13:01:46 <andreasn> #startmeeting
13:01:46 <zodbot> Meeting started Mon Apr 13 13:01:46 2015 UTC.  The chair is andreasn. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:46 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
13:01:59 <andreasn> .hello andreasn
13:02:00 <zodbot> andreasn: andreasn 'Andreas Nilsson' <anilsson@redhat.com>
13:02:00 <dperpeet> .hello dperpeet
13:02:03 <zodbot> dperpeet: dperpeet 'Dominik Perpeet' <dperpeet@redhat.com>
13:02:12 <sgallagh> .hello sgallagh
13:02:13 <zodbot> sgallagh: sgallagh 'Stephen Gallagher' <sgallagh@redhat.com>
13:02:19 <mvollmer> .hello mvo
13:02:20 <zodbot> mvollmer: mvo 'Marius Vollmer' <marius.vollmer@gmail.com>
13:02:31 <stefw> .hello stefw
13:02:32 <zodbot> stefw: stefw 'Stef Walter' <stefw@redhat.com>
13:02:52 <andreasn> petervo: around as well?
13:03:33 <andreasn> #topic Agenda
13:03:38 <andreasn> what do we have today?
13:03:42 <stefw> * Metrics, Graphs, cAdvisor, Kubernetes update
13:03:44 <andreasn> * Fedora 22
13:03:52 <stefw> * OSTree update, if Peter's around
13:04:06 <mvollmer> * Services rework
13:04:53 <petervo> yep
13:05:14 <mvollmer> * Multi OS testing
13:06:04 <stefw> * New CI migration progress
13:06:14 <stefw> * Translations to Zanata
13:07:34 <andreasn> sounds like we have quite the agenda to fill the hour
13:07:38 <andreasn> ok, lets start
13:07:51 <andreasn> * Metrics, Graphs, cAdvisor, Kubernetes Update
13:07:54 <andreasn> err..
13:08:01 <andreasn> #topic  Metrics, Graphs, cAdvisor, Kubernetes Update
13:08:28 <stefw> I've been writing some javascript code to handle populating time series data from various sources, including our own metrics channels, and also things like cAdvisor
13:08:32 <stefw> some work here: https://github.com/cockpit-project/cockpit/pull/2131
13:09:05 <stefw> I'll continue working on this and adapting it until I have working graphs on the kubernetes page and also pulling from some other source
13:09:22 <stefw> the goals of the work are to a) Put something other than 'TODO Graphs' on the cluster dashboard
13:09:33 <stefw> and b) migrate the remaining graphs throughout cockpit to use the metrics channels
13:09:51 <stefw> One other interesting thing to note, is that cAdvisor now comes as part of kubernetes (in the kubelet process)
13:09:56 <andreasn> sounds like good goals
13:10:05 <stefw> the other interesting thing to note is that 'docker stats' sucks various things
13:10:06 <andreasn> happy to hear the TODO going away
13:10:33 <stefw> i think that's the brief highlevel, i don't mind going into more detail if people have questions
13:10:59 <mvollmer> just to confirm:
13:11:09 <andreasn> does the kubernetes page still use the old patternfly version?
13:11:18 <mvollmer> this will allow us to port the 'multi' graphs away from cockpit-wrapper, right?
13:11:30 <stefw> andreasn, i don't think so
13:11:36 <stefw> mvollmer, yes, that's right
13:11:43 <mvollmer> multi graph = multiple network devices in one plot.
13:11:52 <andreasn> stefw: so we're rid of the old patternfly? Nice!
13:11:52 <mvollmer> right, very good.
13:11:55 <stefw> andreasn, yes
13:12:02 <stefw> mvollmer, the code contains ways to combine information from multiple channels into a single 'grid' ... and then perform calculated rows
13:12:05 <stefw> as well
13:12:20 <mvollmer> nice.
13:12:41 <stefw> so in theory you could even do crazy things like combining data from cAdvisor and metric channels into a single graphs
13:12:59 <stefw> /s$//
13:13:02 <mvollmer> doesn't sound too crazy.
