atomic
MINUTES
20:04:18 <jzb> #startmeeting
20:04:18 <zodbot> Meeting started Tue Sep  9 20:04:18 2014 UTC.  The chair is jzb. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:04:18 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
20:04:49 <jzb> #chair stickster walters dustymabe number80 oddshocks dgilmore nirik
20:04:49 <zodbot> Current chairs: dgilmore dustymabe jzb nirik number80 oddshocks stickster walters
20:04:52 <jzb> miss anybody?
20:05:08 * nirik is lurking around
20:05:32 <stickster> .hellomynameis pfrields
20:05:33 <zodbot> stickster: pfrields 'Paul W. Frields' <stickster@gmail.com>
20:05:39 <dustymabe> .hellomynameis dustymabe
20:05:40 <zodbot> dustymabe: dustymabe 'Dusty Mabe' <dustymabe@gmail.com>
20:05:45 <jzb> .hellomynameis jzb
20:05:46 <zodbot> jzb: jzb 'Joe Brockmeier' <jzb@zonker.net>
20:06:05 <jzb> so, it may have been a few weeks since we had have an actual meeting.
20:06:30 <stickster> Yeah, congrats to walters is implicated in that ;-)
20:06:39 <jzb> heh, indeed.
20:06:43 <walters> thanks =)
20:06:43 <dustymabe> walters: glad to have you back
20:07:09 <jzb> walters: the important question - are you catching up on your sleep in the office? :-)
20:07:42 <walters> heh
20:07:52 <jzb> ok, nirik - last week, dgilmore pointed me at a TC6 (I think it was) compose for Atomic
20:08:06 <jzb> nirik: so are we now producing Atomic images regularly?
20:08:21 <nirik> with the test composes, yeah. But not nightly yet.
20:08:33 <jzb> nirik: any sticking points or blockers left there?
20:08:41 <number80> .hellomynameis hguemar
20:08:42 <zodbot> number80: hguemar 'Haïkel Guémar' <karlthered@gmail.com>
20:08:57 <nirik> not sure. I think we just need to get it added to the compose scripts.
20:09:50 <stickster> If oddshocks is around... maybe he'd be inclined to help with some of the rel-eng bits too. There's a lot going on in rel-eng due to multi-product Fedora, and this is one component.
20:09:56 <walters> that's this script right? https://git.fedorahosted.org/cgit/releng/tree/scripts/run-pungi
20:10:17 <walters> or is that only for the manually-initiated composes, and something else is nightly?
20:11:12 <nirik> https://git.fedorahosted.org/cgit/releng/tree/scripts/buildbranched and https://git.fedorahosted.org/cgit/releng/tree/scripts/buildrawhide
20:11:22 <nirik> which call run-pungi for the pungi part.
20:11:59 * stickster amazed but not surprised by the fact that nirik seems to know everything
20:12:08 <stickster> about *everything*
20:12:14 <nirik> heh. I'm old. ;)
20:13:06 <stickster> #info note, we are producing Atomic images regularly in test composes, not yet in nightlies
20:14:10 <jzb> nirik: so - adding to the script, is this a lot of work, a formality, or...?
20:14:28 <nirik> a bit of work....
20:14:34 <nirik> but nothing too vast. ;)
20:15:24 <jzb> nirik: so - is that something you or dgilmore will handle, or do you need assistance?
20:15:36 <nirik> assistance would be great. Patches to the releng list.
20:16:16 <stickster> #info Help is needed patching the compose scripts!
20:16:24 <stickster> #url https://git.fedorahosted.org/cgit/releng/tree/scripts/run-pungi
20:16:29 <stickster> #url https://git.fedorahosted.org/cgit/releng/tree/scripts/buildbranched
20:16:37 <jzb> any volunteers?
20:16:38 <stickster> argh, wrong command.
20:16:44 <stickster> #link https://git.fedorahosted.org/cgit/releng/tree/scripts/run-pungi
20:16:49 <stickster> #link https://git.fedorahosted.org/cgit/releng/tree/scripts/buildbranched
20:16:55 <stickster> #link https://git.fedorahosted.org/cgit/releng/tree/scripts/buildrawhide
20:17:06 <stickster> jzb: I'm nominating oddshocks  :-)
20:17:19 <jzb> is oddshocks around?
20:17:43 <stickster> He doesn't appear to be here right now -- might be lunchtime his TZ -- but I'm dropping him an email to request his help
20:17:47 <jzb> OK
20:18:16 <jzb> anything else on the composes?
20:18:21 <jzb> I think that's pretty much it.
20:18:31 <walters> is the current compose mirrored?
20:18:39 <jzb> ah, good question
20:19:01 <jzb> walters: I don't think the last one was - it's just in Koji.
