qadevel
LOGS
14:04:27 <tflink> #startmeeting
14:04:27 <zodbot> Meeting started Mon Mar 30 14:04:27 2015 UTC.  The chair is tflink. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:04:27 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
14:04:27 <tflink> #meetingname qadevel
14:04:27 <tflink> #topic Roll Call
14:04:27 <zodbot> The meeting name has been set to 'qadevel'
14:04:34 <tflink> #chair mkrizek kparal
14:04:34 <zodbot> Current chairs: kparal mkrizek tflink
14:04:38 * mkrizek is here
14:04:49 <tflink> I remembered to #chair this time :)
14:04:53 * kparal here
14:05:32 <tflink> who wants to go first for status?
14:05:41 <mkrizek> I'll go
14:05:49 <mkrizek> #topic mkrizek status report
14:05:56 <mkrizek> #info finished research on paramiko
14:06:01 <mkrizek> #link https://phab.qadevel.cloud.fedoraproject.org/T436
14:06:07 <mkrizek> #info infra playbook fixes
14:06:12 <mkrizek> #info found out that taskotron-dev playbook was run deploying execdb bits on buildmaster making dev machine stop executing jobs
14:06:23 <mkrizek> questions?
14:06:44 <tflink> I suppose that makes getting execdb built so it can deploy that much more important
14:06:59 <mkrizek> yep :)
14:07:16 <mkrizek> I can try building it
14:07:23 * tflink knew it was causing problems with the playbook, didn't realize it was causing problems with task execution
14:07:31 <tflink> but now that you mention it, that makes sense
14:07:49 <kparal> mkrizek: where's the initial implementation for paramiko wrapper you mention in T436?
14:07:53 <mkrizek> buildmaster tries to create job in execdb
14:08:11 <mkrizek> kparal: on my laptop :P
14:08:16 <kparal> ok
14:08:42 <kparal> I thought I missed something
14:08:57 * danofsatx is here if you need him
14:09:46 <mkrizek> anything else? otherwise that's about it from me
14:09:57 <tflink> i don't think so
14:10:06 <tflink> nothing from me, anyways
14:10:24 <kparal> thanks mkrizek
14:10:33 <kparal> I'll go
14:10:44 <kparal> #topic kparal status report
14:10:51 <kparal> #info looking at testCloud and other tools for VM boot and customization (cloud-init, imagefactory, oz, virt-builder)
14:10:56 <kparal> #link https://phab.qadevel.cloud.fedoraproject.org/T407
14:11:05 <kparal> I really like virt-builder at the moment, because it seems it could provide us with the exact images we would like to use (configured for our needs), it should be easy generate updated images very often, and we could have several different images for different use cases (a minimal one, one with preinstalled GNOME, etc). But I'm still exploring how exactly it works, it seems to boot the image in qemu for certain configuration steps,
14:11:05 <kparal> so this approach might not be easily usable inside a virtualized environment without any extra configuration. It depends whether we would run this on bare metal or in a VM. We could also try to make this into a taskotron task, of course, but it's more work, so probably not initially.
14:11:54 <kparal> there are some bugs that need to be fixed, in virt-builder, as well
14:11:56 <tflink> i think it makes sense to eventually run it as a task
14:12:01 <kparal> but I managed to work around them for the moment
14:12:10 <tflink> but yeah, not sure it's required to start with
14:12:12 <kparal> tflink: I agree, eventually
14:12:23 <tflink> kparal: is there a list of the stuff that needs to be fixed?
14:12:59 <kparal> tflink: not at the moment. if we decide to go with virt-builder, I'll start filing bugs into the project. I haven't found a showstopper yet
14:13:14 <kparal> one question, though
14:13:21 <tflink> I forget, can virt-builder update images?
14:13:48 <kparal> tflink: it can create them anew with updated packages. which should be pretty much the same
14:13:56 <tflink> nvm, I was thinking of something different
14:15:05 <kparal> Q: does it make sense to think about having multiple VM images for different use cases? e.g. a minimal one for overhead, but a GNOME one for GNOME-related tasks? otherwise we would have to install GNOME packages every time we run this task, and that's a lot of packages
14:15:21 <kparal> or am I overthinking this?