13:13:50 <stefw> we do this by quantizing the time series data
13:13:51 <stefw> so that it lines up based on the interval
13:14:28 <stefw> we already had some elements of this in our metric channel code, but this takes it a step further, so that you can intelligently calculate stuff between various sources
13:14:49 <stefw> #info https://github.com/cockpit-project/cockpit/pull/2131
13:15:14 <andreasn> sounds good
13:15:30 <andreasn> #topic Fedora 22
13:15:30 <mvollmer> very nice indeed.
13:15:49 <andreasn> sgallagh: when is the beta date?
13:15:57 <mvollmer> one thing on the list is "partial exposures" to make scrolling smoother
13:15:59 <mvollmer> does that fit into #2131?
13:16:19 <mvollmer> err, I mean, did you take it into consideration already?
13:16:32 <stefw> in the D3 world the smoother scrolling happens at the view level
13:16:42 <mvollmer> (don't want to say that it should be part of it)
13:16:42 <stefw> and has nothing to do with the data
13:16:44 <sgallagh> mvollmer: We are in Beta freeze right now
13:16:53 <sgallagh> We hoped to release tomorrow, but we had to slip a week.
13:16:57 <andreasn> all right
13:17:02 <mvollmer> me missed the topic change, sorry.
13:17:28 <sgallagh> We need to have a new Release Candidate today or tomorrow to approve the release at Go/No-Go on Thursday
13:17:31 <stefw> mvollmer, so i'm personally not ready to figure that out, as i don't understand the tradeoffs
13:17:51 <sgallagh> (oops, I responded to mvollmer and meant to respond to andreasn)
13:18:04 <stefw> no worries, now we should be on the Fedora 22 topic
13:18:14 <mvollmer> it's all messed up, let's start again!
13:18:28 <mvollmer> :-)
13:18:52 <github> [cockpit] stefwalter opened pull request #2134: tools: Allow make-srpm and make-rpms to be used without git clone (master...make-srpm-without-git) http://git.io/vv3wT
13:18:59 <andreasn> do we have everything that we want in F22?
13:19:20 <stefw> andreasn, we want to push some more changes
13:19:24 <andreasn> would be nice to get https://github.com/cockpit-project/cockpit/pull/2061 in, so I'll review that asap
13:19:26 <stefw> there are several that help with Fedora Atomic
13:20:01 <stefw> I think we're totally within the guidelines pushing our recent releases into Fedora after the beta freeze.
13:20:47 <andreasn> right
13:20:55 <andreasn> anything else on that topic?
13:21:39 <andreasn> #topic OSTree update
13:22:06 <petervo> still don't have any ui to show
13:22:29 <petervo> finalizing the changes based on the last feedback i got from the ostree guys
13:22:48 <andreasn> how are things with the dependencies?
13:23:18 <petervo> ostree is fine, packagekit was the one with the issues
13:23:26 <andreasn> right
13:23:42 <petervo> but we decided to not do any implementation there for now and focus on ostree
13:23:56 <andreasn> yeah, I mixed them up
13:24:06 <github> [cockpit] mvollmer closed pull request #2133: shell: Correctly transcribe !startsWith to indexOf !== 0. (master...fix-starts-with) http://git.io/vvOAr
13:24:31 <andreasn> let me know as soon as you have any ui I can play around with, anxous to try that
13:24:44 <andreasn> play around/destroy systems
13:24:49 <stefw> andreasn, petervo, by the way /usr/local/share/cockpit is the ticket
13:25:02 <andreasn> ticket?
13:25:05 <stefw> as well as ~/.local/share/cockpit
13:25:12 <stefw> when trying to develop this stuff directly on top of Atomic
13:25:16 <stefw> as you can't 'make install'
13:25:19 <andreasn> oh, good to know
13:25:20 <andreasn> right
13:25:33 <andreasn> so you just compile things and dump them into that?
13:25:52 <stefw> just copy or link the pkg/ostree (if that's what peter calls it) directory into those locations
13:25:58 <stefw> and ideally that'd be enough
13:26:01 <andreasn> good
13:26:16 <stefw> compiling is a good step to check javascript correctness and stuff like that
13:26:40 <andreasn> all right. Anything else on ostree?
