19:03:12 #startmeeting infra 19:03:13 Meeting started Tue Dec 20 19:03:12 2016 UTC and is due to finish in 60 minutes. The chair is fungi. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:03:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:03:16 The meeting name has been set to 'infra' 19:03:20 #link https://wiki.openstack.org/wiki/Meetings/InfraTeamMeeting#Agenda_for_next_meeting 19:03:22 #topic Announcements 19:03:30 #info The OpenStack Infra team will not meet on Tuesday, December 27, 2016; our next meeting will occur on Tuesday, January 3, 2017 at 19:00 UTC in the #openstack-meeting channel as usual. 19:03:40 (as decided in last week's meeting) 19:03:45 as always, feel free to hit me up with announcements you want included in future meetings 19:04:03 #topic Actions from last meeting 19:04:10 #link http://eavesdrop.openstack.org/meetings/infra/2016/infra.2016-12-13-19.03.html previous meeting minutes 19:04:18 pabelanger add openstackci::zuul_launcher puppet class 19:04:22 Yes! 19:04:24 how did you get on? 19:04:30 I have started working on this locally 19:04:33 o/ 19:04:38 just doing a few tests before pushing it up 19:04:43 but, looks to be straightforward 19:04:59 will be ready for reviews on thursday 19:05:02 okay, cool. maybe we can mark it off by the time we meet again in a couple weeks 19:05:14 ++ 19:05:34 i likely won't be doing much in the way of reviews on thursday, but give me a heads up and i'll try to look when i'm online 19:05:44 sure 19:05:52 thanks for working on that! 19:05:57 #topic Specs approval 19:06:06 #info APPROVED "Zuul v3: update with Ansible role information" 19:06:12 #link https://review.openstack.org/381329 "Zuul v3: update with Ansible role information" spec update 19:06:18 also i finally marked the artifact signing spec as implemented 19:06:21 o/ 19:06:23 #link https://review.openstack.org/409906 "Artifact signing is now implemented" change 19:07:08 #link http://lists.openstack.org/pipermail/openstack-dev/2016-December/109111.html "Signing Keys for Ocata Release" announcement 19:07:39 doesn't look like we have any new specs going up for a vote this week 19:07:59 #topic Priority Efforts 19:08:28 nothing is called out on the agenda, but our next discussion topic covers something we may want to consider as an addition 19:08:40 #topic Cross-project goal to support Python 3.5 in Pike (fungi) 19:08:47 #link https://review.openstack.org/349069 proposed cross-project goal to support Python 3.5 in Pike 19:09:08 the latest patchset for that proposed pike goal mentions infra 19:09:15 we have quite a few repos without python35 jobs 19:09:24 * AJaeger hasn't done any detailed analysis 19:09:41 we do have rather a lot of python-based tooling, at least some of which is probably not tested at all for py3k compliance, and certainly most of it is not regularly running all its tests under 3.5 19:10:29 i haven't responded positively to the goal wearing my infra ptl hat yet, since i wanted to bring it up with the team first and figure out if we think it's feasible by the end of pike 19:10:59 fungi: i don't think we belong in there 19:11:30 i can certainly make that case, given sufficient explanation 19:11:59 i imagine that this sort of goal is to make openstack itself more operationally consistent 19:12:01 exemption along the same lines as the licensing argument? 19:12:21 fungi: indeed, and even language -- i mean, our java code isn't going to be py3 any time soon 19:13:12 well, sure, this is primarily about making sure we don't get caught without a solution when python 2.7 reaches end of life 19:13:34 i'm not disputing its importance, merely the need to synchronize with openstack itself on tihs 19:15:16 its probably not a bad idea to start thinking about it though? 19:15:28 oh, we've been thinking about it, and indeed doing a lot of work 19:15:49 yeah, so for infra projects we generally don't use coordinated global requirement synchronization, have projects under other (even apache-incompatible) free licenses, have projects in languages not officially sanctioned by the tc for openstack deliverables 19:16:00 but we've also been deferring some of it because we know that the initial cut of zuulv3 is more urgent 19:16:00 Might also want to update the creator's guide templates to include or require 3.5 http://docs.openstack.org/infra/manual/creators.html#configure-zuul-to-run-jobs 19:16:48 one thing we could do is require any new code to be python3 19:16:55 no we can't 19:16:58 like possible new bug organizing report dashboard 19:17:01 jeblair: why not? 19:17:02 oh like that 19:17:04 DavidPurcell: yep, agreed, in fact we've already started including gate-{name}-python35 in the python-jobs job-template in jjb and have talked about adding it to the python-jobs project-template in zuul's config 19:17:06 DavidPurcell: I just updated that page, wonder why it's not up yet... 19:17:07 i thought you meant new patches 19:17:16 sorry not new patches, new project repos/code bases 19:17:41 clarkb: possibly. i mean, we should at least try. :) 19:18:13 so i guess there's a couple of major axes to this problem from our (infra's) perspective 19:18:18 * AJaeger investigates why infra-manual publishing fails 19:18:21 we could also likely port our jobs that check things about our repos to python3 easily as that should be self testing maybe 19:18:22 AJaeger: It is now... That was quick. 