20:19:09 <walters> the repo is mirrored here? http://dl.fedoraproject.org/pub/alt/stage/atomic/
20:19:34 <nirik> that area is mirrored... but not by the full pile of mirrors, only some few that want to
20:20:10 <walters> #topic metalink XML, mirroring
20:20:16 <walters> (is that right?)
20:20:37 <walters> to recap from before the meeting, http://koji.fedoraproject.org/koji/buildinfo?buildID=576104 should support metalinks and has a new "summary" file
20:20:40 <jzb> walters: yeah, I don't think the meetbot has perms to change the channel topic though.
20:21:02 <walters> to make this work we'd need to have the updated ostree on the compose server (should just happen via mock next run I assume)
20:21:07 <stickster> #info http://koji.fedoraproject.org/koji/buildinfo?buildID=576104 should support metalinks and has a new "summary" file
20:21:11 <walters> then we need to patch the scripts to run ostree summary -u
20:21:21 <walters> both are easy
20:21:34 <walters> what's unclear to me is the next step to hook in with mirrormanager
20:21:34 <nirik> cool.
20:21:48 <walters> and have it generate the XML
20:22:07 <nirik> mirrormanager has a cron called 'update master directory list' where it crawls the filesystem looking for new repos/things or updated repos/things.
20:22:26 <nirik> I think we need to teach that to identify ostree dirs/trees.
20:22:33 <walters> if we can iterate on this soon that'd be good, because while I *think* the metalink code will work, it'd be nice to test it in something resembling reality soon
20:22:39 <walters> ah
20:23:18 <nirik> I'm not sure where it is in code, but can look/ask around.
20:23:26 <walters> a suitable pattern would probably be "repo/summary" (if there can be conditionals, if repo/objects exists too)
20:24:32 <nirik> alright
20:25:15 <stickster> nirik: Might be this? https://git.fedorahosted.org/cgit/mirrormanager/tree/server/update-master-directory-list
20:26:36 <nirik> sounds right..
20:29:20 * oddshocks looks around
20:29:21 <walters> once the file mapping is updated, this process would automatically handle filtering the generated XML for which mirrors carry the atomic repo?
20:29:48 <oddshocks> sorry for being late. alarm might have been messed up by me listening to music on the phone.
20:29:50 <nirik> walters: not 100% sure, but I think so. it should save it in the db and then check sites against it.
20:31:15 * stickster wonders if people are looking at code now and not really IRC :-) Sorry for derailing, if so.
20:31:18 <jzb> oddshocks: you might want to scroll up
20:31:48 * oddshocks is now caught up
20:31:53 <walters> jzb, you said there had been a compose?
20:31:57 <walters> do you have a link?
20:32:03 <jzb> walters: yeah, somewhere, hang on...
20:32:57 <oddshocks> I'm glad I didn't completely miss this because I was hoping to get an update -- now that I'm done moving -- on what still needs to be done with MM, or if there is something else I could do that is higher priority. are these aforementioned scripts something I should take a look at?
20:33:45 <jzb> walters: http://koji.fedoraproject.org/koji/taskinfo?taskID=7531374
20:34:06 <walters> oddshocks, we'll need to run ostree summary -u on the compose server, then MM will need to generate a metalink to the repo/summary file resulting from that
20:34:16 <walters> (in the future, there will be a GPG signing step in there too)
20:34:29 <stickster> walters: I think oddshocks is asking about prior scripts as well for nightly generation
20:34:44 <walters> ah right
20:34:49 <stickster> oddshocks: In which case -- yes, some help would be appreciated there too
20:35:29 <stickster> oddshocks: I'm pretty sure that dgilmore has a lot going on w/ multiple Fedora products and so any script fixing you can assist with would be a good idea (and hopefully welcome!)
20:36:46 <oddshocks> got it. are any of these completed or changed?: http://ur1.ca/i5qk7 http://ur1.ca/i5qkg
20:37:55 <walters> yep, the ostree metalink part is done
20:38:00 <oddshocks> walters: cool
20:38:14 <oddshocks> are all of these scripts in the releng scripts repo or..?
20:39:05 <walters> i think so
20:39:20 <jzb> 20:16 < stickster> #link https://git.fedorahosted.org/cgit/releng/tree/scripts/buildbranched
20:39:23 <jzb> 20:16 < stickster> #link https://git.fedorahosted.org/cgit/releng/tree/scripts/buildrawhide
20:39:43 <jzb> oddshocks: ^^ I think those are what we're talking about?
20:40:10 <nirik> yep. those two for nightly composes
20:40:41 <jzb> oddshocks: so we can put you down for tackling adding Atomic?