13:26:55 <stefw> petervo, any obstacles come up?
13:27:25 <petervo> i have one issues i wouldn't mind getting advice on, but we can talk after the meeting
13:27:29 <stefw> k
13:27:42 <andreasn> #topic Services rework
13:28:14 <mvollmer> alright
13:28:35 <mvollmer> I have been rewriting the old services page
13:28:37 <mvollmer> to directly talk to systemd
13:28:46 <mvollmer> this is still work in progress
13:29:04 <mvollmer> https://github.com/cockpit-project/cockpit/pull/2128
13:29:18 <andreasn> nice
13:29:25 <mvollmer> works well enough, and it's not much code at all at the end of the day
13:29:30 <stefw> cool
13:29:46 <mvollmer> stefw has given us a very nice architecture, it's a lot of fun
13:30:23 <mvollmer> the graphs are gone for now, but should come back after "the grid", I guess
13:30:28 <stefw> thanks. glad it's fun :)
13:30:33 <stefw> so did you move it into a component?
13:30:36 <stefw> or is that a seperate step?
13:30:39 <stefw> oh wait, i see it here
13:30:39 <mvollmer> yes, systemd/init
13:30:42 <stefw> nice
13:31:16 <stefw> we can add the graphs back later
13:31:26 <mvollmer> i am still not happy with the code structure, but, yes, it's not a lot of code, so...
13:31:31 <stefw> they need to also be conditional, and have a way of turning on and off the cgroup stuff that makes them work
13:31:36 <mvollmer> one open issue is whether or not to use DBusProxies for the long lists.
13:32:06 <mvollmer> i don't use them now, but I should try and if anything makes it problematic, we should try to fix that.
13:32:09 <stefw> DBusProxies works well with org.freedesktop.DBus.ObjectManager based services
13:32:30 <mvollmer> which systemd is not, but they should work well regardless
13:32:32 <stefw> so since systemd doesn't use ObjectManager, if DBusProies doesn't work
13:32:40 <stefw> then it wouldn't be *too* much of a surprise
13:33:14 <mvollmer> i am worried about excessive traffic, and the mysterious UnitNew/GetAll loop.
13:33:22 <stefw> indeed
13:33:29 <andreasn> what is "the grid"? Isn't that a TRON thing?
13:33:33 <stefw> since the systemd code forces us to do roundtrips
13:33:37 <stefw> andreasn, the first agenda point
13:33:50 <andreasn> ah, ok
13:34:24 <mvollmer> "the grid" is the metric channel/plot interfacing thing that stef is doing.
13:34:43 <andreasn> right
13:34:50 <andreasn> ok, next topic?
13:35:10 <andreasn> #topic Multi OS testing
13:35:22 * andreasn rushes along with the agenda
13:35:27 <andreasn> let me know if it's too rushed
13:35:34 <mvollmer> stefw, yeah, so we would call ListUnits to trigger the filling of the cache, which would flood the bus with 200 or so GetAll messages.
13:35:34 <mvollmer> yep
13:37:26 <mvollmer> it's fine, I guess.
13:37:31 <mvollmer> so, Multi OS testing.
13:37:44 <mvollmer> we started testing rhel 7 along with fedora 22 on github
13:37:58 <mvollmer> and everything is red until we figure out the issues.
13:38:07 <mvollmer> should happen soonish.
13:38:22 <mvollmer> https://github.com/cockpit-project/cockpit/pull/2122
13:38:25 <jscotka> So from my side, I'm working to have also test installing sources from mock. I have plan to rewrite all sigle host tests ASAP
13:38:32 <mvollmer> also, hubbot now checks master.
13:38:37 <mvollmer> dperpeet did that.
13:38:44 <stefw> yeah, that's cool
13:38:50 <stefw> i wish the status icons could be higher on the page
13:38:54 <mvollmer> results are in the cockpit readme.
13:38:57 <stefw> but probably at the top of README is the best we can do
13:39:00 <mvollmer> yes.
13:39:06 <stefw> mvollmer, why isn't https://github.com/cockpit-project/cockpit/pull/2122 green?
