openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Spec for cross-cell resize https://review.openstack.org/616037 | 00:29 |
---|---|---|
*** mriedem has quit IRC | 00:36 | |
*** irclogbot_2 has quit IRC | 00:39 | |
*** tetsuro has joined #openstack-placement | 00:42 | |
*** takashin has joined #openstack-placement | 01:22 | |
*** tetsuro has quit IRC | 01:33 | |
*** tetsuro has joined #openstack-placement | 06:14 | |
*** lyarwood has quit IRC | 06:53 | |
*** e0ne has joined #openstack-placement | 07:49 | |
*** helenafm has joined #openstack-placement | 08:19 | |
*** dims has quit IRC | 08:52 | |
*** dims has joined #openstack-placement | 08:53 | |
*** lyarwood has joined #openstack-placement | 08:55 | |
*** dims has quit IRC | 08:58 | |
*** dims has joined #openstack-placement | 08:59 | |
*** ttsiouts has joined #openstack-placement | 09:10 | |
*** takashin has left #openstack-placement | 09:30 | |
*** tetsuro has quit IRC | 09:33 | |
*** ttsiouts has quit IRC | 09:44 | |
*** ttsiouts has joined #openstack-placement | 09:44 | |
*** ttsiouts_ has joined #openstack-placement | 09:48 | |
*** ttsiouts has quit IRC | 09:49 | |
*** cdent has joined #openstack-placement | 09:58 | |
*** e0ne has quit IRC | 10:02 | |
*** e0ne has joined #openstack-placement | 10:04 | |
*** ttsiouts_ has quit IRC | 10:07 | |
*** ttsiouts has joined #openstack-placement | 10:07 | |
*** ttsiouts_ has joined #openstack-placement | 10:10 | |
*** ttsiouts has quit IRC | 10:12 | |
*** ttsiouts_ has quit IRC | 10:13 | |
*** ttsiouts has joined #openstack-placement | 10:20 | |
*** tssurya has joined #openstack-placement | 10:28 | |
*** e0ne has quit IRC | 10:45 | |
*** e0ne has joined #openstack-placement | 10:54 | |
*** ttsiouts has quit IRC | 11:36 | |
*** ttsiouts has joined #openstack-placement | 11:37 | |
*** cdent has quit IRC | 11:40 | |
*** ttsiouts has quit IRC | 11:42 | |
*** e0ne has quit IRC | 12:04 | |
*** ttsiouts has joined #openstack-placement | 12:08 | |
*** e0ne has joined #openstack-placement | 12:08 | |
*** e0ne has quit IRC | 12:09 | |
*** e0ne has joined #openstack-placement | 12:27 | |
*** cdent has joined #openstack-placement | 12:51 | |
*** cdent has quit IRC | 13:37 | |
*** ttsiouts has quit IRC | 13:39 | |
*** e0ne has quit IRC | 13:41 | |
openstackgerrit | Merged openstack/placement master: Add bandwidth related standard resource classes https://review.openstack.org/615687 | 13:42 |
*** e0ne has joined #openstack-placement | 13:45 | |
*** ttsiouts has joined #openstack-placement | 13:57 | |
*** mriedem has joined #openstack-placement | 13:58 | |
*** e0ne has quit IRC | 14:13 | |
*** e0ne has joined #openstack-placement | 14:14 | |
*** cdent has joined #openstack-placement | 14:36 | |
dansmith | mriedem: so the thing I'm concerned about with the grenade change is just that (if) all grenade jobs will be using new placement implicitly under the covers, | 15:03 |
dansmith | since we install it in the upgrade step instead of during the new phase devstack | 15:03 |
dansmith | I see that regular devstack jobs won't use new placement until the devstack change lands, | 15:04 |
dansmith | but we're kindof subverting the process it seems right? | 15:04 |
dansmith | or will running new devstack after the grenade upgrade just revert us back somehow? although I don't see that happening in the logs so I would think not | 15:04 |
mriedem | grenade will still run new devstack after the upgrade script which until https://review.openstack.org/#/c/600162/ lands should be overwriting whatever the upgrade script did | 15:04 |
dansmith | okay I did not see a third start of placement back to nova-placement in the logs, let me look again | 15:05 |
dansmith | yeah, it's not there | 15:05 |
dansmith | we run with placement-uwsgi for the rest of the run | 15:05 |
mriedem | looking | 15:07 |
dansmith | the problem with that I see is that all new side grenade jobs run with new placement it seems, but there's no way for me to get a new placement devstack without doing a full upgrade cycle I think | 15:07 |
dansmith | until the devstack patch lands | 15:07 |
mriedem | ok i guess you're right, this is first start of placement-api on the old side | 15:08 |
mriedem | http://logs.openstack.org/54/604454/21/check/neutron-grenade/a0f5d15/logs/screen-placement-api.txt.gz#_Oct_24_15_25_01_945201 | 15:08 |
mriedem | Oct 24 15:25:01.945201 ubuntu-xenial-inap-mtl01-0003381914 devstack@placement-api.service[21496]: uwsgi socket 0 bound to UNIX address /var/run/uwsgi/nova-placement-api.socket fd 4 | 15:08 |