14:15:42 <kparal> *a minimal one for minimal overhead
14:15:55 <tflink> I think it makes sense to use different images for some different things
14:15:59 <mkrizek> I guess that would make sense, eventually but I'd start simple
14:16:11 <tflink> but I'm not sure it'll make sense to use cloud-style images for gnome testing
14:16:31 <tflink> I suspect that the workstation folks would want to see the live image used
14:16:42 <tflink> but that'd require livecd support
14:17:04 <kparal> ok, it's just whether I should take this into account when exploring tools and their possibilities. whether their ability to create/customize different images with different contents is desired
14:17:44 <tflink> i don't think it would hurt to keep it in mind when exploring tools
14:18:05 <kparal> tflink: we would have to install in onto disk anyways. the livecds are quickly out of memory if you start touching the disk
14:18:27 <tflink> just because live images seem to make more sense right now doesn't mean it'll definitely happen
14:18:48 <tflink> kparal: hopefully, we wouldn't be installing much on top of the livecds but that's a good point
14:19:00 <kparal> ok, thanks for answers
14:19:12 <kparal> that's it from me
14:19:16 <tflink> k, thanks
14:19:36 <tflink> #topic tflink status report
14:19:48 <tflink> not a spectacularly productive week for me :-/
14:19:59 <tflink> #info some more work on digging into depcheck
14:19:59 <tflink> #info looking into script to reproduce depcheck failures
14:20:17 <tflink> #info some debugging and new builds for ansible_utils (new rbac-playbook)
14:20:46 <kparal> tflink: one of the failures is here, just in case: https://admin.fedoraproject.org/updates/FEDORA-2015-4896/mate-control-center-1.8.3-2.fc21,mate-desktop-1.8.2-2.fc21
14:21:03 <tflink> I did find out that old mash data is archived for a bit
14:21:23 <tflink> so that'll help with reproducing
14:21:23 <tflink> kparal: thanks
14:22:01 <kparal> tflink: is this the archive you talked about, or is there some place where you can find even older composes? http://koji.fedoraproject.org/mash/
14:22:19 <tflink> that's just branched and rawhide
14:22:52 <kparal> and updates repo for stable releases
14:23:05 <tflink> yeah, I spoke too quickly
14:23:10 <tflink> that's what I was talking about
14:23:19 <kparal> ok, thanks
14:23:52 <tflink> but it's at least a week, so at least we aren't required to get repodata for reproducers the day it happens
14:24:23 <kparal> good news, didn't occur to me before
14:25:22 <tflink> with that archive, I'm thinking about writing a script to reproduce runs using the taskotron.log from failed runs - that lists all the rpms involved in a failure and gives the date that the task was run (to get repodata)
14:25:28 <tflink> but that's about all from me
14:25:37 <tflink> any other comments/questions?
14:25:52 <kparal> not here
14:25:55 <mkrizek> nope
14:26:00 <tflink> ok, moving on
14:26:03 <tflink> #topic planning
14:26:25 <tflink> one thing that I'd like to get done is rebuild stg with f21, keeping history this time
14:26:35 <tflink> which we didn't do when rebuilding dev
14:26:53 <tflink> I think beta freeze starts tomorrow, so I'd like to get that done today
14:27:06 <tflink> since rebuilding would require freeze break exceptions otherwise
14:27:32 <nirik> FYI, I think the dev playbooks are still not completing...
14:27:38 <tflink> any objections or concerns with that?
