| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-04-17 | |||
| 19:32:14 | jaypipes | mriedem: yes, please go ahead on that service catalog spec. | |
| 19:37:31 | lifeless | efried: so 'mock' is the backport mechanism, if there is a fix in git, we'll pick it up and make it available | |
| 19:38:02 | efried | lifeless Okay, cool. I'll keep an eye on those two bug reports. | |
| 19:38:04 | efried | Thanks | |
| 19:39:43 | mordred | jaypipes: it's not as good as my summit talk that didn't get accepted | |
| 19:40:11 | jaypipes | lol | |
| 19:51:09 | mriedem | alex_xu: this must be yingxin's perf benchmark tool for the nova scheduler https://github.com/cyx1231st/nova-scheduler-bench | |
| 19:53:57 | mriedem | nice, readme says you can use devstack single node | |
| 20:00:58 | comstud | mriedem: quotas are not my fault. | |
| 20:01:10 | comstud | :) | |
| 20:01:16 | mriedem | comstud: is that something i said 3 weeks ago? | |
| 20:01:24 | comstud | at least | |
| 20:01:27 | mriedem | ha | |
| 20:01:38 | mriedem | comstud: but your name is all over the TODOs :) | |
| 20:01:53 | comstud | My name was over a race condition TODO re: instance delete | |
| 20:02:04 | mriedem | guilt by association | |
| 20:02:08 | comstud | but that TODO existed prior to quotas | |
| 20:02:12 | comstud | :) | |
| 20:02:24 | mriedem | comstud: riddle me this: who do i blame for instance delete? | |
| 20:02:38 | comstud | well | |
| 20:02:41 | comstud | a lot of people | |
| 20:02:49 | comstud | I did put effort into trying to resolve races | |
| 20:03:13 | jroll | whoa, comstud is here | |
| 20:03:15 | comstud | that particular TODO you were asking about... was definitely to cover the case of 2 concurrent API requests for deleting the same instance | |
| 20:03:20 | comstud | jroll: :) | |
| 20:03:31 | mriedem | comstud: you should realize by now that anything i mention and call you out is just an attempt to lure you back | |
| 20:03:43 | comstud | The quotas code in general needed a lot of love when I stopped openstacking. | |
| 20:03:49 | comstud | and I'm guessing it has never been improvemd | |
| 20:03:51 | comstud | improved | |
| 20:03:54 | mriedem | comstud: wip | |
| 20:04:03 | comstud | heheh | |
| 20:04:07 | mriedem | wip == the status melwitt gives in each week's cells meeting | |
| 20:04:21 | comstud | you asked me to smack you around... I'd gladly do it | |
| 20:04:26 | mriedem | comstud: but https://specs.openstack.org/openstack/nova-specs/specs/pike/approved/cells-count-resources-to-check-quota-in-api.html | |
| 20:04:29 | comstud | although there was no other context at the time | |
| 20:04:32 | comstud | hehe | |
| 20:05:33 | comstud | IIRC, quotas were very fragile even without cells | |
| 20:05:52 | comstud | the soft deleting of instances in particular caused issues | |
| 20:06:23 | mriedem | yes, soft delete is terrible | |
| 20:06:24 | comstud | i know I'd made a number of comments on how to fix it | |
| 20:06:27 | comstud | but | |
| 20:06:35 | comstud | i don't remember exactly what they were | |
| 20:06:35 | mriedem | well, | |
| 20:06:38 | comstud | i think it was... | |
| 20:06:44 | mriedem | soft delete + quotas + local delete code in the api | |
| 20:06:47 | comstud | move some code very close to the real delete. | |
| 20:07:01 | comstud | but that's the other issue... | |
| 20:07:10 | comstud | edge cases when compute host is down | |
| 20:07:13 | comstud | etc | |
| 20:07:29 | mriedem | yeah see https://review.openstack.org/#/c/386756/ | |
| 20:07:40 | mriedem | we've also duplicated a lot of that local delete logic too | |
| 20:07:47 | mriedem | since you've been gone | |
| 20:08:10 | mriedem | oh, let's not forget shelve offloaded | |
| 20:08:33 | comstud | ya, i dunno when shelved was added | |
| 20:08:41 | comstud | i think it was after I stopped focusing on nova | |