mriedem | and then the new side | 15:08 |
mriedem | Oct 24 15:42:59.207451 ubuntu-xenial-inap-mtl01-0003381914 devstack@placement-api.service[17536]: uwsgi socket 0 bound to UNIX address /var/run/uwsgi/placement-api.socket fd 4 | 15:08 |
dansmith | I'm not really sure why devstack isn't configuring both to run, but I haven't looked | 15:08 |
dansmith | yeah | 15:08 |
mriedem | i'm glad i have https://review.openstack.org/#/c/600548/ so i can read this again | 15:10 |
dansmith | +W | 15:11 |
cdent | I guess I should read that, because I'm confused. What is "back to nova-placement"? | 15:11 |
mriedem | from-base upgrade scripts run after installing stein code but before starting it | 15:11 |
mriedem | so maybe that is the issue, | 15:11 |
mriedem | and maybe the grenade change should be a within-stein upgrade script | 15:12 |
mriedem | which runs before installing new code | 15:12 |
dansmith | how is the placement service even being started? | 15:12 |
mriedem | and probably makes more sense for how the extraction is happening | 15:12 |
dansmith | isn't it called something different? | 15:12 |
dansmith | or is it being started by systemd because of a dep? | 15:12 |
dansmith | and if that's the case, why aren't both being started? | 15:12 |
cdent | the systemd unit file is being changed | 15:13 |
cdent | run-process does that | 15:13 |
dansmith | the grenade change says a new unit file will be created | 15:13 |
dansmith | but also, isn't it new devstack that starts those services not grenade? so maybe that comment is wrong | 15:14 |
cdent | grenade does restart placement, but sounds you're thinking about a different restart than the one my comment was asserting? | 15:15 |
*** ttsiouts has quit IRC | 15:15 | |
*** ttsiouts has joined #openstack-placement | 15:16 | |
*** ttsiouts has quit IRC | 15:16 | |
*** ttsiouts has joined #openstack-placement | 15:16 | |
mriedem | grenade starts placement here https://github.com/openstack-dev/grenade/blob/master/projects/60_nova/upgrade.sh#L91 | 15:17 |
mriedem | the upgrade script runs here https://github.com/openstack-dev/grenade/blob/master/projects/60_nova/upgrade.sh#L68 | 15:17 |
dansmith | ah right I forgot, it has to start the services because it controls the order in some cases... derp. | 15:17 |
cdent | My understanding was that we were doing something "special" here because we cannot do "New code should work with old configs" | 15:17 |
cdent | thus the "theory of upgrade" doesn't fully apply | 15:18 |
mriedem | the from-rocky script runs after devstack installs placement on the new side | 15:18 |
mriedem | the within-rocky upgrade script, if we had one, would run here https://github.com/openstack-dev/grenade/blob/master/projects/60_nova/upgrade.sh#L61 | 15:18 |
mriedem | i can push a change on top that moves from-rocky to within-rocky but i'm not sure if that will work | 15:19 |
dansmith | cdent: yes we're definitely skirting the upgrade rules here, but my concern is just that all grenade jobs will now "secretly" run with new placement while there's no way to get a full devstack job or install running with new placement | 15:19 |
cdent | isn't running with new placement what we want, secret or not? | 15:19 |
dansmith | mriedem: can we not just gate this behavior on something set by devstack in cdent's patch? | 15:19 |
mriedem | you mean to noop if cdent's patch isn't merged, but also make his devstack change not fail grenade? | 15:20 |
mriedem | yeah maybe | 15:20 |
dansmith | just the way we always do these interactions.. where we gate something in grenade on devstack config | 15:20 |
dansmith | like we did for the cells transitions, etc | 15:21 |
mriedem | i.e. the NOVA_CONFIGURE_CELLSV2 flag | 15:22 |
dansmith | yeah | 15:22 |
dansmith | because we did the same thing, enabling grenade but not effecting that change until the devstack (and d-g, IIRC) things had landed | 15:22 |
mriedem | well, looking at cdent's devstack change, i could noop if $PLACEMENT_REPO is not defined | 15:22 |
mriedem | https://review.openstack.org/#/c/600162/12/stackrc | 15:22 |
dansmith | yup, that would be a good signal | 15:23 |
dansmith | because that's what's telling us to get it from somewhere else | 15:23 |
mriedem | or PLACEMENT_DIR https://review.openstack.org/#/c/600162/12/lib/placement | 15:23 |
mriedem | lemme see what old matty can do here | 15:23 |
mriedem | ok updated https://review.openstack.org/604454 | 15:34 |
cdent | mriedem: that seems sane | 15:38 |
cdent | and simple | 15:38 |
