16:13:56 | gmann | melwitt: lyarwood master patch is merged for functional tox fix, train one is ready - https://review.opendev.org/#/c/719120/ | |
| 16:41:58 | bauzas | can someone tell me why libvirt returns me a former guest that was resized to another host ? | |
| 16:42:15 | bauzas | kashyap: stephenfin: sean-k-mooney: in case you know ? ^ | |
| 16:42:36 | bauzas | tbc, the resize was *confirmed* | |
| 16:42:51 | bauzas | and the allocations diseappeared accordingly | |
| 16:42:57 | bauzas | lyarwood: maybe ? | |
| 16:44:05 | kashyap | bauzas: Hi, need more context. But I'm hungry and just about to cook a meal. | |
| 16:44:13 | bauzas | kashyap: https://github.com/openstack/nova/blob/9001d7e3459d97f507e8ce638d1fc3935401252d/nova/virt/libvirt/driver.py#L7023 | |
| 16:44:16 | kashyap | Trying to maintain a Good Hermit Schedule(tm) :) | |
| 16:46:12 | bauzas | kashyap: anyway, take your time, I'll investigate tonight so we could discuss this tomorrow | |
| 16:46:25 | kashyap | bauzas: Hmm, that requires me to read code and delve deeper into context. I'm just recovering from 3 hours of debugging another libvirt bug | |
| 16:46:30 | kashyap | Yep, thanks! | |
| 17:44:17 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Add new default roles in quota class policies https://review.opendev.org/719100 | |
| 17:49:33 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Pass the actual target in quota class policy https://review.opendev.org/720042 | |
| 17:58:08 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Pass the actual target in flavor access policy https://review.opendev.org/719607 | |
| 18:06:12 | openstackgerrit | Rafael Weingärtner proposed openstack/nova master: Enhance os-simple-tenant-usage VM hours accounting https://review.opendev.org/711113 | |
| 18:06:12 | openstackgerrit | Rafael Weingärtner proposed openstack/nova master: Enhance os-simple-tenant-usage VM hours accounting https://review.opendev.org/711113 | |
| 18:16:02 | artom | melwitt, I just noticed your have 3 Ts in your Gerrit email | |
| 18:16:22 | artom | Took some head banging to understand why it wasn't accepting my script's query | |
| 18:16:23 | artom | :) | |
| 19:40:17 | bauzas | erm, do folks here have any ideas on how to get a right CONF.host value within a functional test ? | |
| 19:40:51 | bauzas | I mean, I could self.flags it but given it's a global object, all the created computes in the functest will get the same valur | |
| 19:40:53 | bauzas | value* | |
| 20:29:45 | melwitt | bauzas: yes, you do that by passing the 'host' kwarg when you start_service https://github.com/openstack/nova/blob/9001d7e3459d97f507e8ce638d1fc3935401252d/nova/test.py#L381-L391 | |
| 20:30:23 | bauzas | melwitt: sure, I saw it | |
| 20:30:37 | bauzas | melwitt: but for example, I start two computes in a functest | |
| 20:30:42 | bauzas | (using this helper) | |
| 20:31:04 | melwitt | artom: yeah ... the 2 Ts version is owned by someone else and I was uncreative so I just tacked on another t. probably should have picked something else, in hindsight | |
| 20:31:50 | bauzas | melwitt: but then, when I want to confirm resize an instance, the global CONF.host is only having the last host value | |
| 20:31:51 | bauzas | https://github.com/openstack/nova/blob/9001d7e3459d97f507e8ce638d1fc3935401252d/nova/virt/libvirt/driver.py#L1519 | |
| 20:32:47 | melwitt | bauzas: oh, I see. I don't know of a way around that. you would probably need the ghost of mriedem. or maybe gibi might have ideas | |
| 20:33:12 | bauzas | melwitt: so, for example, if I create an instance for host1, resize it to host2 and then confirm the resize, it won't remove the guest in host1 because instance.host (host2) == CONF.host (host2) even I checked that it was calling host1 | |
| 20:33:38 | bauzas | melwitt: no worries, I'll just then modify the CONF before calling confirm_resize | |
| 20:36:04 | mriedem | isn't there another handle within the libvirt driver to the host value used when the service was started? | |
| 20:36:06 | mriedem | the compute service i mean | |
| 20:36:16 | mriedem | if so, you can change that code to use that and not deal with global config problems | |
| 20:38:11 | mriedem | the compute driver has a handle to the virtapi which has a handle to the compute manager which has a self.host value | |
| 20:39:09 | bauzas | mriedem: yeah, no worries, I can provide a change for it | |
| 20:39:25 | mriedem | so i think it's just `self.virtapi._compute.host` | |
| 20:39:40 | melwitt | oh yeah, self.host, that will be different even in func tests. good call | |
| 20:40:06 | mriedem | you could add a "host" property getter method to the virtapi so that the driver doesn't need to know about self._compute | |
| 20:40:23 | mriedem | self.virtapi.host | |
| 20:40:25 | mriedem | badabing | |
| 20:41:09 | bauzas | thanks | |
| 20:41:28 | bauzas | fwiw, we have a lot of CONF.host values in libvirt | |
| 20:41:54 | bauzas | so I'll just create a change for all of them | |
| 20:45:03 | artom | bauzas, yeah, there's not much you can do with the global CONF.host... | |
| 20:45:19 | artom | Although the computes should keep their hostnames... | |
| 20:45:24 | artom | So maybe there's a bug somewhere? | |
| 20:45:48 | bauzas | artom: that's not really a bug | |
| 20:45:51 | artom | Oh, wait, melwitt linked the "wrong" helper | |
| 20:45:55 | bauzas | artom: it's just a problem for tests | |
| 20:46:15 | artom | bauzas, are you using the fake driver, or the libvirt driver? | |
| 20:46:22 | bauzas | the latter | |
| 20:46:31 | bauzas | hence the problem | |
| 20:46:58 | bauzas | anyway, I found the solution | |
| 20:47:00 | melwitt | artom: what do you mean "wrong" helper? | |
| 20:47:12 | artom | bauzas, https://github.com/openstack/nova/blob/9001d7e3459d97f507e8ce638d1fc3935401252d/nova/tests/functional/libvirt/base.py#L117 | |
| 20:47:32 | bauzas | I all know about it | |
| 20:47:58 | bauzas | but again, CONF.host will be overrided by the last value | |
| 20:48:16 | bauzas | given it's a global variable | |
| 20:48:21 | artom | bauzas, no way around it, then | |
| 20:48:26 | artom | CONF is global per process | |