Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-03-18
15:21:27 kashyap (And fixed the wrong info in the Nova conf doc, which came in via another misinfo from IBMs's doc, which disappeared)
15:23:37 kashyap Huh, let me reorder them patches correctly. Sorry for the noise.
15:25:29 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: libvirt: Use 'writeback' QEMU cache mode when 'none' is not viable https://review.openstack.org/641981
15:25:29 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: libvirt: vzstorage: Use 'writeback' QEMU cache mode https://review.openstack.org/643376
15:25:30 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: libvirt: smbfs: Use 'writeback' QEMU cache mode https://review.openstack.org/643377
15:27:26 gibi melwitt, mriedem: the bandwidth documentation patch for neutron is ready to review https://review.openstack.org/#/c/640390
15:30:04 mriedem gibi: ok, should probably get mlavalle to review that since he worked on it
15:32:12 gibi mriedem: sure
15:32:26 gibi I just wanted to note that it is not wip any more
15:54:18 tssurya thanks gmann
16:16:52 kashyap efried: Just to tie-up the loose end: your observation was correct earlier -- self._disk_cachemode does _not_ get fed in by the CONF.libvirt.disk_cachemodes
16:20:24 kashyap mdbooth: Thanks for the review, again. It's like a whack-a-mole. Just when I think I fixed all occurrences...
16:21:09 mriedem aspiers: efried: i'm doing some testing in devstack with the compute capabilites as traits stuff and i'm not seeing where / when a sighup is needed https://review.openstack.org/#/c/538498/22/releasenotes/notes/driver-capabilities-to-traits-152eb851cd016f4d.yaml
16:21:46 mdbooth kashyap: Hehe.
16:22:14 mdbooth kashyap: Incidentally, if at all possible can you avoid rebasing?
16:23:09 kashyap mdbooth: Sure. I was fixing commit messages by re-ordering patches, and forgot to put them back in the same order one of the times
16:23:28 mdbooth Ack
16:23:29 kashyap (Once I needed a genuine rebase due to that "compress" flag landing)
16:24:19 tssurya mriedem: whenever you have time and if you remember, do you know why we didn't go ahead with https://review.openstack.org/#/c/218975 ? I can't find it in the history in case there was any opposition
16:24:47 mdbooth kashyap: Did you see your release note build failure, btw?
16:25:21 kashyap mdbooth: I didn't, will look in the AM. Have to bike to my Dutch class shortly, and stay alert for 3 1/2 hours...
16:25:32 mdbooth kashyap: That's on the bottom patch, so you're going to have to beg for your +2s back.
16:25:39 mriedem tssurya: go ahead with that patch or the alternative discussed between myself and alaski?
16:25:57 tssurya go ahead with that patch I mean
16:26:03 openstackgerrit Balazs Gibizer proposed openstack/nova master: Refactor _heal_allocations_for_instance to make place for port healing https://review.openstack.org/637953
16:26:03 tssurya to add the config
16:26:03 openstackgerrit Balazs Gibizer proposed openstack/nova master: Prepare _heal_allocations_for_instance for nested allocations https://review.openstack.org/637954
16:26:04 openstackgerrit Balazs Gibizer proposed openstack/nova master: nova-manage: heal port allocations https://review.openstack.org/637955
16:26:04 openstackgerrit Balazs Gibizer proposed openstack/nova master: cache neutron ports in heal allocation https://review.openstack.org/638207
16:26:07 mriedem see the comments between myself and laski :)
16:26:08 kashyap Thanks for the heads-up. mdbooth: And yes, I'll add a rel note.
16:28:01 tssurya mriedem: shoot my bad sorry :)
16:29:19 mriedem tssurya: are you hitting an issue with baremetal nodes?
16:29:46 tssurya mriedem: the power sync stuff https://review.openstack.org/#/c/636132/
16:30:04 tssurya working on it since we need it downstream
16:30:10 tssurya so was going through the history
16:30:24 mriedem ok you're aware of https://github.com/openstack/nova/commit/19cb8280232fd3b0ba0000a475d061ea9fb10e1a right?
16:30:40 tssurya yea
16:44:48 cfriesen process question...when does the stein branch get created?
16:46:38 edleafe cfriesen: usually when RC1 is cut
16:53:26 mriedem cfriesen: iow thursday
16:54:04 cfriesen thx, was just looking at the release timeline.
16:54:44 efried mriedem: Set your association refresh interval to 0 and you'll need that SIGHUP
16:55:03 efried mriedem: I reckon we're deprecating that option soon
16:55:34 openstackgerrit Balazs Gibizer proposed openstack/nova master: Remove put_allocations() from report client https://review.openstack.org/644292
16:55:34 efried I think we maybe wanted to give CERN a little while to play first
16:56:35 mdbooth Eurgh: http://logs.openstack.org/52/626952/5/check/openstack-tox-py37/47f5826/job-output.txt.gz
16:56:57 mdbooth eventlet monkey patching for command line tools which import nova?
