| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-04-09 | |||
| 18:48:25 | mriedem | the default matches what's in the out of tree driver, | |
| 18:48:30 | mriedem | but not what you'd get today with the in-tree driver | |
| 18:48:36 | mriedem | i don't know how you want to resolve that | |
| 18:48:42 | edmondsw | mriedem that's part of the reason for this change | |
| 18:48:42 | mriedem | i'm inclined to not care about the out of tree default | |
| 18:49:03 | edmondsw | the default in the nova driver was not great... and we definitely want it to be consistent | |
| 18:49:43 | edmondsw | it was always meant to be 0.1 by default | |
| 18:49:59 | mriedem | why not change the default in the library? | |
| 18:51:19 | edmondsw | mriedem because pypowervm is used by folks, whereas the powervm driver in nova isn't yet | |
| 18:51:54 | mriedem | if folks are using the out of tree driver, they are getting the out of tree default config opt value, not what's in the library | |
| 18:51:55 | mriedem | yeah? | |
| 18:52:00 | mriedem | but lowly in-tree driver is | |
| 18:52:51 | edmondsw | I mean folks are using pypowervm directly, not only via the oot driver | |
| 18:53:02 | mriedem | folks == powervm? | |
| 18:53:05 | mriedem | *powervc | |
| 18:53:19 | edmondsw | not powervc... it uses nova-powervm | |
| 18:53:20 | edmondsw | but others | |
| 18:53:26 | mriedem | oVirt on Power | |
| 18:53:27 | mriedem | got it :) | |
| 18:53:38 | edmondsw | non-OpenStack folks | |
| 18:54:21 | edmondsw | basically, this is a goof, we release changing the default in nova is lousy, but it was just the best solution in this case | |
| 18:54:29 | edmondsw | s/release/realize/ | |
| 18:54:31 | mriedem | ok, so you don't want to change the default in the library, i get it. then esberglu it's an upgrade release note, not feature | |
| 18:54:43 | edmondsw | yep | |
| 18:55:42 | mriedem | well, it's honestly probably both | |
| 18:56:00 | mriedem | both a new option (feature) and an upgrade since the default for that new option changes the default | |
| 18:56:05 | mriedem | on an existing extra spec | |
| 18:56:23 | mriedem | just do both in a single reno file | |
| 18:56:36 | esberglu | mriedem: Sounds good | |
| 18:56:57 | esberglu | edmondsw: tx for the help, I didn't have as much context, just wanted to fix CI :) | |
| 18:58:04 | edmondsw | anytime :) | |
| 19:11:20 | mriedem | huh, it probably surprises no one that when we shelve offload a server, we null out the host but not it's AZ value | |
| 19:12:47 | mriedem | this is the time where cfriesen_ is supposed to ask if anyone uses shelve | |
| 19:16:29 | openstack | Launchpad bug 1759924 in OpenStack Compute (nova) "Port device owner isn't updated with new host availability zone during unshelve" [Medium,Triaged] - Assigned to Matt Riedemann (mriedem) | |
| 19:16:29 | mriedem | jmlowe_: hey i'm trying to recreate https://bugs.launchpad.net/nova/+bug/1759924/ on a single node devstack system | |
| 19:16:52 | mriedem | jmlowe_: and things are find until i shelve, rename the az from az1 to az2, and then try to unshelve the instance | |
| 19:17:02 | mriedem | it blows up because the AZ filter is looking for az1 which is now gone | |
| 19:17:08 | jmlowe_ | yes | |
| 19:17:29 | mriedem | oh in yours you unshelved to a different az/host | |
| 19:17:31 | jmlowe_ | unshelve will put it anywhere if the original spec doesn't have an az | |
| 19:17:44 | jmlowe_ | correct | |
| 19:17:55 | mriedem | oh yeah, i created this instance with az1 specifically, so it's in the request spec, yeah i'll fix that and retry | |
| 19:18:35 | mriedem | this kind of seems like a bug of it's own, but not sure what we'd do about it | |
| 19:19:05 | mriedem | we'd have to update any shelved instances in an az if we rename that az | |
| 19:20:33 | jmlowe_ | if nothing else, where ever an instance is unshelved it needs to update the device_owner because it is there and that is truth | |
| 19:20:50 | mriedem | jmlowe_: yeah i'm trying to get to that part | |
| 19:21:03 | mriedem | jmlowe_: while you've got az renames on the brain, maybe you'd like to reply to http://lists.openstack.org/pipermail/openstack-operators/2018-March/015029.html with your thoughts | |
| 19:22:06 | jmlowe_ | oh, I effectively did that by putting all my computes into rackwise az | |
| 19:22:28 | jmlowe_ | to get migration to work again I wound up editing the specs | |
| 19:22:47 | mriedem | i bet updating json blobs in the db is super fun | |
| 19:28:50 | mriedem | jmlowe_: ack confirmed http://paste.openstack.org/show/718769/ | |
| 19:36:08 | openstackgerrit | Surya Seetharaman proposed openstack/nova master: Cleanup RP and HM records while deleting a compute service. https://review.openstack.org/554920 | |
| 19:36:52 | jmlowe_ | mriedem: it was all the fun you might imagine it to be and then some | |
| 19:38:32 | jmlowe_ | if you could never change the az name because you could never change the spec then az's really become a nightmare for operators and should be avoided at all costs, probably not what people really want | |
