| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-08-09 | |||
| 01:09:51 | dansmith | sweet | |
| 01:10:14 | mriedem | it also helped me notice that we're not indexing the super-cond/cond-cell1 logs in logstash | |
| 01:10:16 | mriedem | fix is up for that | |
| 01:10:39 | dansmith | oh interesting | |
| 01:10:55 | mriedem | system-config repo has a yaml of the files to index | |
| 01:11:01 | mriedem | with the new names we didn't account for those | |
| 01:19:35 | clarkb | mriedem: ya we emailed them today | |
| 01:19:59 | clarkb | then being citycloud | |
| 01:27:19 | mriedem | cool | |
| 01:56:38 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add release note for shared storage known issue https://review.openstack.org/491582 | |
| 01:56:38 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Mark max microversion for Pike in history doc https://review.openstack.org/491581 | |
| 01:56:39 | openstackgerrit | Matt Riedemann proposed openstack/nova master: doc: provide more details on scheduling with placement https://review.openstack.org/491900 | |
| 01:56:39 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add a prelude section for Pike https://review.openstack.org/491424 | |
| 02:01:26 | mriedem | stephenfin: can you drop this -2 now? https://review.openstack.org/#/c/482216/ | |
| 03:01:22 | openstackgerrit | Merged openstack/nova master: Add functional test for local delete allocations https://review.openstack.org/470578 | |
| 03:10:50 | openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi: Live migration failed in xapi pool https://review.openstack.org/489451 | |
| 03:28:23 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted_certs to Instance object https://review.openstack.org/489408 | |
| 03:29:38 | openstackgerrit | Jackie Truong proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949 | |
| 03:34:53 | openstackgerrit | Alex Xu proposed openstack/nova master: placement: ensure RP maps to those RPs that share with it https://review.openstack.org/480379 | |
| 03:46:37 | openstackgerrit | Michael Still proposed openstack/nova master: Avoid chowning console logs in libvirt https://review.openstack.org/472229 | |
| 03:46:38 | openstackgerrit | Michael Still proposed openstack/nova master: Move execs of touch to privsep. https://review.openstack.org/489190 | |
| 03:46:38 | openstackgerrit | Michael Still proposed openstack/nova master: First attempt at adding a privsep user to nova itself. https://review.openstack.org/459166 | |
| 03:46:39 | openstackgerrit | Michael Still proposed openstack/nova master: Move execs of tee to privsep. https://review.openstack.org/489438 | |
| 03:46:39 | openstackgerrit | Michael Still proposed openstack/nova master: Move libvirts dmcrypt support to privsep. https://review.openstack.org/490737 | |
| 03:46:40 | openstackgerrit | Michael Still proposed openstack/nova master: Move libvirt usages of chown to privsep. https://review.openstack.org/471972 | |
| 03:46:41 | openstackgerrit | Michael Still proposed openstack/nova master: Refactor libvirt.utils.execute() away. https://review.openstack.org/489816 | |
| 03:46:41 | openstackgerrit | Michael Still proposed openstack/nova master: Read from console ptys using privsep. https://review.openstack.org/489486 | |
| 06:59:40 | openstackgerrit | huangtianhua proposed openstack/python-novaclient master: Allow boot server with multiple nics https://review.openstack.org/492003 | |
| 07:24:06 | openstackgerrit | Maciej Jozefczyk proposed openstack/nova master: Remove host filter for _cleanup_running_deleted_instances periodic task https://review.openstack.org/491808 | |
| 08:16:05 | openstackgerrit | huangtianhua proposed openstack/python-novaclient master: Remove substitutions for command error msg https://review.openstack.org/490705 | |
| 08:34:21 | gibi | good morning | |
| 09:05:35 | jianghuaw | Is the DB nova-api allowed to be accessed by nova-compute service? | |
| 09:07:47 | jianghuaw | I met an error as "RemoteError: Remote error: CantStartEngineError No sql_connection parameter is established" when nova-compute tries to query data from aggregate which belong to nova-api db. | |
| 14:00:52 | openstackgerrit | Matthew Edmonds proposed openstack/nova master: update policy UT fixtures https://review.openstack.org/398610 | |
| 14:12:19 | mriedem | i'm going through https://review.openstack.org/#/c/491850/ now | |
| 14:13:22 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Improve stable-api doc with current API state https://review.openstack.org/489926 | |
| 14:18:52 | openstackgerrit | Maciej Jozefczyk proposed openstack/nova master: Remove host filter for _cleanup_running_deleted_instances periodic task https://review.openstack.org/491808 | |
| 14:20:29 | openstackgerrit | Matthew Edmonds proposed openstack/nova master: use conf for keystone session creation https://review.openstack.org/485121 | |
| 14:21:45 | dtantsur | dansmith: hi! re your comment on the ironic-related patch: should we block changing node.resource_class for active nodes in Ironic? | |
| 14:22:07 | dansmith | dtantsur: if that's possible I think that would be an excellent idea | |
| 14:22:50 | dtantsur | dansmith: it's not impossible, but our beloved API microversion will kick in here. meaning, we'll only be able to block it starting with the next API version :( | |
| 14:23:13 | dtantsur | I would actually block it in all versions, given that it's going to screw up nova | |
| 14:23:25 | dansmith | dtantsur: presumably you can return a 409 for mostly any reason right? | |
