20:01:55 #startmeeting tripleo 20:01:56 Meeting started Mon Jul 8 20:01:55 2013 UTC. The chair is lifeless. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:01:57 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:01:59 The meeting name has been set to 'tripleo' 20:02:02 #topic agenda 20:02:06 morning lifeless :) 20:02:08 bugs 20:02:09 Grizzly test rack status 20:02:09 CI virtualized testing progress 20:02:14 open discussion 20:02:40 #topic bugs 20:02:42 #link https://bugs.launchpad.net/tripleo/ 20:03:25 #link https://bugs.launchpad.net/os-config-applier 20:03:31 #link https://bugs.launchpad.net/diskimage-builder/ 20:03:38 bug 1182249 and bug 1183223 are still kind of in limbo due to ordering things.. once I finish integrating orc and oac with occ I can address them 20:03:40 Launchpad bug 1182249 in tripleo "quantum configuration is overly hardcoded" [Critical,In progress] https://launchpad.net/bugs/1182249 20:03:41 Launchpad bug 1183223 in tripleo "nova-compute.yaml missing parameters" [Critical,In progress] https://launchpad.net/bugs/1183223 20:03:51 #link https://bugs.launchpad.net/os-refresh-config 20:04:02 cool 20:04:12 occ "works" now 20:04:15 bug 1182241 I have in hand 20:04:16 Launchpad bug 1182241 in tripleo "first-boot.d rules are running on every boot" [Critical,Triaged] https://launchpad.net/bugs/1182241 20:04:24 With my stack of patches landing 20:04:46 all the boot-stack first-boot rules are now orc rules 20:05:06 swweeeeeeetttt 20:05:18 except one fedora one which I've mentioned to lucasgnomes/derekh 20:05:26 Ng: https://review.openstack.org/#/c/35943/ doesn't look like you actually clicked on it 20:06:45 Ng: I may have misunderstood your IRC feedback 20:07:28 bug 1189385 and bug 1184484 seem stalled neutron side 20:07:30 Launchpad bug 1189385 in tripleo "quantum-server hung up it's listening port" [Critical,Triaged] https://launchpad.net/bugs/1189385 20:07:32 Launchpad bug 1184484 in tripleo "Quantum default settings will cause deadlocks due to overflow of sqlalchemy_pool" [Critical,Triaged] https://launchpad.net/bugs/1184484 20:07:38 #action lifeless follow up on those bugs 20:09:02 and thats all our criticals. 20:09:21 I haven't actually rearranged highs yet, OTOH I started closing some myself. 20:09:34 Perhaps we don't need to rearrange, just have folk start cherrypicking highs ? 20:10:37 +1 for cherry pick 20:10:46 a bug a week will keep us from getting burried 20:11:21 ok 20:11:56 anymore bugs discussion ? 20:12:12 do we need a bug tracker for orc and occ? 20:12:26 we have one for orc 20:12:32 didn't I link it? 20:12:40 lifeless: fixed. I thought I had clicked that one 20:12:42 I did : https://bugs.launchpad.net/os-refresh-config 20:12:43 ah yes later 20:12:49 lifeless: ok so occ will need one too 20:13:13 lets us the same lp teams as tripleo does 20:13:25 I don't think we need per-project granularity - or do we? 20:13:34 no definitely not 20:13:41 Ng: thanks! 20:13:56 1 team should handle tripleo-* and os-* for now 20:14:04 and dib :) 20:14:11 okies 20:14:13 yeah KISS :) 20:14:20 #topic Grizzly test rack status 20:14:25 Still working. 20:14:42 We had some reports of instance startup glitchiness, but haven't actually confirmed. 20:14:48 lifeless: I've not seen any more mails about dying quantum/etc bits - are they still throwing load at it? 20:14:55 SpamapS: you were going to file bugs for the things we need to do to make it long lived. 20:15:00 (or am I just terrible at email ;) 20:15:10 Ng: there are instances running, I too have not heard any more. 20:15:36 lifeless: indeed I was. action me? 20:15:43 #action SpamapS to do what SpamapS does 20:15:56 Any other POC rack business? 20:16:36 #topic CI virtualized testing progress 20:16:53 pleia2: oh hai 20:18:10 ok, I'm going to time out that ping :) 20:18:20 I haven't heard anything more from pleia2 - has anyone else? 20:18:52 no 20:18:59 guessing the US holiday interfered 20:19:02 ah 20:19:04 likely 20:19:12 It's kind of related 20:19:19 so I'll take a second to talk about devtest 20:19:31 devtest now covers all the way up to heat spawning an overcloud 20:19:47 of course, we don't have a scalable heat template for an overcloud 20:20:00 so its a bit of a lie 20:20:02 hm 20:20:21 I have been distracted elsewhere, but the main reason we didn't use nova-compute-group.yaml was the hostname issue, right? 20:20:22 SpamapS: as I recall we kindof do 20:20:27 SpamapS: yes I think so 20:20:34 ok, that is solved 20:20:45 - is used instead of ., and i believe there is a random component now too 20:20:46 SpamapS: I may need a little hand holding today, if you can 20:21:40 lifeless: sure, I'll be out of pocket for 2 hours after the meeting but then should be around to help 20:21:44 lifeless: no updates from me, holiday plus other things (needed to refresh my dev environment to flush out some network gremlins) 20:21:53 SpamapS: awesome 20:21:57 pleia2: ack! 20:22:05 toci is catching stuff too which is good 20:22:06 ok 20:22:09 #topic open discussion 20:22:23 toci is failing at the moment, probably needs some eyeballs on that 20:22:45 last I saw was some 404 on pypi 20:22:52 ah, good 20:22:59 (well, not good, but you know) 20:23:18 its the kind of not good we know how to handle I think 20:23:23 right 20:24:44 so 20:24:56 I guess we need to figure out how to get toci reliable enough to make it a gate 20:25:20 that is separate out 'infra failed' vs 'actually broken by $other' 20:25:20 some of the issues are just general di-b issues 20:25:50 like how to get already checked out repos into it and how to use a static mirror for certain things 20:26:36 yup 20:26:43 derekh has a stack of patches I'm going to review today 20:26:58 which will give us the basis for git caching 20:27:04 is there any reason we manually have people run all the steps in devtest.md versus just "run toci" ? 