| 19:45:38 | mriedem | jmlowe_: what if, for starters, as mentioned in that ML thread, we just didn't allow az renames while there were (alive) instances in those AZs? | |
| 19:45:49 | mriedem | so operators can't shoot themselves in the foot that way? | |
| 19:49:35 | jmlowe_ | Lets say you start with the default "nova", specs wind up with nova in it, how to you create any AZ if you didn't plan on it from the beginning? Do you have to get this right before any people start using your cloud and make sure your guesses never having run a cloud before were correct for all time? | |
| 19:49:55 | mriedem | is nova put into the request spec? | |
| 19:50:21 | mriedem | https://github.com/openstack/nova/blob/master/nova/compute/api.py#L895 | |
| 19:50:35 | mriedem | 'nova' shouldn't be in the request spec unless the user specifically requested --availability-zone nova | |
| 19:51:09 | jmlowe_ | Maybe not anymore, but my "no free hosts" live migration problems were because nova az was in the specs | |
| 19:51:42 | jmlowe_ | Fixes you didn't know you needed but always wanted | |
| 19:51:44 | mriedem | hmm, i'm not saying users couldn't wedge themselves, | |
| 19:51:51 | mriedem | the server create API does say to not specify 'nova' here https://developer.openstack.org/api-ref/compute/#id10 | |
| 19:52:36 | jmlowe_ | I think I would make more mistakes with renames disallowed than I would with it allowed | |
| 19:53:15 | mriedem | huh, this inconsistency in the CLI for non-admins isn't good either http://paste.openstack.org/show/718771/ | |
| 19:53:32 | jmlowe_ | it's kind of bad data modeling to have it all based on name and not id | |
| 19:53:48 | mriedem | jmlowe_: yeah that's jaypipes' point in the ML thread, | |
| 19:53:53 | mriedem | but is also a bigger hairier change | |
| 19:54:26 | jmlowe_ | I really don't have any answers over here, just the ability to create more problems for myself | |
| 19:54:39 | mriedem | heh, fair enough | |
| 19:55:44 | jmlowe_ | I'm looking at the possibility of going all in on host aggregates to keep certain customers on hardware they bought sometime later this summer | |
| 19:56:07 | jmlowe_ | none of this is going to get easier for me | |
| 19:56:07 | openstackgerrit | Merged openstack/nova master: Fix comments at the 'save' method of objects.Instance https://review.openstack.org/559743 | |
| 19:57:07 | jmlowe_ | I've observed that cli behavior before | |
| 19:57:51 | openstack | Launchpad bug 1762534 in python-openstackclient "openstack availability zone list shows default 'nova' AZ to non-admin users" [Undecided,New] | |
| 19:57:51 | mriedem | https://bugs.launchpad.net/python-openstackclient/+bug/1762534 | |
| 19:57:57 | mriedem | i'll have to dig into that later | |
| 20:09:46 | openstackgerrit | Merged openstack/nova master: Fix cancel_all_events event name parsing https://review.openstack.org/558059 | |
| 20:18:41 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Update port device_owner when unshelving https://review.openstack.org/559828 | |
| 20:18:43 | mriedem | jmlowe_: ^ | |
| 20:20:13 | jmlowe_ | That looks pretty good | |
| 20:29:00 | mriedem | oh it doesn't fix the bug in the single node devstack case... | |
| 20:29:16 | mriedem | because on shelve offload, we don't clear the port binding host id, so the conditional on unshelve isn't true | |
| 20:29:17 | jmlowe_ | really? looks like it should work | |
| 20:29:26 | mriedem | it would work in a >1 compute env :) | |
| 20:29:33 | jmlowe_ | oh, that's what somebody was on about when I first reported it | |
| 20:29:44 | jmlowe_ | clearing the port on shelve | |
| 20:29:52 | mriedem | that was me :) | |
| 20:30:03 | mriedem | and that's the todo i left in the code | |
| 20:30:06 | jmlowe_ | how convenient | |
| 20:32:37 | mriedem | i wanted to avoid fixing the shelve / clear case in the same patch since that gets slightly more complicated, and not really needed to fix this bug | |
| 20:32:56 | mriedem | unless your production env is 1 compute host... | |
| 20:33:01 | mriedem | maybe it's a mainframe | |
| 20:53:42 | openstackgerrit | Jay Pipes proposed openstack/nova master: mirror nova host aggregate members to placement https://review.openstack.org/553597 | |
| 21:12:39 | tblakes | dansmith: Could you please take a look at https://review.openstack.org/#/c/559158/ if you have a chance. It's gotten a +2 from Matt Riedemann but still needs to get a +1 on workflow. It's a cherry pick from master to stable/queens. | |
| 21:51:35 | imacdonn | dansmith: if you have a moment, could you evaluate https://review.openstack.org/#/c/558089/ , please? | |
| 22:00:16 | openstackgerrit | Eric Berglund proposed openstack/nova master: PowerVM: Add proc_units_factor conf option https://review.openstack.org/554688 | |
| 22:04:15 | dansmith | imacdonn: mriedem: sorry, I just can't get on board with that | |
| 22:04:41 | imacdonn | dansmith: OK.... do you have another idea ? | |