19:18:36 first is whether _our_ tools need to all support python 3 asap, and it sounds like that's maybe a harder sell 19:18:38 DavidPurcell: ;) 19:19:03 the other is whether our ci supports and workflow documentation encourages python 3 support in other projects 19:19:11 which i think we've already mostly met? 19:19:36 yes, I don't think anything prevents our users from running python3 19:19:41 we should double check the documents 19:20:35 certainly implying that it's acceptable for completely new python projects to start out without python 3 support is something we should avoid 19:22:18 fungi: there's a reference to an un-logged #openstack-python3 channel in there 19:22:31 a rough count says we have somewhere around 40 infra repos with a setup.py in them, and even auditing those for py3k compliance before the end of the pike cycle strikes me as a tall order 19:22:46 * AJaeger just did https://review.openstack.org/413211 to promote python35-jobs a bit 19:22:48 jeblair: oh, fun 19:23:03 fungi: which violates the tc irc channel policy: https://governance.openstack.org/tc/reference/irc.html 19:23:56 okay, that took me way to long to find Infrastructure in that review 19:24:10 before we enforce it, we need to extend meetbot setup to handle all channels... 19:24:32 Agreed, that is going to be a fair amount of work for us I think. 19:24:39 with zuulv3 effort 19:24:51 jeblair: agreed, i've asked dhellmann and haypo in that channel whether there's a change to add logging for it 19:25:10 dhellmann thinks it's an oversight we should correct 19:25:17 right I think a good place to start is where it is easy 19:25:32 require new code bases be python3, update simple scripts that are already self testing, and so on 19:25:35 there are some significant py3 challenges with gear that need to be resolved before systems that use it can/should switch to py3 19:25:56 this is one of the things we have deferred so that we can make progress on the basics in zuulv3 19:26:36 i think they are solvable, but they are going to take some real care and thought, and should not be rushed 19:26:37 at least all infra repos that have python35 set up , have it as voting ;) 19:26:42 jeblair: yeah, i think the fact that we don't use global requirements and the py3k goal hinges on making sure the global requirements list of deps support it is a good indication of the situation we're in there 19:28:26 fungi: yeah, and i like the idea of the goals, and i like this one. i don't want us to be a distraction for it. 19:29:22 #agreed The Infra team is actively working toward Python 3 support across its tooling, but has far more diverse requirements from the general body of OpenStack services and libraries and so does not expect to complete that effort by the end of the Pike development cycle; however making sure our CI automation supports the Python 3 testing needs of other projects is a very high priority (and already 19:29:23 basically solved). 19:29:28 hrm, that was too long 19:29:30 #undo 19:29:31 Removing item from minutes: 19:29:52 #agreed The Infra team is actively working toward Python 3 support across its tooling, but has far more diverse requirements from the general body of OpenStack services and libraries and so does not expect to complete that effort by the end of the Pike development cycle. 19:29:53 * jlvillal notices another thing to to add a __str__ method to 19:30:20 #agreed Making sure our CI automation supports the Python 3 testing needs of other projects is a very high priority (and already basically solved). 19:30:41 everyone comfortable with those two stated agreements? 19:30:51 yes 19:30:52 yup 19:30:56 ++ 19:31:08 ++ 19:31:15 Yep 19:31:26 thanks! this was very helpful 19:31:47 fungi: thank you for interfacing! 19:32:03 anybody else have anything they want to say on the python 3 goal front before i yield teh floor to clarkb's topic? 19:32:49 #topic Switch review.o.o openid to login.ubuntu.com (clarkb) 19:33:14 we were originally talking about doing this late last week or early this week i guess 19:33:34 ya basically we should be thinking about this so lp doesn't berak us again as supposedly we are using the wrong url 19:33:44 (we should also be thinking about storyboard and wiki) 19:33:57 yeah, I've +1'd a patch for storyboard 19:33:59 'wrong' but not 'deprecated' :) 19:33:59 I can't merge it 19:34:10 since it's in the system config repo 19:34:22 zara_the_lemur__: ya and the tricky part is we have to update the config, and db at roughly the same time 19:34:47 gerrit is a bit easier because we can update db while it is running as gerrit allows for more than one openid 19:35:20 right, with gerrit we can basically just leave the "old" openid urls in the db forever 19:35:28 for gerrit I think our rough plan should be 1) update db 2) merge config change 3) restart gerrit when config is applied. 