14:27:49 <tflink> nirik: I was about to get to that :)
14:28:04 <mkrizek> no objections
14:28:10 <kparal> no objections from me either :)
14:28:21 <tflink> execdb still isn't built as an rpm which is why I think the dev playbook is failing
14:28:30 <mkrizek> yes
14:28:40 <tflink> we can add execdb into stg later - that won't require a freeze break since that would only affect stg systems
14:28:53 <tflink> rebuilding requires host key changes that affect all infra systems, though
14:29:04 <nirik> yeah, we should detangle that sometime. ;(
14:29:33 <tflink> #info will be rebuilding taskotron-stg with f21 today (before beta freeze starts)
14:30:18 <tflink> a hopefully minor task is wrapping up the execdb stuff - building the rpm, getting it into the infra-testing repo and deployed on dev so that it starts functioning
14:30:49 <tflink> I think that some more playbook changes are going to be required, as well
14:31:07 <tflink> the taskotron-proxy role doesn't handle the frontend for execdb
14:32:35 <tflink> I'll get tickets filed for that if anyone wants to poke at it
14:32:48 <tflink> but getting dev functioning is more important :)
14:33:25 * tflink also wants to move the execdb repo from josef's user to fedoraqa so it matches the rest of the projects, but that isn't incredibly urgent
14:33:46 <tflink> any other questions or thoughts on execdb?
14:33:56 <tflink> even though josef is not here today
14:34:11 * kparal agrees to moving the repo
14:35:21 * tflink assumes nothing more for execdb
14:35:40 <kparal> not here
14:35:50 <tflink> #info execdb to be built as rpm, some more playbook changes are likely to be needed to complete deployment in dev
14:36:17 <tflink> #info different playbook changes will be needed for deployment in stg/prod due to different proxy setup
14:36:43 <tflink> moving on to disposable clients
14:37:19 <tflink> mkrizek created a new remote tracking branch for disposable client work last week but it's not diverged from develop yet
14:37:38 * tflink has been thinking about how to best approach moving forward with code and came up with a couple approaches
14:38:16 <tflink> 1) submit most of the PoC code for review as a starting point, knowing that it'll need to be changed before it's complete
14:38:45 <tflink> 2) come up with a skeleton for functionality so that folks can start working on different bits
14:39:01 <tflink> unless there are other ideas on how to start moving forward
14:39:53 <kparal> sounds reasonable
14:40:18 <tflink> any thoughts on which approach is better?
14:40:32 <tflink> keeping in mind that I don't remember if the test suite passes on the PoC
14:40:34 <mkrizek> 2) seems cleaner
14:41:03 <kparal> aha, I thought having both of them :)
14:41:26 <tflink> kparal: how would both work?
14:42:29 <kparal> skeleton is probably cleaner and easier to look at
14:42:43 <kparal> provided we have the full picture and can write it
14:43:05 <tflink> kparal: not sure I understand what you mean by "full picture and can write it"
14:43:32 <kparal> whether we know how exactly the different pieces fit together
14:44:12 <tflink> most of it makes sense in my head - not sure if it's been communicated well enough, though
14:44:24 <tflink> you all would know that better than I would, though :)
14:44:31 <kparal> maybe I'm just confused a bit, don't mind me and go on, I'll catch up :)
14:46:02 <mkrizek> anyway I vote for 2) :)
14:46:12 <tflink> it sounds like 2) is the winner
14:46:28 * tflink agrees that it's a cleaner start. there's a reason he did the PoC in a different repo
14:46:49 <tflink> the next question is who works on the skeleton
14:47:12 <tflink> I suspect that rebuilding stg will keep me busy for most of the day
14:47:45 <tflink> so it'd be tuesday/wednesday before I could have anything ready for review
14:48:11 <tflink> unless someone else wants to work on the skeleton
14:48:45 <mkrizek> not sure if we have that much insight to work on that
14:49:03 <kparal> I still have plenty to do. are we in a hurry about this?
14:49:31 <tflink> i think that mkrizek is mostly blocked until it's done
14:50:36 <mkrizek> I can work on infra tickets until then
14:50:41 <tflink> and I'm not sure how much jskladan has left in his queue without the skeleton, either but that's a question for him :)
14:51:34 <kparal> tflink: if you're willing to create that skeleton, that would be great. I'd have to re-read a lot of docs to get into the picture again
14:51:55 <tflink> ok, I'll work on that as soon as stg is redeployed
14:51:59 <tflink> since that has to be done today
14:52:27 <tflink> I'd rather not wait until after beta so that we have more time to make sure everything's stable
14:53:43 <tflink> #action tflink to work on disposable client skeleton after stg is redeployed as f21
14:54:08 <tflink> I think that's about it for planning from me
14:54:12 <tflink> any other topics?
