Wednesday, 2017-08-16

*** edmondsw has quit IRC00:00
*** jwcroppe has joined #openstack-powervm00:06
*** jwcroppe has quit IRC00:12
*** thorst has joined #openstack-powervm00:42
*** jwcroppe has joined #openstack-powervm01:01
*** svenkat has quit IRC01:27
*** esberglu has quit IRC01:34
*** esberglu has joined #openstack-powervm01:35
*** esberglu has quit IRC01:39
*** esberglu has joined #openstack-powervm01:43
*** esberglu has quit IRC01:43
*** edmondsw has joined #openstack-powervm01:44
*** edmondsw has quit IRC01:48
*** jwcroppe has quit IRC01:55
*** thorst has quit IRC01:59
*** jwcroppe has joined #openstack-powervm03:15
*** edmondsw has joined #openstack-powervm03:32
*** edmondsw has quit IRC03:37
*** chhavi has joined #openstack-powervm04:58
*** chhavi has quit IRC05:00
*** edmondsw has joined #openstack-powervm05:20
*** edmondsw has quit IRC05:24
*** k0da has joined #openstack-powervm06:47
*** edmondsw has joined #openstack-powervm07:08
*** edmondsw has quit IRC07:13
*** k0da has quit IRC07:17
*** thorst has joined #openstack-powervm08:04
*** thorst has quit IRC08:09
*** jwcroppe has quit IRC08:55
*** edmondsw has joined #openstack-powervm08:57
*** edmondsw has quit IRC09:01
*** thorst has joined #openstack-powervm09:05
*** thorst has quit IRC09:09
*** jwcroppe has joined #openstack-powervm09:16
*** thorst has joined #openstack-powervm10:05
*** thorst has quit IRC10:11
*** edmondsw has joined #openstack-powervm10:45
*** edmondsw has quit IRC10:49
*** thorst has joined #openstack-powervm11:04
*** jwcroppe has quit IRC11:50
*** svenkat has joined #openstack-powervm11:53
*** svenkat_ has joined #openstack-powervm11:56
*** svenkat has quit IRC11:57
*** svenkat_ is now known as svenkat11:57
*** edmondsw has joined #openstack-powervm11:57
*** jwcroppe has joined #openstack-powervm13:00
*** esberglu has joined #openstack-powervm13:08
*** apearson has joined #openstack-powervm13:55
*** openstackgerrit has joined #openstack-powervm14:09
openstackgerritMerged openstack/networking-powervm master: Replace six.iteritems() with .items()
*** svenkat has quit IRC14:13
esbergluefried: Looks like uwsgi is just working now. At least stacking. My internet cut out and killed the tempest run so I didn't get a full run. Trying again now for master OOT and IT14:13
efriedesberglu Lovely.  I guess.14:13
efriedesberglu But it means us poor schlubs with regular neos and 2GB images are gonna fail until mtreinish gets his stuff in.14:14
efriedCause our /tmp dirs are small by default.14:14
efriedwell, too small for a 2GB image.14:14
efriedI should clarify: are gonna fail unless we mod_wsgify our local.confs - we won't be able to just copy the CI's.14:14
edmondswcould we carry his change as a patch?14:15
esbergluedmondsw: We don't need his change for CI14:15
edmondswoh, efried is just worried about his own system, got it14:16
efriedAnd yours too, edmondsw.  This workaround will have to go in the etherpad.14:16
edmondswefried right14:16
efriedesberglu Did we drop a change where we're copying logs out of /var/log/apache2 ?14:17
esbergluefried: Yeah14:18
efriedk.  I would say we wouldn't need that anymore after we switch to uwsgi... but there are still logs in there that we might care about.14:19
efriedSo as long as we're wildcarding them rather than copying them explicitly by name, we should leave it alone.14:19
esbergluefried: Yep we're just wildcarding /var/log/apache2/*14:19
*** svenkat has joined #openstack-powervm14:22
esbergluI wonder if it's possible to set up gerrit so that you don't need fresh votes if the only thing updated is the commit message14:31
*** dwayne has joined #openstack-powervm14:35
edmondswesberglu what are we waiting on before merging 5688 & 5689?14:37
esbergluRight now every pike run would fail because it will try to checkout pike ceilometer which doesn't exist yet14:38
edmondswcould we make it checkout master ceilometer?14:38
edmondswtemporarily, of course14:39
esbergluedmondsw: We could. I didn't think it was going to take this long for them to branch14:39
esbergluI might ask around again before making that change though14:39