19:35:53 and then question si do we want to schedule a more formal downtime, get everything done ahead of time so that when GC problems occur we do restart then or something else? 19:36:21 clarkb: that's still a little racy if people make new accounts during then 19:36:30 the tricksy timing for gerritses comes down to minimizing the window between the db additions and the config change, since new accounts created between them could end up broken/duplicated 19:36:37 er, what jeblair just said 19:36:39 fungi: :) 19:36:44 jeblair: oh good point 19:37:00 so maybe is best to stop gerrit, update db, update config, start gerrit 19:37:02 clarkb: so we may want a second 'update the db' change to catch any new accounts; or maybe we just want to take a brief outage and do the db update and config change while offline 19:37:27 it's not as if gerrit's under heavy use this time of year 19:39:12 jeblair: problem with catch new accounts is they won't work in the interim 19:39:28 actually maybe they will I think monty mentioned he tested the old and new things accept the old and new urls both 19:39:30 yeah, i lean toward the outage idea 19:39:31 rather, they'll get duplicated 19:39:34 but outage will be safer 19:40:17 i did test that the lp api calls used in jeepyb support user lookups via openid interchangeably with both 19:40:34 so we can patch those any time we like 19:41:29 are people around tomorrow? 19:41:39 yep 19:41:48 i am not around tomorrow, but if other people are and want to work on it then please feel free to do so 19:41:59 i'll be driving a good chunk of the day 19:42:08 mordred: did the review-dev change, any chance you will be around tomorrow and we can do the prod server? 19:42:14 I will be around 19:42:28 * jhesketh isn't but can be if needed 19:42:32 jeblair: we could do it if you are willing to 19:42:38 clarkb: yah 19:43:05 clarkb: wfm 19:43:16 great what time works tomorrow? say 1800UTC? 19:43:33 clarkb: I'm free all day 19:43:51 ok I should be awake by 1800UTC but will ping mordred and we can get it done 19:44:43 1800 is good for me 19:44:50 I will also likely poke at storyboard-dev at some point tomorrow as well 19:45:28 sounds good. Thats mostly what I wanted out of this topic. How do we want to go about it and when and those have both been answered 19:46:24 :) 19:47:27 i'm already comfortable this should be pretty straightforward. the main trick which mordred already solved is the etl to add/replace rows keeping some substring constant while altering another substring 19:47:55 I won't be around much tomorrow to help, a heads up 19:48:03 happy to just find out the next time i get to the internet that it's done and behind us ;) 19:48:09 \o/ 19:50:18 okay, anything else need hashing out for the openid url replacement? 19:50:28 not that i have any great insights into it, but i will be around .au time if everything goes cray cray and you want to go to bed :) 19:50:55 thanks, everyone who's working on updating this 19:50:58 #topic Open discussion 19:51:03 I notice on #openstack-meeting-5 that 'openstack' is not an operator. Which seems incorrect. 19:51:32 I think that got traced back to the accessbot not having its nick set properly? 19:51:39 I know I +2'd the change to fix that 19:51:40 jlvillal: i believe it gets topic modifying permission anyway so shouldn't bother 19:51:55 s/bother/matter/ 19:52:03 maybe we do that through chanops? 19:52:19 i thought we just added the specific mode for topic 19:52:27 I just did a 'testing' meeting there and the topic stays the same 19:53:23 And doing the #topic command didn't seem to do anything 19:53:35 oh, meetbots get +O so are auto-opped 19:53:46 but that only takes effect on /join i think 19:54:19 so probably needs a restart or might be able to just /kick and let it rejoin? 19:54:34 or else it'll take effect the next time a change merges to its own configuration 19:54:42 and make sure it is setting its nick properly now 19:55:07 er I guess if it has the proper access in that channel that that did work /me checks 19:55:27 it's not on the access list for that channel 19:55:43 so accessbot is broken or its config is broken or has not run 19:58:16 https://review.openstack.org/#/c/410975/ did merge so possibly we just have to restart the accessbot? 19:58:35 oh, as mentioned i'm not around much tomorrow. i'll be mostly spotty with my online activity for the coming week, but should be around again normally from the 29th onward 19:59:36 and change your freenode password 19:59:36 happy holidays everyone! 19:59:53 thanks dims 20:00:04 aaaaand we're out of time--thanks everyone! 20:00:06 #endmeeting