20:41:02 <oddshocks> sure, I'll go at it
20:41:49 <jzb> #action oddshocks work on adding Atomic to releng scripts
20:42:10 <jzb> walters: any other blockers or issues right now you're aware of?
20:43:05 <walters> how we pre-configure the generated images to point to the mirrors is as yet somewhat fuzzy to me
20:43:21 <walters> i had been waiting to tackle this until we had content mirrored and metalink XML generated
20:43:38 <jzb> walters: for updates, you mean?
20:43:40 <walters> basically in mainline, we ship a fedora-release package which contains yum repositories which link to metalinks
20:43:41 <walters> yes
20:44:29 <jzb> walters: on a scale of 1-10, how bad is it if they have to adjust after the alpha to get updates?
20:44:36 <walters> 3?
20:45:04 <walters> anyways not a blocker to my knowledge
20:45:04 <jzb> walters: with 1 being "no worries" and 10 being "you're being eaten by a shark"?
20:45:12 <stickster> heh
20:45:25 <jzb> OK
20:45:36 <jzb> Is this something we might be able to sort out before the beta?
20:45:43 <walters> hopefully yes
20:45:46 <jzb> ok
20:45:48 <dustymabe> so will the alpha image contain a location where we can get updates (during the alpha)?
20:46:14 <jzb> I would "vote" to have a non-mirrored location for the alpha if we can.
20:46:14 <dustymabe> I know that might sound funny but it would be nice to have an 'update' to perform for testing purposes
20:46:29 <jzb> dustymabe: right. That is sort of core functionality.
20:46:31 <walters> jzb, yeah, it's pretty easy to provide that
20:46:47 <walters> i'm a bit more concerned about the default experience/layout
20:48:39 <jzb> for the alpha, I think if Docker + Cockpit are working, that's a good start.
20:48:47 <jzb> but I agree.
20:49:41 <jzb> so - let's try to answer for next week how we get to pointing to mirrors?
20:49:51 <walters> yep
20:50:31 <jzb> any other topics this week?
20:50:35 <jzb> #topic open floor
20:51:13 <jzb> dustymabe: anything on your end?
20:51:17 <jzb> stickster: ^^?
20:51:22 <dustymabe> yeah actually
20:51:33 <roshi> missed the meeting didn't I?
20:51:37 <dustymabe> a few questions
20:51:54 <dustymabe> so in the test images from koji I notice we are using an LV structure
20:52:15 <stickster> jzb: The only thing I wanted to mention is that we are still looking slippy/not-good for Alpha. So if there's anything not working yet for Atomic wrt. Alpha release that *is* releease-blocking and not communicated elsewhere, don't wait to talk about it
20:52:21 <dustymabe> so we have sda1 and sda2 with sda2 chunked up into LVs
20:53:04 <jzb> roshi: little bit
20:53:25 <dustymabe> I notice centos7 at least at the moment was using sda1 sda2 sda3 and no LVM structure.. should we get together with them and have something consistent?
20:53:32 <roshi> sorry, had an unavoidable appt that ran long :(
20:53:55 <dustymabe> I don't theres is finalized either.. just wondering if we should collaborate
20:53:58 <jzb> dustymabe: I can certainly see the value in being consistent.
20:54:03 <jzb> dustymabe: so +1 yes.
20:54:30 <dustymabe> jzb: thanks
20:54:38 <oddshocks> walters: sorry, to clarify on the things that are done, is the repomd.xml parsing done, as well?
20:54:41 <walters> dustymabe, this is a really important topic but i see it as a bit outside of the atomic/fedora-infra meeting scope, though this meeting is sometimes pure atomic specific
20:55:12 <dustymabe> walters.. should we discuss after the meeting?
20:55:31 <walters> oddshocks, i'm not sure what that referred to exactly; ostree doesn't use repomd.xml.  Perhaps that task was about teaching MM to look for repo/summary (ostree repo style) instead of yum/repomd.xml ?
20:55:56 <oddshocks> walters: yeah I believe it was a mirrormanager task
20:56:00 * stickster is sorry he has to duck out for another appt.
20:56:17 <jzb> stickster: thanks for helping
20:56:24 <stickster> what little I did, np
20:56:25 <walters> dustymabe, i know i owe you a followup on that but can i get back to you by tomorrow?
20:57:05 <dustymabe> walters: sure. The messages on the list don't necessarily address collaboration with centos. I was thinking about sending a new email about that
20:57:23 <dustymabe> that is.. after we talked and perhaps came up with what we thought was the ideal layout
20:59:42 <walters> ok, i can do a chat after this
20:59:58 <dustymabe> walters: thanks.
21:00:06 <walters> anything else on the fedora-infra/atomic discussion?
21:01:39 <walters> thanks all!
21:01:41 <walters> #endmeeting