13:01:20 <mvollmer> #startmeeting
13:01:20 <zodbot> Meeting started Mon May 18 13:01:20 2015 UTC.  The chair is mvollmer. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:20 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
13:01:31 <mvollmer> .hello mvo
13:01:32 <zodbot> mvollmer: mvo 'Marius Vollmer' <marius.vollmer@gmail.com>
13:01:46 <mvollmer> #topic Welcome
13:01:48 <mvollmer> Welcome!
13:02:00 <mvollmer> let's have the weekly cockpit irc meeting
13:02:03 <andreasn_> .hello andreasn
13:02:06 <zodbot> andreasn_: andreasn 'Andreas Nilsson' <anilsson@redhat.com>
13:02:22 <mvollmer> petervo, around?
13:02:45 <mvollmer> phatina, maybe you are interested.
13:03:01 <mvollmer> #topic Agenda
13:03:11 <mvollmer> * Storage update / packaging
13:03:31 <mvollmer> * CI status
13:03:47 <andreasn_> * ostree updates
13:04:12 <petervo> i'm here
13:04:43 <mvollmer> jscotka, meeting?
13:04:55 <jscotka> mvollmer, yep, I'm ready
13:05:03 <mvollmer> anything for the agenda?
13:05:26 <phatina> .hello phatina
13:05:27 <zodbot> phatina: phatina 'Peter Hatina' <phatina@redhat.com>
13:05:49 <jscotka> mvollmer, probably not, need to talk with Dominit :-)
13:05:54 <mvollmer> ok!
13:06:03 <mvollmer> #topic Status update / packaging
13:06:19 <mvollmer> ok, so now I should be close to feature complete... :-)
13:06:31 <mvollmer> I added some graphs
13:06:41 <mvollmer> but tey don't show individual disks anymore
13:07:07 <mvollmer> i hope we can figure that out more generally, with "The Grid" and D3, etc.
13:07:34 <mvollmer> there are bugs, maybe only with jquery amend.
13:07:45 <mvollmer> so its time to think about packaging
13:08:04 <mvollmer> i guess we could have a cockpit-storaged package, and a cockpit-udisks2 package
13:08:36 <mvollmer> I am not sure how to split the old code out of the shell, maybe it can stay there.
13:08:39 <github> [cockpit] stefwalter reopened pull request #2291: bridge: Add "owned" message to dbus-json3 (master...name-changed) http://git.io/vUhYj
13:09:12 <mvollmer> next up are getting the tests to run again.
13:09:25 <andreasn_> sounds good
13:09:33 <mvollmer> any opinions about the packaging?
13:09:50 <mvollmer> i guess we need to support the old udisks2 code for some while still.
13:10:59 <mvollmer> ok, next topic?
13:11:13 <mvollmer> #topic CI status
13:11:17 <mvollmer> it rains again...
13:11:33 <mvollmer> i tried to fix the frequent failures
13:11:45 <mvollmer> https://github.com/cockpit-project/cockpit/pull/2301
13:11:53 <mvollmer> https://github.com/cockpit-project/cockpit/pull/2300
13:12:07 <mvollmer> no definite results yet...
13:12:32 <mvollmer> but it's not as bad as it looks.
13:12:47 <mvollmer> stefw is fixing the Travis "stream" failure
13:13:08 <mvollmer> and the kube failure, forgot about that
13:13:42 <mvollmer> so, no green right now, but that is because of a few known failures.
13:14:13 <mvollmer> next? :-)
13:14:42 <andreasn_> sure
13:15:17 <mvollmer> #topic ostree updates
13:15:26 <andreasn_> petervo has a branch
13:15:38 <petervo> i put a tree up on cockpit-files
13:15:46 <andreasn_> and since I'm terrible with these things, I'm still struggling to get things working :)
13:16:48 <petervo> should be mostly working needs review and testing, the daemon still needs some work as well
13:17:24 <petervo> i'll help andreasn get it working
13:17:33 <andreasn_> really close now
13:17:53 <sgallagh> mvollmer: I have a topic for later in the meeting related to *non*-ostree updates (for Fedora Server)
13:18:12 <mvollmer> sgallagh, noted, thanks.
13:18:26 <mvollmer> petervo, where is the daemon?
13:18:56 * mvollmer reads good things about ember in the interwebs
13:20:37 <mvollmer> ok, next?
13:20:42 <andreasn_> sure
13:20:54 <andreasn_> I guess that's sgallagh's item then
13:21:02 <mvollmer> #topic  *non*-ostree updates
13:21:09 <sgallagh> Hello
13:21:11 <andreasn_> as in rpm-updates?
13:21:27 <sgallagh> Yes
13:21:51 <sgallagh> In a recent Fedora Server meeting, we addressed the possibility of having Cockpit be capable of managing standard RPM updates (using PackageKit)
13:22:31 <sgallagh> There are (as always) multiple opinions on how this should be done.
13:23:18 <andreasn_> such as offline updates or not?
13:23:24 <sgallagh> Option 1) Cockpit behaves like DNF, just updates the packages live (possibly offering an option to reboot afterwards)
13:23:35 <sgallagh> Option 2) Offline updates just like Fedora Workstation
13:23:51 <andreasn_> does option 1 include restarting the affected services and stuff?
