| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-09-21 | |||
| 13:12:32 | kashyap | mdbooth: Ah, thanks | |
| 13:12:39 | edmondsw | stephenfin wasn't that always the plan when we first deprecated them? | |
| 13:13:10 | mriedem | mdbooth: yes there are 2 patches | |
| 13:13:17 | edmondsw | either way... the powervm CI is going to be busted until we get something merged for this | |
| 13:13:31 | stephenfin | edmondsw: Yes and no. We do emit warnings that option have been renamed, but we've never gone and removed the old aliases | |
| 13:13:35 | mriedem | mdbooth: https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bug/1718295 | |
| 13:13:45 | mdbooth | mriedem: I found 1. Interested to know how this ended up on your radar, btw. | |
| 13:13:45 | edmondsw | stephenfin k | |
| 13:14:31 | stephenfin | efried: The fix I've suggested seemed far simpler and involved less other changes (fixing the the powervm CI patch, for one) so that's why I went with it | |
| 13:14:32 | mriedem | mdbooth: because the live migration job shit the bed the other day when devstack was changed to use pike uca packages | |
| 13:14:34 | mriedem | it's in the ML | |
| 13:14:45 | mdbooth | mriedem: Ah, ok. | |
| 13:15:03 | efried | stephenfin wfm. But how long are we planning to keep the original opts around? | |
| 13:15:03 | mriedem | jamespage pointed out the change in 2.10 | |
| 13:15:28 | stephenfin | That's a question for the PTL. mriedem - do you think we should ever remove old deprecated opt aliases? | |
| 13:15:49 | stephenfin | mriedem: i.e. where we moved an opt from '[DEFAULT]' to, say, '[libvirt]' | |
| 13:15:49 | efried | Last night his only complaint was the proliferation of renos :) | |
| 13:15:59 | kashyap | mdbooth: It's this bug - https://bugs.launchpad.net/nova/+bug/1718295 | |
| 13:16:01 | openstack | Launchpad bug 1718295 in OpenStack Compute (nova) "Live migration fails with qemu-img >= 2.10: "Failed to get shared "write" lock\nIs another process using the image?"" [High,In progress] - Assigned to Sean Dague (sdague) | |
| 13:16:13 | stephenfin | :D | |
| 13:16:18 | mdbooth | Interesting that UCA has newer libvirt/qemu than RHEL. | |
| 13:16:42 | stephenfin | efried: I was going to note that you can edit a reno in the same cycle, then noticed you actually did that. I can see where he's coming from, heh | |
| 13:17:36 | mdbooth | Or Fedora, even | |
| 13:17:42 | mdbooth | That's pretty darned fresh | |
| 13:17:43 | sdague | mriedem: I reved the patch to pass unit tests | |
| 13:17:56 | kashyap | Yeah, UCA is supposed to be stable/long-term | |
| 13:17:58 | kashyap | Isn't it? | |
| 13:18:02 | efried | stephenfin Yeah... I was following the instructions in the reno template which said you need to duplicate information. x This change seemed like it was applicable to both the 'upgrade' and the 'deprecation' sections. x I was doing both things (removing the old deprecated opt and renaming the new one). | |
| 13:18:17 | sdague | it turns out the broken hypervisor error pops a little earlier with this code path | |
| 13:18:20 | stephenfin | efried: gotcha | |
| 13:18:39 | sdague | kashyap: UCA is how they get latest bits to users. For the ubuntu package they unconditionally patched it | |
| 13:18:41 | stephenfin | efried: In any case, if you'd prefer to just remove the old alias then go for it. I would've proposed a separate patch altogether but I didn't want to de-emphasise your work reporting the bug/working on an initial fix | |
| 13:18:52 | sdague | because they know what version of qemu is required | |
| 13:19:11 | stephenfin | Though I do prefer this approach, given that we've never removed deprecated aliases before #justsaying | |
| 13:19:22 | efried | stephenfin No, I think it's appropriate to split those up. Removing the original opts would seem to be a separate effort following on from the original bp that did the split-out-of-DEFAULT in the first place. | |
| 13:19:31 | kashyap | sdague: Ah, latest bits; shipping QEMU 2.10 is a bit pre-mature with OpenStack, no? (But good for finding bugs, though.) | |
| 13:19:39 | stephenfin | efried: +1 | |
| 13:19:43 | efried | "never removed deprecated aliases before" ? | |
| 13:20:02 | efried | For what value of "never"? | |
| 13:20:04 | stephenfin | i.e. dropped 'deprecated_group' or 'deprecated_name' arguments | |
| 13:20:18 | sdague | kashyap: well someone has to ship things | |
| 13:20:20 | stephenfin | Like, ever, from what I can tell | |
| 13:20:49 | kashyap | Sure; question is not about shipping :-) There can be different repos for bleeding edge users. | |
| 13:21:19 | efried | I'm *sure* I've seen it done. But I'm not one to stand on principle and go find an example :) | |
| 13:21:28 | sdague | kashyap: yeh, well, the work around was easy for them. Honestly, I'm never going to ding people for deciding to run with what latest upstream things is fine. :) | |
| 13:21:57 | kashyap | Yep, agreed on that bit. | |
| 13:22:14 | kashyap | We still have to figure out that CI job we've been discussing on the list forever | |
| 13:22:42 | kashyap | Where latest libvirt / QEMU (from Git, newest upstream tarballs) are tested w/ Nova Git. | |