13:39:09 <mvollmer> we can have less files in the tree. :-)
13:39:23 <dperpeet> I haven't found any other way, except to look at http://files.cockpit-project.org/hubbot/
13:39:40 <mvollmer> stefw, I don't know.  It was green, but not any more.
13:39:55 <stefw> ah ok. good to know that it was green at one point though
13:40:13 <mvollmer> maybe I broke it with the recent update...
13:41:02 <mvollmer> jscotka, very good!
13:41:18 <mvollmer> hubbot doesn't run the avocado tests anymore, unfortunately.
13:41:30 <stefw> right, i think that's the next agenda item
13:41:33 <mvollmer> it got stuck and we didn't get around to make it unstuck...
13:41:42 <mvollmer> okay
13:41:43 <jscotka> mvollmer, np
13:41:44 <dperpeet> I can get on that soon-ish
13:41:58 <andreasn> should we move on to CI then?
13:42:33 <stefw> yup
13:42:44 <andreasn> #topic New CI migration progress
13:43:10 <jscotka> so few words
13:43:14 <stefw> So now that Fedora and RHEL stuff is mostly done, I was wondering how we can make progress on this?
13:44:05 <jscotka> migration to new CI is in most cases very easy, supporting libs are writtend for disc anc networks
13:46:15 <stefw> so i guess it'll happen by itself?
13:46:24 <stefw> on my end, when writing new kubernetes tests, i'll try and use the new arch
13:46:42 <stefw> but we need to figure out how to run them reliably, and make progress there
13:46:47 <jscotka> stefw, cool, perfect
13:46:48 <stefw> as well as migrate old tests over
13:47:11 <stefw> unfortunately we haven't reached the tipping point, where the work has begun to pay off
13:47:25 <stefw> we should aim to get there, i guess
13:47:25 <jscotka> yes, the problem is from my POV, that libvirt is sometimes unreliable in
13:48:54 <dperpeet> mvollmer, didn't we have failed tests that crashed the new CI?
13:49:05 <dperpeet> or rather suspended it
13:49:08 <mvollmer> We need make a small change to how we make VM snapshots
13:49:15 <dperpeet> ok
13:49:46 <mvollmer> no, avocado was hanging in the VMs.
13:50:02 <mvollmer> (the snapshot comment was general)
13:50:24 <jscotka> it sometimes happen, discussed with avocado devel, there is (will be added feature for remote timeouts)
13:50:44 <mvollmer> ahh, timeouts.... :-)
13:51:19 <jscotka> it is connected with troubles, that for example network can be gone down -> ssh connection is lost -> no results returned as an example
13:51:47 <dperpeet> well, let's look into when we have a more significant sample base (more tests)
13:52:08 <jscotka> https://github.com/avocado-framework/avocado/issues/513
13:52:18 <stefw> jscotka, what is the next test to be ported to the new CI?
13:53:00 <stefw> and mvollmer, dperpeet ... what do you think needs to happen before we can remove an old style test that was migrated to the new CI?
13:53:01 <dperpeet> can we put a comment in the old tests if they've been ported? that way there's a lower chance of losing changes when we delete old ones
13:53:04 <jscotka> hmm, it is up to us, do you have some suggetion, or I shoulduse alphabetival order :-)
13:53:20 <stefw> jscotka, i would suggest check-storage
13:53:34 <jscotka> okay, to test also new disc library inside
13:53:34 <stefw> as we discussed earlier, doing the hard ones first makes sense
13:53:38 <stefw> yes
13:54:13 <dperpeet> stefw, the new CI needs to get a clean track record of not dying on us during a significant amount of runs
13:54:23 <dperpeet> at least a week
13:54:31 <dperpeet> and/or have automatic recovery
13:54:57 <dperpeet> I can get on enabling the new tests this week
13:55:04 <dperpeet> and work with jscotka on that
13:55:10 <jscotka> yes probably add timeouts and as you mention some recovery (restart libvirt or similar)
13:55:12 <dperpeet> since I've familiarized myself with hubbot
13:55:53 <jscotka> dperpeet, We should do that on another server, not on recept hubbot, to not waste resources on hubbot and stuct actual resutl process
13:56:09 <dperpeet> jscotka, I think we should test it where it will be used
13:56:37 <dperpeet> then it can run alongside the other tests
13:57:02 <stefw> next topic?