13:23:56 <sgallagh> Option 3) "Warm" updates where the system descends into the update-only mode and then resumes to default.target
13:24:26 <sgallagh> andreasn_: Doing that successfully is a complicated problem.
13:24:41 <andreasn_> I assumed so
13:24:49 <sgallagh> For example, it's extremely difficult to know which services to restart if you update a single libraryt.
13:25:07 <sgallagh> (With glibc, rebooting is really the only sane option)
13:25:19 <sgallagh> Or rather, restarting everything.
13:25:39 <sgallagh> Option 2) is the most resilient, while resulting in the most downtime/disruption to the user.
13:26:07 <sgallagh> It guarantees a known state of the system, at the expense of two reboots.
13:26:36 <sgallagh> This can be extremely costly on a lot of server hardware which may have multiple-minute EFI memory checks.
13:27:02 <andreasn_> right
13:27:34 <sgallagh> Actually, I should add Option 4): Option 3, but with a reboot after updating instead of a resume to default.target
13:28:08 <sgallagh> If we implemented Option 3, we should have Option 4 around in any case, for glibc/kernel updates.
13:28:22 <andreasn_> I thought we had an issue open about rpm updates, but I can't seem to find it
13:28:46 <petervo> there were issues with packaging
13:29:03 <petervo> and dependcies
13:29:08 <sgallagh> Anyway, the approach is still up for discussion, but I wanted to open the conversation with Cockpit upstream
13:29:13 <petervo> but those should be resolved now i think
13:29:19 <sgallagh> petervo: Sorry, can you be more descriptive?
13:30:06 <petervo> Packagekit was pulling in x11 and a bunch of graphics libraries as dependencies
13:31:26 <petervo> i haven't had a chance to look at the latest but i think those were removed
13:31:47 <sgallagh> petervo: I'm not sure if they have been or not, but Server is carrying PackageKit at this point anyway
13:32:08 <petervo> is that new with f22?
13:32:09 <sgallagh> /me notes he's also working on a project to detect large dependency sets so we can narrow those down.
13:33:21 <mvollmer> cockpit right now isn't very good with anticipated reboots.
13:33:23 <sgallagh> petervo: I'm not sure; I *thought* it was also in F21, but I might be mistaken
13:33:34 <mvollmer> so that would be one thing to fix as part of this, I guess
13:33:45 <sgallagh> I should note that it's not explicitly part of the compose; something we do include pulls it in, apparently
13:33:45 <petervo> it wasn't back in jan when i was testing, but i was using cloud images
13:34:04 <sgallagh> petervo: Cloud images are a different compose set than Fedora Server
13:34:21 <petervo> that might be it
13:34:57 <sgallagh> Actually, I take that back. PackageKit *is* explicitly included as part of the "headless-management" group in comps
13:34:59 <petervo> mvollmer, we are going to work on that as part of the ostree work
13:35:02 <sgallagh> It's there for openlmi
13:35:09 <petervo> since that requires reboot
13:35:12 <sgallagh> And yeah, it would have been there in F21 also
13:35:16 <mvollmer> petervo, ahh, yes, that makes a lot of sense.
13:38:37 <sgallagh> OK, anyway I mostly brought this up to have you guys think about it (and ideally work up some mock-ups on how the user-experience should be)
13:38:45 <andreasn_> I'll keep it in mind
13:38:49 <sgallagh> Thank you
13:38:52 <mvollmer> looks like we are not scared of updates via PackageKit
13:38:58 <sgallagh> I'm *hoping* for this to be an F23 target.
13:39:05 <sgallagh> mvollmer: Hmm?
13:39:13 <andreasn_> and I'm not sure what the best answer is in the end with regards to reboots and all. It's a bit of pick-your-poison
13:39:15 <andreasn_> :)
13:39:18 <andreasn_> (as always)
13:39:38 <mvollmer> sgallagh, i didn't want to say "we are ready", but I don#t see any blockers.
13:40:49 <sgallagh> andreasn_: Yeah, there are no good answers. The actual mechanism under the hood should hopefully be invisible to Cockpit
13:40:54 <andreasn_> right
13:41:16 <sgallagh> Except insofar as we should provide feedback to Cockpit whether a reboot will be necessary, so it can inform the user
13:41:36 <sgallagh> *invisible to the Cockpit UI (not to the Cockpit implementation, necessarily)
13:42:02 <andreasn_> if we end up with offline updates, I think it can be fairly similar to the ostree updates we're working on right now
13:42:59 <sgallagh> Except for Option 1, all of the approaches I listed above would be *effectively* offline updates from Cockpit's point of view
13:43:07 <sgallagh> Since Cockpit wouldn't be reachable during that time.
13:43:56 <andreasn_> I have no idea if we would have time to get it to work for F23
13:44:06 <andreasn_> but it would be good to have at some point
13:44:25 <sgallagh> It's a pretty significant missing piece from Fedora Server's perspective.
13:44:35 <andreasn_> yeah
13:45:32 <sgallagh> That's all I had for todahy
13:45:45 <mvollmer> ok!
13:45:54 <mvollmer> anything else?
13:46:17 <andreasn_> nothing from me
13:46:25 <mvollmer> ok.
13:46:32 <mvollmer> quick meeting, nice.
13:46:36 <mvollmer> #endmeeting