edmondswI've tried on IRC with no luck... we need to send something on the mailing list about that14:39
edmondswyou want to write that?14:40
esbergluI don't want to put that change in if they are branching soon14:40
esbergluedmondsw: Sure14:40
edmondswthe answer I got from their PTL to my question about an outload was "I don't know" :(14:40
edmondswesberglu flag both [telemetry] and [release] in the subject when you hit the ML14:42
edmondswI want to keep this in front of the eyes of the release guys as much as the telemetry team14:43
efriedthorst edmondsw Got a sec to chat about PCI?14:43
thorstefried: I don't  :-(14:44
*** tjakobs has joined #openstack-powervm14:44
edmondswefried sure14:44
efriedIt may be a very quick question.14:44
efriedIs the thing we're trying to accomplish just related to passthrough of standard PCI devices like GPUs and FPGAs14:45
efriedor are we also trying to muck with SR-IOV PFs/VFs?14:45
thorstfor now, full devices.  Not semi-virtualized devices (those are almost a case by case thing)14:45
edmondswfor now14:46
thorstGPUs, FPGAs, IB, etc...14:46
thorst'weird stuff'14:47
edmondswefried, you're going to have to get over this yaml thing sooner or later :)14:51
efriededmondsw yaml is for other people.14:52
edmondswuh huh.. sure...14:52
efriedYou show me a creditable berimbolo and I'll learn yaml.14:52
esbergluefried: Lucky for you I'm getting rid of a ton of yamlf14:53
*** svenkat_ has joined #openstack-powervm15:22
*** svenkat has quit IRC15:23
*** svenkat_ is now known as svenkat15:23
*** svenkat has quit IRC15:31
efriedesberglu both CIs failed, and the links are 404.  Known issue?15:42
edmondswefried I looked out what berimbolo was... that's a lot more complicated than yaml :)15:42
efriededmondsw :)15:43
esbergluefried: Yeah the logserver ran out of space. I cleared a bunch of old logs out so we should be good now. Might have to tweak our logrotate config15:43
efriededmondsw If it was just yaml, that'd be one thing.  It's trying to understand the thing the yaml is talking to...15:43
efriedSo like, yaml itself is putting on the gi.  zuul is the berimbolo.15:44
efriedor whatever tf it's actually talking to.  jenkins, ansible, chef, puppet, other things I don't get.15:44
efriedesberglu So can I recheck?15:46
esbergluefried: Yep15:47
*** svenkat has joined #openstack-powervm16:16
*** svenkat_ has joined #openstack-powervm16:27
*** svenkat has quit IRC16:28
*** svenkat_ is now known as svenkat16:28
*** apearson has quit IRC17:00
*** apearson has joined #openstack-powervm17:03
*** apearson has quit IRC17:03
*** apearson has joined #openstack-powervm17:07
*** svenkat has quit IRC17:12
*** svenkat has joined #openstack-powervm17:24
*** svenkat_ has joined #openstack-powervm17:41
*** svenkat has quit IRC17:42
*** svenkat_ is now known as svenkat17:42
*** apearson has quit IRC17:52
*** jwcroppe has quit IRC17:54
*** k0da has joined #openstack-powervm17:54
*** apearson has joined #openstack-powervm18:01
esbergluefried: edmondsw: Got a response about ceilometer stable/pike. Apparently they can't cut a branch or release a rc1 for their release model? So we won't get one until the pike release18:18
efriedweird, okay.18:18
efriedI'm in no hurry.18:18
esbergluWe need to put in a workaround if we want to run CI on any patches going into pike between now and release18:19
esbergluI'm curious why they can't put out release candidates. Doesn't make sense to me18:21
edmondswhmm... they are missing
edmondswesberglu efried did we not have this trouble last release? Doesn't seem like something has probably changed with their release model in 6 months18:24
esbergluI don't recall facing this issue. Maybe we just didn't enable ocata CI until the official release? But I doubt that because I know we backported things to ocata after the first rc18:26
esbergluAnd would be surprised if we were merging stuff without CI runs18:26
edmondswhere's where the release model is defined:
edmondswdefined here:
efriedWe didn't have this problem because this is the first time in history we've remembered to update the tox.ini settings right as the release candidates were being branched.18:28