| 20:08:46 | mriedem | really? | |
| 20:08:48 | mriedem | i figured it was older | |
| 20:08:53 | mriedem | what i refer to as "the before times" | |
| 20:08:59 | comstud | haha | |
| 20:09:15 | comstud | well, it's new enough that I don't remember much about it | |
| 20:09:19 | mriedem | nova, BC (before comstud) | |
| 20:09:20 | comstud | it is familiar to me, so I know it must have existed | |
| 20:09:35 | comstud | but I might have been more focused on Ironic at the time | |
| 20:09:40 | mriedem | ok, | |
| 20:09:43 | mriedem | well, fyi, don't use it | |
| 20:10:03 | comstud | haha | |
| 20:10:08 | comstud | don't use openstack? | |
| 20:10:09 | comstud | yeah, I know. | |
| 20:10:23 | mriedem | ouch | |
| 20:11:29 | comstud | in any case, i remember feeling like the quota code needed a complete re-do | |
| 20:11:44 | mriedem | that's what melwitt is working on | |
| 20:12:01 | mriedem | taking out reservations and commit/rollback | |
| 20:12:34 | mriedem | if i mention her enough, she's bound to show up too :) | |
| 20:13:48 | comstud | hehe | |
| 20:16:48 | Apoorva | Cinder CI tests are failing with following error http://paste.openstack.org/show/606920/ Can someone help? | |
| 20:17:10 | Apoorva | Failures are in n-cpu.log http://openstack-ci.tintri.com/tintri/refs-changes-58-430858-3/logs/screen-n-cpu.log.txt | |
| 20:17:26 | mriedem | Apoorva: is your CI using the ubuntu cloud archive? | |
| 20:17:49 | mriedem | this came up in the ML, something about the version of libvirt and python-libvirt not matching | |
| 20:17:50 | Apoorva | mriedem I recently upgraded to Ubuntu 16.04 | |
| 20:19:15 | mriedem | Apoorva: see http://lists.openstack.org/pipermail/openstack-dev/2017-April/114912.html - i think somewhere in that thread someone mentions this issue with python-libvirt | |
| 20:20:24 | rfolco | Apoorva, we solved this by rebuilding python-libvirt as mriedem said. Basically our pip wheels were out of sync. | |
| 20:21:25 | mriedem | gah the CellDatabases fixture isn't in newton | |
| 20:40:16 | Apoorva | rfolco thanks. I have libvirt-python (3.2.0) installed on the machine running devstack | |
| 20:55:47 | mriedem | jaypipes: commented on the claims in scheduler spec too https://review.openstack.org/#/c/437424/ | |
| 21:00:04 | openstackgerrit | Monty Taylor proposed openstack/python-novaclient master: Add novaclient client_name and client_version to user-agent https://review.openstack.org/456638 | |
| 21:25:17 | openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Use service catalog to get endpoint URLs https://review.openstack.org/454983 | |
| 21:25:19 | mriedem | jaypipes: updated ^ | |
| 21:25:32 | jaypipes | mriedem: cool, thank you sir. | |
| 21:34:25 | jaypipes | mriedem: k, +2 from me. | |
| 21:34:50 | jaypipes | mriedem: that one would be great starter work for a newcomer to Nova. | |
| 21:35:07 | jaypipes | mriedem: do you mind if I reach out on the ML to find someone for that one? | |
| 21:35:20 | jaypipes | mriedem: not sure if you'd already talked to anyone. | |
| 21:35:34 | efried | mriedem jaypipes I threw a comment in there | |
| 21:36:00 | efried | And am suspiciously present as you're discussing giving an assignment to a newcomer... | |
| 21:36:22 | jaypipes | efried: you consider yourself a newcomer? :) | |
| 21:36:51 | efried | Been sequestered in my powervm cubby for two or three years. Know practically nothing about nova at large. | |
| 21:36:59 | jaypipes | efried: you've been around IRC for at least a month now. That makes you an old Nova hat :P | |
| 21:37:16 | efried | Oh, okay. One month is long time, got it. | |
| 21:37:21 | jaypipes | efried: if you're interested in that one, and mriedem says cool, go for it ;) | |
| 21:38:21 | efried | If it would be more appropriate for an even noobier noob, I won't cry in my beer. | |