*** mriedem has quit IRC | 15:53 | |
*** mriedem has joined #openstack-placement | 16:26 | |
*** e0ne has quit IRC | 16:44 | |
*** e0ne has joined #openstack-placement | 16:44 | |
*** e0ne has quit IRC | 16:44 | |
*** efried is now known as efried_rollin | 17:08 | |
*** helenafm has quit IRC | 17:32 | |
*** mriedem has quit IRC | 17:33 | |
*** ttsiouts has quit IRC | 17:36 | |
*** ttsiouts has joined #openstack-placement | 17:39 | |
*** ttsiouts has quit IRC | 17:44 | |
cdent | dansmith, mriedem: looks like the grenade change behaves as expected: from placement on the devstack patch, not on the grenade patch for the second placement-api run | 18:19 |
dansmith | cdent: yep, +2 | 18:29 |
cdent | yay, progress | 18:29 |
*** e0ne has joined #openstack-placement | 18:36 | |
*** e0ne has quit IRC | 18:38 | |
*** mriedem has joined #openstack-placement | 18:38 | |
edleafe | cdent: it looks from the docstring that you're running the migration tests under postgres. Correct? | 18:45 |
edleafe | I'm getting a MySQL exception | 18:46 |
*** irclogbot_2 has joined #openstack-placement | 18:46 | |
cdent | that's the docstrong that you copied from wherever it was copied. I've run it against both mysql and postgresql | 18:46 |
cdent | what exception? | 18:46 |
edleafe | oslo_db.exception.DBError: (pymysql.err.InternalError) (1071, u'Specified key was too long; max key length is 767 bytes') [SQL: u'\nCREATE TABLE projects (\n\tcreated_at DATETIME, \n\tupdated_at DATETIME, \n\tid INTEGER NOT NULL AUTO_INCREMENT, \n\texternal_id VARCHAR(255) NOT NULL, \n\tPRIMARY KEY (id), \n\tCONSTRAINT uniq_projects0external_id UNIQUE (external_id)\n)\n\n'] (Background on | 18:46 |
edleafe | this error at: http://sqlalche.me/e/2j85) | 18:47 |
*** tssurya has quit IRC | 18:47 | |
edleafe | If I run the command with the MySQL cli I get the same thing | 18:47 |
cdent | that's a unicode related problem | 18:47 |
edleafe | yeah, the multibyte issue | 18:48 |
cdent | how did you create your db? cuz it works for me | 18:48 |
openstackgerrit | sean mooney proposed openstack/placement master: Harden placement init under wsgi https://review.openstack.org/612444 | 18:48 |
cdent | edleafe: you'll at some point need to rebase on ^ | 18:49 |
edleafe | I *thought* I had utf-8, but I dunno now. I'll re-create it and try again. | 18:49 |
edleafe | Wanted to run it by you first | 18:49 |
* edleafe curses heavy meeting days and their interruptions. | 18:50 | |
cdent | i'll try it here just to make sure it is still working | 18:50 |
cdent | (and try p27, which I don't usually do anymore) | 18:51 |
sean-k-mooney | cdent: edleafe the rebase may not be required as i did not chage the public interface in a syntaic way so there should be no merge confict later if you dont rebase | 18:51 |
sean-k-mooney | the semantics have change slightly however but they are more concreatly stated in comments now. | 18:53 |
cdent | edleafe: yeah still works here | 18:53 |
cdent | both dbs | 18:53 |
edleafe | Must have been a wonky database. It's working here now. | 18:54 |
cdent | I don't think WonkyDB will pass jepsen | 18:54 |
cdent | sean-k-mooney: will you be at summit? | 19:05 |
sean-k-mooney | cdent: no. if i really wanted to be or needed to be i have a room i could stay in for the week but i was not planning on gong this time | 19:06 |
sean-k-mooney | cdent: was there a topic you wanted to sync with me on? | 19:07 |
cdent | probably a wise choice. summits are weird. | 19:07 |
cdent | no mostly just wanted to be sure to wave and say hi if you were there | 19:07 |
sean-k-mooney | i got a little burnt out with travel in the last 10 monts. | 19:08 |
cdent | yeah | 19:08 |
sean-k-mooney | dublin ptg (stuck for 2 days) , then vancover summit, then internal intel face to face 1 week later, then join redhat then denver ptg(stuck in phlidefa for a night with canceled flight), then new heir orentation in munich | 19:11 |
sean-k-mooney | cdent: i actully kindo of feel rechared enough to go to the summit but that has only been in the last week or so but ill hopefully see you in denver | 19:12 |
*** e0ne has joined #openstack-placement | 19:24 | |
*** e0ne has quit IRC | 19:31 | |
*** mriedem is now known as mriedem_afk | 19:53 | |
*** efried_rollin is now known as efried | 20:18 | |
*** mnaser has quit IRC | 20:48 | |
*** mnaser has joined #openstack-placement | 20:49 | |
*** cdent has quit IRC | 21:19 | |
*** mriedem_afk is now known as mriedem | 22:22 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!