20:27:24 I think there are three concerns 20:27:36 one is 'get developers to understand the process' 20:27:45 there is 'automated in a specific context' 20:27:51 and there is 'deploy somewhere' 20:28:02 toci is a version of 'automated in...' 20:28:38 to me, devtest is the former; we should get considerably more automation in it, but not to the extent of reducing clarity. 20:29:07 this has come up before 20:29:16 and for deploy somewhere, that should be clearly broken up into what runs on the seed host, hardware enrollment etc. 20:29:17 has toci been kept up to date with what is specifically in devtest.md? 20:29:42 people want to start doing something with tripleo, but the steps in 19 seem to be a stumbling block 20:29:44 pleia2: largely yes, and one potential benefit of saying 'its all toci' would be to remove conceptual duplication 20:30:26 mordred: the recent report of failure looks like something toci would totally fail on too 20:30:54 probably so 20:30:57 my big concern about telling folk to run toci is that toci, like devstack, assumes it owns the local machine. 20:31:05 just being clear to what the desire is 20:31:16 ack 20:31:58 SpamapS: so for instance - the 'ssh in and copy this seed rc out' - thats something we should have a utility for IMO. 20:32:22 SpamapS: -or- we should populate the rc file from outside, for boot-stack init (Which I do for undercloudrc now) 20:32:45 yeah there are a few other steps that I don't think devs need help understanding 20:32:48 like all the git clones 20:33:00 yup 20:33:12 we should just have an mr // config-manager definition for that 20:33:22 alright so in general.. make steps simpler, but keep the progression obvious 20:33:37 yes 20:34:02 IMO treat devtest as a manual for onboarding new tripleo devs. 20:34:53 needs to be robust and reliable; needs to be safe to run on a laptop's normal environment. 20:35:19 needs to show -just enough- working folk get a handle on the overall process and where they can/should poke things. 20:35:25 what about new tripleo users who arent' necessarily devs 20:35:31 mordred: they are deployers 20:35:45 they might be _potential_ devs or deployers 20:35:53 try-before-you-buy kindda thing 20:36:15 or am I just too impatient 20:36:18 ? 20:36:45 The archetype in my head of a deployer 20:36:48 vs a dev 20:37:06 a dev is someone tasked with making tripleo better, in terms of *first contact*. 20:37:16 a deployer can obviously segue into doing dev. 20:37:57 a deployer usually wants to know enough about the plumbing to be confident driving it, but not to know all the gory details. 20:38:31 so - in my head there are a small number commands you run to just deploy. 20:38:42 you raise an interesting question of 'deploy to virt' 20:39:17 but frankly, we can't yet deploy a full overcloud from scratch without manual tweaking; let alone upgrade the software on it. 20:39:44 I think we need to close the loop and get the whole thing intact-but-improvable before we polish the entry 20:39:58 any deployer right now will be disappointed because not being able to upgrade is kindof a big deal 20:40:24 [or they will be doing upgrades via a CM-that-does-packages, and thus the whole design makes a lot less sense for them] 20:40:36 mordred: perhaps a better way to address this: 20:40:49 - we have enough people to either work on functionality, or polish. 20:40:54 - which should we pick :) 20:42:02 more polish makes testing easier and easier to get more folks involved 20:42:09 but I think it may be a harder sell 20:42:40 pleia2: I certainly agree; OTOH folk getting involved who don't know all the big picture structure can work on polish with good results 20:43:02 the bridge still has a hole in it 20:43:07 pleia2: as a way to get involved : thats no reason to have it unpolished, but it mitigates the 'can't get involved' aspect. 20:43:31 lets fill the hole, (updates) and then we can choose to improve the rest or add more. 20:43:58 SpamapS: yah. so here is what I think we should drive at 20:44:16 we should drive at end to end CI tests of deploy, and of upgrade-under-load. 20:44:29 and we should be implementing and automating precisely whats needed to deliver that 20:45:13 there is a tonne more to do once thats done, but at that point we no longer need to say 'and there is science fiction in this presentation' 20:47:41 so, mordred is distracted by a pep8 thing. 20:47:48 sorry 20:47:48 Any other business? 20:48:16 mission statement 20:48:18 mordred: np; just don't think we have enough here to make a useful decision other than 'we have work to do and we want to automate all the things' 20:48:24 mordred: thank you! I shall draft today 20:48:27 please do 20:48:33 #action lifeless to draft missing statement. 20:48:40 tripleo needs to apply for program status in a mail to the dev list 20:48:47 and it needs to happen 4 days before a tc meeting 20:48:54 for it to be an eligible vote item 20:49:21 that needs to happen before the infra team is really going to be available in mindspace to help think about ci integration 20:49:32 gotchya. 20:49:35 Ok, anything else? 20:50:01 Folk should put in paper submissions for HK. 20:50:21 Any other other things? 20:51:33 be well, Robert Collins 20:51:39 #endmeeting