Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-03-09
10:45:37 owalsh asettle: AFAIK need to run api_db sync before map_cell0 or create_cell
10:46:39 asettle owalsh: the person to best answer that is bmoss as he has been testing. I'm just waiting for him to send me his error logs so can I appropriately address questions.
10:46:52 asettle I am currently just escalating to ensure we have eyes on it :)
10:57:04 asettle owalsh and johnthetubaguy - Brian confirmed that he initially used the dev docs in-tree to install, and those didn't work either *head desk*
10:57:39 gibi hi nova! there is a high severity bug fix that needs core attention https://review.openstack.org/#/c/437222/ . Could you check it please?
10:57:44 johnthetubaguy OK, that has worked for some people, at least apparently
10:58:29 johnthetubaguy gibi: its open in my browser, just chasing down this docs issue right now
10:58:36 gibi johnthetubaguy: thanks!
10:58:41 jaosorior gibi: thanks for the fix; that actually breaks us.
11:00:20 gibi jaosorior: hi! breaks me too :)
11:00:50 johnthetubaguy gibi: there are related quota issues to that fix, I believe
11:01:12 johnthetubaguy not that your fix breaks quotas, just that you are pointing at a wider bug / lack of code sharing
11:02:25 gibi johnthetubaguy: you mean this fix needs to be extended with some quota fix as well?
11:02:43 johnthetubaguy gibi: no, just there are other problems in that area I think
11:02:53 johnthetubaguy there is a quota bug about that, someone was working on it
11:03:19 johnthetubaguy gibi: I would love dansmith or melwitt to take a look at that bug fix, they have better context around that code
11:03:48 gibi johnthetubaguy: OK, I will ping them today
11:03:49 johnthetubaguy gibi: when I last checked it I was a bit worried about the cut and paste -ness of it, but not sure how to restructure it
11:04:43 gibi johnthetubaguy: yeah, that code path is ugly. I was affraid to touch it as I felt I will break something if I try to factor out the duplication
11:05:39 gibi johnthetubaguy: do you have the bug number for the quota bug? I would review that patch to learn something new
11:06:26 johnthetubaguy gibi: I don't was in chat previously, something about hitting over quota errors, and the quota never being removed
11:06:36 asettle johnthetubaguy: I see you pushed a patch! What's up?
11:06:40 asettle Just replied to the ML with brian's logs
11:06:44 cdent asettle, johnthetubaguy: do we have a more clear picture of the context that leads to "wrong" and whatever "wrong" is with that docs stuff? It feels a bit like we're all guessing
11:06:44 asettle bauzas: ^^
11:06:47 cdent oh good
11:06:54 cdent that'll probably answer my questions, thank you asettle
11:06:55 asettle cdent: yep take a look.
11:06:59 johnthetubaguy gibi: maybe just a function that does destroy and does the notification
11:07:26 johnthetubaguy asettle: I was worried about the order not being very clear, but also very, very important, so I refactored the docs patch while we were waiting for the logs
11:07:44 asettle cdent: the in-tree docs you guys have were also tested and although I don't have the error logs to give you, my understanding is that those instructions immediately error with "placement API is not available, you must configure the placement API."
11:07:49 asettle johnthetubaguy: you're a gem ;)
11:07:56 openstack Launchpad bug 1670627 in OpenStack Compute (nova) "quota is always in-use after delete the ERROR instances " [Critical,In progress] - Assigned to Matt Riedemann (mriedem)
11:07:56 gibi johnthetubaguy: this could be the quota bug https://bugs.launchpad.net/nova/+bug/1670627
11:08:00 sean-k-mooney tonyb: hi you about? sorry my rely to your comment on https://review.openstack.org/#/c/413325/2/tools/tox_install.sh is a bit ranty
11:08:34 johnthetubaguy asettle: did you include the dev list?
11:08:40 asettle johnthetubaguy: sure did
11:08:50 johnthetubaguy hmm, not getting the email, will try the archive
11:08:51 cdent asettle: I suspect that is because of this of the <Directory> thing (I added a link to an os-dev posting on the review)
11:09:04 gibi johnthetubaguy: yeah that could be done. If that helps then I'm happy to do that
11:09:06 sean-k-mooney tonyb: that said the fact my cygwin workflow for running tox is now broken really annoyed me when i disscovered it
11:09:22 johnthetubaguy gibi: makes it easier not to forget next time I think
11:09:24 cdent the reason things happen to work in devstack but not elsewhere is because keystone stuff is taking care of adding a <Directory> entry that covers placement too
11:09:27 gibi johnthetubaguy: OK
11:09:37 asettle johnthetubaguy: http://lists.openstack.org/pipermail/openstack-dev/2017-March/113608.html
11:10:17 asettle cdent: in the in-tree docs or the docs.o.o ones?