efriedEvery other time we missed it until something broke because of an incompat constraint in master.18:29
efriedand only then updated the settings.18:29
efriedThis is so not urgent.18:29
esbergluAnd would be surprised if we were merging stuff without CI runs18:30
edmondswefried but this is holding us up from being able to run CI on pike... not just delaying that tox.ini change18:32
efriedesberglu keep in mind, in the past (before our CI) it would only cause us a problem if a) we were trying to develop in a non-master branch (very rare, especially for ceilometer-powervm), and b) some master constraint change caused a breakage (also pretty rare).18:32
esbergluedmondsw: Well to get stuff into pike we first propose to master. So is the master CI run good enough before backporting?18:33
esbergluefried: This breaks the checkouts in prep_devstack though18:33
efriedRight, okay, so we can't run any CI on pike because it's trying to clone pike ceilometer?18:33
esbergluYeah. Easy workaround to put in18:33
efriedSo yeah, special case that one for now.18:33
efriedit's only for a few weeks.18:33
edmondswyep, I suggested that this morning. But I still want to push on the ceilometer guys in tandem, and see if we can avoid this next time18:34
edmondswI don't think they're right about their release model...18:34
edmondswreplying on the ML18:34
*** apearson_ has joined #openstack-powervm19:17
*** apearson has quit IRC19:19
efriedthorst edmondsw Got a successful PCI passthrough claim with no mods to nova base.19:54
edmondswhacky, or sensible?19:55
efriedAlmost completely sensible.19:55
edmondswcan you write up what you've done / figured out?19:56
efriedAssigning it to the partition will be some code, but that'll all be in our driver.19:56
*** jwcroppe has joined #openstack-powervm19:58
efriedthorst edmondsw I'm gonna need an actual system with more than one free PCI device.  Preferably multiple types, with multiple instances of at least one of those types.20:08
edmondswefried I'll see what I can do there20:10
edmondswthorst if you have tips, let me know... you've got far more experience scrounging up hardware :)20:10
efriededmondsw I think I can play around for a little while with what I've got.  Basically mucking with the claims engine without actually trying to assign the devices.  So I'll set up the driver to think it's got access to all the devices in the system (even the ones that are assigned).20:11
efriedBut when it comes to actually assigning 'em around, I'll need something more realistic.20:11
edmondswefried do you have the LP bug handy for the /tmp wsgi issue?20:33
openstackLaunchpad bug 1709970 in devstack "image create fills up /tmp and fails under uwsgi" [High,In progress] - Assigned to Matthew Treinish (treinish)20:33
edmondswty sir20:33
efriededmondsw mtreinish has two changes out for it.  If you punch through to the noted change set, it has a Depends-On.20:34
efriedThey seem to be working, btw.20:34
*** k0da has quit IRC20:35
edmondswefried added to
efriedThough I haven't tried them out in the original setup.20:35
efriededmondsw Cool, thanks.20:35
*** jwcroppe has quit IRC20:36
*** k0da has joined #openstack-powervm20:48
*** esberglu has quit IRC20:50
*** esberglu has joined #openstack-powervm20:50
*** esberglu has quit IRC20:55
*** esberglu has joined #openstack-powervm21:03
*** thorst has quit IRC21:19
*** edmondsw has quit IRC21:32
*** svenkat has quit IRC21:36
*** thorst has joined #openstack-powervm21:53
*** thorst has quit IRC21:53
*** k0da has quit IRC22:18
*** esberglu has quit IRC22:22
*** thorst has joined #openstack-powervm22:24
*** thorst has quit IRC22:29
*** apearson_ has quit IRC22:33
*** edmondsw has joined #openstack-powervm22:49
*** edmondsw has quit IRC22:54
*** esberglu has joined #openstack-powervm22:59
*** esberglu has quit IRC23:03
*** thorst has joined #openstack-powervm23:11
*** thorst has quit IRC23:13
*** tjakobs has quit IRC23:24
*** svenkat has joined #openstack-powervm23:59

Generated by 2.15.3 by Marius Gedminas - find it at!