13:57:12 <andreasn> sure
13:57:12 <jscotka> dperpeet, understant, but for me clean enviroment is better, because it also shows that there are some undocumented things, like (installing packages, what are on hubbot already installed)
13:57:23 <jscotka> yep, okay, next topic :-)
13:57:29 <andreasn> #topic Translations to Zanata
13:57:45 <andreasn> what's Zanata?
13:57:47 <stefw> so at some point, somebody setup cockpit with Transifex fedora infrastructure
13:57:52 <stefw> and now they want us to migrate to Zanata
13:58:00 <stefw> because somebody reinvented transifex i guess?
13:58:05 <stefw> and there's all these reminder emails about it.
13:58:51 <stefw> is anyone interested in working on the i18n situation in Cockpit in general, or perhaps just syncing translations with Zanata?
13:58:54 <andreasn> all of Fedora is moving to this?
13:58:58 <stefw> yes
13:59:08 <stefw> obviously we are not fedora
13:59:11 <petervo> transifex is sort of dying/dead
13:59:22 <stefw> but since someone put our stuff in transifex, people want it moved to zanata
13:59:32 <stefw> andreasn, https://fedoraproject.org/wiki/L10N_Move_To_Zanata
13:59:53 <petervo> i can take it as a low priority todo
14:00:17 <stefw> cool. i hope you have the necessary access
14:00:24 <petervo> i'm sure i don't
14:00:26 <stefw> if not, i'm not sure who does ... we'd have to hunt them down
14:00:31 <petervo> that will be half the work
14:00:36 <stefw> i don't see why i would either
14:00:39 <stefw> hence my confusion about this
14:01:09 <petervo> is fedora running their own?
14:01:09 <stefw> i see this: https://fedora.transifex.com/projects/p/cockpit/
14:01:22 <stefw> and we're all listed there
14:01:52 <stefw> oh look at that.
14:01:58 <stefw> somehow i accepted to be a maintainer
14:02:02 <stefw> okay i guess i must have access
14:02:09 <stefw> i don't remember this at all
14:02:10 <stefw> let me see
14:02:10 <andreasn> was it accepted for you maybe?
14:02:30 <dperpeet> sorry, gotta run
14:02:38 <stefw> dperpeet, o/
14:02:53 <andreasn> how does this workflow work? Do these become pull requests?
14:03:11 <stefw> i haven't done this in a while, but i imagine they do once they're synced over
14:04:40 <fche> AIUI projects -pull- from zanata via some commands
14:04:52 <stefw> right, and then the result is commited, right?
14:05:00 <stefw> at least that's how i did it in other projects with transifx
14:05:58 <github> [cockpit] stefwalter closed pull request #2132: Compilation test update (master...lib_compilation) http://git.io/vvObx
14:06:06 <andreasn> cool
14:06:13 <andreasn> anything else for this week?
14:06:22 <stefw> alright, i'll try to get access to Transifex again for the migration
14:07:33 <petervo> on a different topic, coverity has some issues with the storaged code
14:08:26 <petervo> is it worth spending time on fixing it up? or just leave it till we remove that code
14:08:35 <stefw> i think it's worth fixing
14:08:52 <petervo> ok
14:09:26 <stefw> but i guess mvollmer would be the one to ask
14:09:34 <stefw> depending on how long we're keeping the code
14:10:38 <mvollmer> yeah, I can't answer that either.
14:10:38 <mvollmer> personally, I would like to attack storage as the next big topic.
14:11:02 <mvollmer> after wrapping up metrics and services.
14:11:02 <mvollmer> and whetever else is ongoing.
14:11:23 <stefw> right, makes sense
14:12:02 <mvollmer> stefw, fyi, https://bugzilla.redhat.com/show_bug.cgi?id=1211287
14:12:08 <andreasn> it's 11 past, so ending meeting
14:12:10 <andreasn> #endmeeting