11:10:42 cdent asettle: I added a link on the docs.o.o review, just as a reference point
11:10:43 johnthetubaguy oh... maybe the docs don't talk about creating any flavors any more
11:11:05 johnthetubaguy thats quite a recent change, we no longer install any flavors by default, you have to create your own
11:11:08 gibi johnthetubaguy: I have to rewoke that it would be easy to do see the block change in https://review.openstack.org/#/c/437222/5/nova/compute/api.py
11:11:21 cdent asettle, johnthetubaguy: I need to coffee-up and read the logs, but I'll be around to try and help more soon
11:11:27 johnthetubaguy cdent: sweet
11:11:29 asettle cdent: thanks dude :)
11:11:38 johnthetubaguy cdent: it sounds very familiar
11:12:01 gibi johnthetubaguy: also it seems there are different error handling around instance.destroy
11:12:15 gibi johnthetubaguy: sometimes we retry after a refresh, sometimes we don'tt
11:12:28 gibi johnthetubaguy: here https://review.openstack.org/#/c/437222/5/nova/conductor/manager.py
11:13:14 johnthetubaguy asettle: I am not seeing that placement related error anywhere in the logs
11:13:20 johnthetubaguy asettle: do you know where you saw it?
11:14:07 johnthetubaguy I see "libvirtError: Failed to connect socket to '/var/run/libvirt/libvirt-sock': Connection refused"
11:14:15 asettle johnthetubaguy: that placement error was specifically for the in-tree docs, unreferenced
11:14:19 gibi johnthetubaguy: anyhow I will ping dansmith and melwitt later with this patch
11:14:25 asettle The error logs I sent you were from the instructions that bmoss tried in the patch
11:14:53 asettle johnthetubaguy: meant to be different things, sorry for confusing you
11:15:10 johnthetubaguy asettle: so the bug he is hitting seems to be listed here: https://github.com/mseknibilel/OpenStack-Grizzly-Install-Guide/issues/71
11:15:18 johnthetubaguy asettle: seems unrelated to placement or cells
11:16:43 openstackgerrit Lucian Petrut proposed openstack/nova master: Avoid using fdatasync() when fetching images https://review.openstack.org/443583
11:16:55 owalsh asettle: I think the current instructions should work, I've added some nits but nothing significant
11:17:15 asettle Okay, this is good. I'd like to get a few more eyes on it.
11:17:25 asettle One of my other concerns is that nobody seemed to agree in that patch set.
11:17:32 asettle There was a lot of differing opinions.
11:17:39 asettle I might have to spin something up and test it, Brian has just gone to bed.
11:18:22 asettle I'm sorry I can't answer all the questions - I'm speaking on behalf of someone.
11:18:41 cdent asettle: you're encountering that sad reality that seems to be coming up alot lately: Half of us making this stuff haven't got enough time to actually use it and test it :(
11:19:59 johnthetubaguy asettle: afraid the logs don't help here, so someone is going to have to walk through the guide I guess?
11:20:31 openstackgerrit Lucian Petrut proposed openstack/nova master: Avoid using fdatasync() when fetching images https://review.openstack.org/443583
11:22:16 asettle johnthetubaguy: volunteering? :D
11:22:23 johnthetubaguy asettle: I am curious what your take is on the doc change after this: https://review.openstack.org/#/c/438328/7/doc/install-guide/source/nova-controller-install.rst
11:22:23 asettle You can do it and I can do it and i'll meet you in the middle
11:22:55 asettle cdent: unfortunately, yep. I've been chatting to mriedem_away about getting a dedicated CPL for this stuff. Nova is one of the only projects that don't have one.
11:23:12 asettle johnthetubaguy: where specifically?
11:23:16 johnthetubaguy asettle: I don't think I can, I am blocking way too many people to do none of my other work I had planned, bauzas sounded like he might
11:23:35 johnthetubaguy asettle: I basically moved all the new instructions into the existing flow
11:24:05 asettle johnthetubaguy: okay that's a good start. My day is relatively clear :)
11:24:11 johnthetubaguy asettle: this is the bulk of the change https://review.openstack.org/#/c/438328/4..7/doc/install-guide/source/nova-controller-install.rst
11:25:00 asettle johnthetubaguy: okay, lemme do the thing.
11:26:20 asettle Let's see if I can get lucky.
11:45:35 alex_xu bauzas: so currently our API is free dict for hints field?
11:46:18 alex_xu bauzas: oops, I saw you said i have valid comment
12:22:52 arne_r markus_z: I addressed the issues in https://review.openstack.org/#/c/436868/ and Jenkins verified
12:24:03 markus_z arne_r: The zkvm CI should also pass
12:25:06 arne_r mriedem_away lyearwood: Is this good enough for a workflow now? https://review.openstack.org/#/c/436868/
12:50:12 alex_xu bauzas: johnthetubaguy, looks like we have no quota for scheduler_hints....so is that something user can input any number of hints?
12:53:13 alex_xu but I guess that limited by the max length of request body
12:56:38 mriedem asettle: we had docs CPL volunteers in the mailing list
12:56:53 asettle mriedem: yes I was waiting to hear back from you regarding confirmation of that.
12:57:06 asettle I saw there were two volunteers - let me know who is available to provide the time
12:57:14 mriedem asettle: Kevin_Zheng and nicolas bock
12:57:23 asettle mriedem: terrific: https://wiki.openstack.org/wiki/CrossProjectLiaisons#Documentation
12:57:38 asettle It would be very much appreciated if you could update that table :)

Earlier   Later