| 13:23:40 | mriedem | efried: stephenfin: deprecated options can go after a release | |
| 13:23:40 | jamespage | sdague, mriedem: I'm going to abandon https://review.openstack.org/#/c/505748/ - its way to brittle relying on calling code to know whether to lock or not. | |
| 13:23:50 | mriedem | jamespage: ok | |
| 13:24:06 | jamespage | mriedem: sdague's approach seems a bit more rugged... | |
| 13:24:18 | mriedem | kashyap: you're the one that's always saying we should run against latest libvirt/qemu :) | |
| 13:24:18 | sdague | well, mriedem actually wrote the patch | |
| 13:24:40 | kashyap | mriedem: Yes, I say it to be done in a _separate_ CI job :-) | |
| 13:24:49 | mriedem | oooooooooooooooooooooooooooooooooohhhhhhhhhhhhhhhhhhhhhhhhhh | |
| 13:24:53 | jamespage | must have missed that.. | |
| 13:24:56 | stephenfin | mriedem: This is more about the deprecated aliases, where we rename an opt or change its group, but still keep it around | |
| 13:24:56 | jamespage | apols | |
| 13:25:19 | stephenfin | efried: Proof or it didn't happen ;) | |
| 13:25:20 | sdague | this only showed up in live migration though, honestly, CI isn't going to catch every edge casee | |
| 13:25:32 | mriedem | stephenfin: deprecated is deprecated what's the difference | |
| 13:25:42 | mriedem | this config option deprecation stuff is grating on my brain | |
| 13:26:33 | stephenfin | mriedem: So I _shouldn't_ push a twenty-patch "remove deprecated opts" series? Aww :'( | |
| 13:28:36 | efried | The difference is just code cleanup / tech debt. We could keep every deprecated option around forever and it wouldn't hurt anything. But it sure would look messy. | |
| 13:31:30 | efried | stephenfin Oh, snap. Here's one. Guess who authored it? I41a6be48380999c49d6db3ee5221174a41014002 | |
| 13:32:45 | efried | reviewed, rather. | |
| 13:33:03 | efried | It was in the same blueprint as the one that moved the vnc opts originally (which was your change set IIRC) | |
| 13:34:23 | openstackgerrit | Merged openstack/os-vif master: Updated from global requirements https://review.openstack.org/502708 | |
| 13:34:41 | openstackgerrit | OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/505839 | |
| 13:35:04 | efried | stephenfin And here's one you did co-author: https://review.openstack.org/#/c/463220/ | |
| 13:35:06 | efried | I'll stop now. | |
| 13:36:56 | openstackgerrit | OpenStack Proposal Bot proposed openstack/os-traits master: Updated from global requirements https://review.openstack.org/503646 | |
| 13:38:36 | mriedem | efried: stephenfin: i've been dealing with getting my kid to school, so haven't been following along, what is the actual issue? | |
| 13:38:42 | mriedem | i know about the vnc rename thing | |
| 13:39:59 | dansmith | gibi: responded to some of your comments, thanks for looking at this stuff | |
| 13:40:58 | openstackgerrit | Radoslav Gerganov proposed openstack/nova master: VMware: implement get_inventory() driver method https://review.openstack.org/506175 | |
| 13:41:30 | efried | mriedem That's all. Just needing https://review.openstack.org/#/c/505831/ to merge pdq for the sake of a couple of 3rd party CIs that were testing with those opts. | |
| 13:43:41 | efried | powervm is kinda bound up between that and https://review.openstack.org/#/c/504194/ -- we need ^ to merge first so our rename of < will pass. | |
| 13:46:37 | mriedem | efried: +2 | |
| 13:46:53 | mriedem | alex_xu: sdague: can you take a look at https://review.openstack.org/#/c/505831/ ? unblocks some 3rd party CIs | |
| 13:47:31 | efried | Thanks! | |
| 13:47:37 | mriedem | efried: are you going to follow that up and drop support for the options in the DEFAULT group/ | |
| 13:47:40 | mriedem | which were deprecated | |
| 13:47:49 | efried | Can | |
| 13:49:25 | sdague | what's the no valid host error all about in CI? | |
| 13:49:30 | bauzas | oomichi: you're around ? see my comments in https://review.openstack.org/#/c/506092/1 ? | |
| 13:51:26 | openstackgerrit | Thomas Bechtold proposed openstack/nova stable/pike: Handle deleted instances when refreshing the info_cache https://review.openstack.org/506183 | |
| 13:52:11 | mriedem | fancy pants nova meeting in 8 minutes | |
| 13:52:42 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Fix definitely move single instance when created concurrently https://review.openstack.org/506093 | |
| 13:52:53 | bauzas | mriedem: you could have interest in ^ | |
| 13:53:31 | takashin | bauzas: Could you review https://review.openstack.org/#/c/408955/ ? | |
| 13:53:56 | bauzas | takashin: yeah, that's in my pipe for a long way :) | |
| 13:54:17 | takashin | bauzas: Thank you. | |
| 13:54:19 | bauzas | takashin: now we merged the spec, let's go back on the implementation side :) | |
| 13:54:31 | mriedem | bauzas: yeah, after the meeting | |
| 13:54:45 | bauzas | mriedem: np, just wanted to make sure you were aware | |
| 13:55:33 | mriedem | dansmith: oh yeah, remind me to talk to you about filtering on IPs | |
| 13:55:35 | mriedem | with your series | |
| 13:55:42 | mriedem | given the discussion we had in denver about how we don't do that in the DB | |