16:57:08 openstackgerrit Matt Riedemann proposed openstack/nova master: Add docs for compute capabilities as traits https://review.openstack.org/644293
16:57:12 mriedem efried: oh.....well i guess i need to update this then ^ but that wasn't mentioned in the reno
16:57:44 mdbooth It really would be better to monkey patch at specific entry points :/
16:59:12 efried mriedem: hm, the default there is five minutes, was there any science behind "60 seconds" or was that anecdotal?
17:01:07 mriedem anecdotal
17:01:11 mriedem i'm trying again now
17:01:13 mriedem with resource_provider_association_refresh=0
17:03:09 mdbooth What's api-paste.ini used for, and what are the implications of making changes there?
17:04:58 cdent mdbooth: it configures middleware used by the api
17:05:21 cdent the difficulty with make changes there is that it is in that hybrid realm of config file and also stuff that's required
17:05:38 mdbooth cdent: So changes to the one we have in tree would have deployment implications?
17:05:44 cdent yes
17:05:48 mdbooth Eurgh
17:05:53 cdent However, we have changed it in the past
17:05:56 cdent so it's not impossible
17:06:23 mdbooth I basically wanted to airlift it into a module which isn't called __init__.py
17:06:24 cdent there's been talk of trying to get rid of it for a long time (similar to policy-defaults-in-code)
17:06:37 mriedem rpms should mark that file as config so updates shouldn't overwrite it
17:06:49 mdbooth However, that would change a bunch of object paths, and therefore require changes in api-paste.ini
17:07:18 mriedem api-paste.ini at this point is mostly just a relic
17:07:20 cdent mdbooth: yeah, this is one of the places I fell down when doing my own roam through __init__ files
17:07:29 mdbooth Or... I could just move monkey patching to the top of it and leave it there...
17:07:31 mriedem when nova and openstack in general were more open to extension points
17:09:15 mdbooth My latest version of the monkey patching change moves monkey patching into nova/__init__.py, and also adds a couple of assertions which will explode if you've imported libraries with known problems before monkey patching
17:09:27 mdbooth Which sounded great, until I realised that command line tools also import nova
17:09:41 mdbooth testing, ftw
17:10:15 mdbooth So now I'm looking again at monkey patching only for certain entry points
17:10:38 mdbooth But we can't monkey patch lower than nova/api/openstack/... because __init__.py imports the world
17:10:59 cdent ayup
17:11:36 mdbooth I was hoping to move that out of the way, but that looks depressingly involved
17:21:42 mriedem efried: hmm, what am i missing with trying to hup n-cpu in devstack: sudo systemctl kill -s HUP devstack@n-cpu.service
17:21:46 mriedem doesn't seem to be doing anything
17:22:39 efried mriedem: have you tried just finding the pid?
17:26:27 mriedem hmm, i must have killed it somehow
17:26:27 mriedem Active: inactive (dead) since Mon 2019-03-18 17:09:49 UTC; 16min ago
17:31:15 mriedem got it
17:31:15 mriedem INFO nova.compute.manager [None req-65477139-47bd-4419-b5b3-d729fdadb369 None None] Reloading compute RPC API
17:32:15 efried mriedem: how did you do it?
17:32:26 mriedem kill -HUP <pid>
17:34:55 efried mriedem: based on the systemctl man page, it's possible you would have needed to say -s SIGHUP (not just HUP) and specify a --kill-who=main (or something?) to target the actual n-cpu process rather than its control process.
17:35:23 efried so yeah, kill -HUP <pid> sounds easier to me :P
17:35:25 openstackgerrit Matt Riedemann proposed openstack/nova master: Add docs for compute capabilities as traits https://review.openstack.org/644293
17:40:08 openstackgerrit Matthew Booth proposed openstack/nova master: Fix SSL infinite recursion https://review.openstack.org/626952
19:06:46 melwitt mriedem: new bug I tagged for stein-rc-potential https://bugs.launchpad.net/nova/+bug/1820588
19:06:47 openstack Launchpad bug 1820588 in OpenStack Compute (nova) "Bandwidth resource is leaked if a bound port is deleted in neutron" [Low,Confirmed]
19:22:26 aspiers mriedem, kashyap: Sorry, had to take a few days break due to RSI issues. (Cautiously) back now and trying to catch up
19:36:30 openstackgerrit Eric Fried proposed openstack/nova master: WIP/PoC: Use openstacksdk for placement https://review.openstack.org/643664
19:36:31 openstackgerrit Eric Fried proposed openstack/nova master: WIP/PoC: Use SDK instead of ironicclient for node.get https://review.openstack.org/642899
19:53:48 openstackgerrit Eric Fried proposed openstack/nova master: qemu: Make disk image conversion dramatically faster https://review.openstack.org/640781
19:54:47 efried kashyap: I added a colon to make that go ^
19:54:54 efried you may (or may not) wish to rebase its successors.
20:12:37 efried mdbooth: I've only superficially been following along, but when you rip out eventlet monkey-patching, some deepcopyZ will start to fail on lock artifacts.

Earlier   Later