14:54:25 <mkrizek> I have one
14:54:37 <mkrizek> do we want to proceed with fedmsg integration?
14:54:52 <mkrizek> it's blocked on depcheck
14:55:51 <kparal> you mean making depcheck report per-build as well as per-update, right?
14:55:55 <mkrizek> yes
14:56:42 <kparal> I have no idea how hard it can be, but I can ask jskladan to have a look at it tomorrow, if he's no longer sick
14:56:49 <kparal> he was looking for a new ticket, I think
14:57:10 <mkrizek> sounds good
14:57:24 <kparal> does it matter if we proceed with fedmsg integration and have just upgradepath reporting for the moment?
14:57:52 <tflink> did we finalize a fedmsg schema for results?
14:58:05 <mkrizek> afaik yes
14:58:11 * tflink remembers talking about it, doesn't remember the conclusion off the top of his head
14:58:30 <mkrizek> on @qadevel
14:59:05 * tflink reads
15:01:46 <tflink> yeah, the schema was mostly decided on list (org.fedoraproejct.taskotron.result)
15:02:05 <tflink> ticket for depcheck exists, not taken yet
15:02:07 <tflink> https://phab.qadevel.cloud.fedoraproject.org/T388
15:03:10 <tflink> er, ticket to get depcheck reporting list of related builds
15:05:08 <tflink> i think it'd be worth running through the entire plan before we deploy to stg (not sure we can do this in dev since there's no dev fedmsg AFAIK)
15:05:42 <mkrizek> I discussed initial steps with ralph some time ago
15:05:54 <mkrizek> have most of them ready
15:06:25 <tflink> yeah, just thinking it would be worth having everything in one place to make sure that everyone's on the same page before we start deploying stuff
15:06:39 * mkrizek agrees
15:08:06 <tflink> any other thoughts on this?
15:08:16 <tflink> or questions/concerns that haven't been covered?
15:08:29 * kparal has none
15:08:38 <mkrizek> nothing here
15:08:43 <tflink> #topic Open Floor
15:08:49 <tflink> anything for open floor?
15:09:02 <kparal> nope
15:09:06 <mkrizek> no
15:09:18 <tflink> almost an hour :)
15:09:20 <roshi> I think I got things working for testCloud
15:09:29 <kparal> roshi: oh, great news
15:09:32 <roshi> the issues I was running into ended up being https://bugzilla.redhat.com/show_bug.cgi?id=1159823
15:09:36 <tflink> roshi: got a new laptop? :-P
15:09:41 <roshi> tldr: nvidia blobs
15:09:42 <roshi> heh
15:10:14 <roshi> so if you're not using nvidia blobs (which I use to drive HDMI) it should work fine ootb now
15:10:44 <roshi> it was a long, annoying yak shave. But at least the yak is finally shaved
15:10:53 <tflink> while we're on the topic - you finalized 'testcloud' as the name for the project, right?
15:11:05 <roshi> yeah, it turns out legal had no issues with it
15:11:14 <tflink> ie, approved by legal and no plans to change the name anytime in the near future
15:11:17 <roshi> so figure I'll just stick with it
15:12:31 <roshi> that's all I had, just wanted to update people
15:12:40 <roshi> since I knew kparal was poking at things
15:12:53 <kparal> thanks for info
15:12:55 <tflink> #info testcloud should be working out of the box now
15:13:01 <kparal> I'll update my git
15:14:05 <tflink> if there's nothing else, I'll close the meeting
15:14:33 <kparal> thanks for leading, tflink
15:14:42 <tflink> thanks for coming, everyone
15:14:48 * tflink will send out minutes shortly
15:14:52 <tflink> #endmeeting