| 14:23:37 | dtantsur | but people tend to feel quite religiously about not bypassing the versioning | |
| 14:24:00 | dtantsur | dansmith: right, but how does it help? | |
| 14:25:22 | dtantsur | it's an interesting corner case of the API WG to discuss. should we leave a feature that clearly breaks things, or should we break the versioning contract | |
| 14:25:46 | dansmith | dtantsur: well can you return 409 for anything else in that call? | |
| 14:26:14 | dtantsur | dansmith: sorry, I think I don't get the question. We cannot just randomly return 409 I think.. | |
| 14:27:02 | mriedem | dtantsur: mearning, can the node update api return a 409 already for something else | |
| 14:27:09 | dansmith | dtantsur: if 409 is already a valid return value then I think it's less problematic | |
| 14:27:10 | dansmith | right | |
| 14:27:11 | mriedem | so the user can already be expecting a 409 in some cases | |
| 14:27:34 | dansmith | and I would expect 409 to be valid for most PUTs to cover situations like this | |
| 14:27:41 | dansmith | like "you're violating some constraint" | |
| 14:27:45 | dtantsur | mriedem, dansmith, I think I get where you're heading. Yes, we can. And no, according to the API versioning ideology, we cannot do it. | |
| 14:28:03 | dtantsur | it's not only about breaking users, it's more about signaling changes /me waits for mordred to jump in | |
| 14:28:13 | mordred | so... | |
| 14:28:13 | mriedem | https://developer.openstack.org/api-ref/baremetal/#update-node | |
| 14:28:17 | mriedem | doesn't mention error codes | |
| 14:28:18 | dtantsur | (for the record: I'm not the biggest fan of the API versioning here) | |
| 14:28:30 | mordred | for signalling changes, it's to signal whether someone can do something or not | |
| 14:28:38 | dansmith | dtantsur: well, regardless of whether ironic lets you do it, the rule to operators should be that they will break stuff if they do it once things are populated | |
| 14:28:41 | mordred | the thing in this case isn't a valid thing to try to do | |
| 14:28:58 | mordred | so a user who does it today isn't actually doing a thing that works | |
| 14:29:13 | mordred | so they don't actually have, you know, an application that is going to break when you do this | |
| 14:29:28 | dtantsur | fair | |
| 14:29:58 | dansmith | that's true, although the application here is *probably* their ansible playbook | |
| 14:29:59 | mriedem | you've got a 409 right here https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L1698 | |
| 14:30:07 | mordred | alternately, they don't need to ask the API if they can avoid doing the bad state -they can, as a client, avoid writing broken code without setting a microversion, since this is mostly about ironic returning an error when they do something stupid | |
| 14:30:09 | mordred | SO | |
| 14:30:19 | dansmith | where they've set a bunch of nodes to be a given class and will blindly blow that into ironic, I'm guessing | |
| 14:30:36 | mordred | I'd say this is a bug fix where requiring a version bump does't actually add value to anyone | |
| 14:30:37 | dtantsur | mriedem: let's please not use 409 though. we've done a big mistake in our youth, and now ironicclient retries it | |
| 14:30:51 | dansmith | eww | |
| 14:30:57 | dansmith | mordred: agreed | |
| 14:30:57 | dtantsur | however, the generic 400 an also be returned from essentially any endpoint, we can use it | |
| 14:31:07 | dtantsur | mordred: I like the way you put it :) thanks! | |
| 14:31:27 | mordred | dtantsur: I will happily advocate for the validity of this change not actually being an API break if you need me to | |
| 14:31:39 | mordred | rules are there to help us do the right thing - they're not there for their own sake | |
| 14:31:45 | dtantsur | right | |
| 14:32:20 | dtantsur | next tricky question :) | |
| 14:32:45 | dtantsur | what should we do about nodes that do not have any resource_class so far? I mean, active nodes? | |
| 14:32:57 | dtantsur | dansmith: ^^ | |
| 14:33:13 | dansmith | dtantsur: if you take the strict meaning of mordred's comments above, | |
| 14:33:28 | dansmith | then you aren't breaking a node that is active with no RC since we're not doing anything with those yet | |
| 14:33:29 | dansmith | however, | |
| 14:33:36 | dansmith | I think it's far more confusing to allow that | |
| 14:33:43 | dansmith | not worth the confusion over consistency | |
| 14:34:13 | mriedem | jaypipes: dansmith: i'm inclined to -1 this for missing tests https://review.openstack.org/#/c/491850/ but given dansmith is out the next two days i realize there is a need to start pushing this code through | |
| 14:34:15 | dtantsur | this is what I'm thinking about. a user creates an instance back in Ocata, no resource_class set. they upgrade to Pike, then to Queens. still no resource_class. | |
| 14:34:39 | dansmith | dtantsur: ah, right, for nodes with an instance | |
| 14:34:41 | mordred | dtantsur: what would the resource_class be if they created it with an empty resource_class today? | |
| 14:34:59 | dtantsur | mordred: None | |
| 14:35:08 | dtantsur | (python None, or JSON null) | |
| 14:35:09 | dansmith | mordred: it's a thing that is required for queens | |
| 14:35:15 | mordred | AH | |
| 14:35:48 | dansmith | dtantsur: so the response probably needs to be "400: You cannot CHANGE the class of an active node" | |
| 14:35:52 | dtantsur | so yeah, it was perfectly valid to not have any resource_class when we introduced it (which is sad, btw) | |
| 14:35:59 | dansmith | dtantsur: sad indeed | |
| 14:36:11 | dansmith | heh | |