17:00:11 <tardyp> #startmeeting weekly
17:00:11 <bb-supy> Meeting started Tue Jan  8 17:00:11 2019 UTC and is due to finish in 60 minutes.  The chair is tardyp. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:11 <bb-supy> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:00:11 <bb-supy> The meeting name has been set to 'weekly'
17:00:36 <tardyp> #topic introductions
17:00:47 <p12tic> hi!
17:00:52 <rjarry> hello
17:00:55 <tardyp> hi all, who is there
17:01:01 <tardyp> meeting agenda: http://bit.ly/2rup31x
17:01:05 * rjarry is here
17:02:14 <tardyp> djmitche is landing at this exact moment so he won't attend today
17:02:36 <tardyp> #topic week in review
17:02:44 <tardyp> two big topics this week
17:03:25 <tardyp> 2.7 deprecation. rodrigc has been sending lots of cleanups PR which start to disable some tests that are only python2.7, and do python3 related cleanups
17:03:44 <tardyp> There is a big PR waiting that is completly removing py2 support
17:04:03 <tardyp> and is removing all the from __future__ stuff, and some of the compat stuff
17:04:12 <skelly> I have one question related to that: any opposition to adding them all to a 2.0 milestone to keep track of them?
17:04:26 <skelly> it's been sort of hard for me to keep track which ones are related to removing py2
17:04:45 <tardyp> basically anything that is sending rodrigc :)
17:04:53 <skelly> true
17:05:39 <tardyp> Anyway, I will release 1.7.1 soon, and then we can say next release is february, so this will be 2.0
17:05:45 <tardyp> and we can merge the big PR
17:05:57 <p12tic> how soon is soon?
17:06:00 <skelly> there's a lock issue that would be good to get into 1.7.1
17:06:10 <skelly> or 1.7.2 I suppose
17:06:23 <tardyp> oh ok. if we need that fix
17:06:38 <tardyp> I wanted to do it this week end
17:06:50 <tardyp> I'll merge the kubernetes stuff for 2.0
17:07:01 <tardyp> that is the second topic.
17:07:07 <skelly> https://github.com/buildbot/buildbot/issues/4491 is the lock issue
17:07:47 <tardyp> p12tic you were mentionning you might be responsible for this regression
17:07:51 <tardyp> do you plan to work on it?
17:08:50 <p12tic> yes
17:09:01 <tardyp> how bad you think it is?
17:09:10 <tardyp> what is the culprit patch?
17:09:59 <p12tic> sorry i've just got a call i needed to take
17:10:29 <bdbaddog> Here
17:10:34 <tardyp> hi bdbaddog
17:10:37 <p12tic> i think it's not really really bad
17:11:34 <p12tic> but i didn't investigate really deeply
17:11:49 <p12tic> i will allocate time on this weekend
17:11:51 <tardyp> https://github.com/buildbot/buildbot/commit/020db4f30cea4497a824c8f8536b40f2944304b1
17:11:59 <tardyp> this is the commit, right?
17:12:11 <p12tic> yes
17:12:38 <tardyp> ok
17:12:54 <tardyp> So second big topic is kubernetes
17:13:00 <tardyp> #topic kubernetes latent worker
17:13:17 <tardyp> We need a replacement for hyper.sh which is shutting down by 15 Jan
17:13:37 <tardyp> I've worked on finishing the PR from Zempashi
17:13:50 <tardyp> things are good so far, and the worker looks like working
17:14:05 <tardyp> I still need to make some more coverage
17:14:31 <tardyp> then I guess we can just replace the hyper workers by kubernetes worker
17:14:46 <tardyp> and probably we will have a few instability to fix over time.
17:15:17 <tardyp> bdbaddog has asked the conservancy if we can have an official GCP account from the conservency
17:15:42 <tardyp> waiting for that, I have created one and got the $300 credits which should work us for the next 10 months
17:16:10 <tardyp> if anybody want to get access to that cluster, I can add you as a developer, it is pretty easy
17:17:21 <tardyp> any question on that topic? or remark/advice?
17:18:04 <tardyp> ok.
17:18:12 <tardyp> #topic discourse instance
17:18:16 <bdbaddog> Is it reasonable to start out trying the preemptable workers first and see if that works?
17:18:29 <tardyp> #undo
17:18:29 <bb-supy> Removing item from minutes: <ircmeeting.items.Topic object at 0x807014650>
17:18:39 <tardyp> preemptable workers?
17:19:15 <tardyp> you mean using premptible VMs for kubernetes nodes?
17:19:15 <bdbaddog> "We can try to decrease that using premptive VM, but I am not sure what will be the reliability. We can afford a few job retry however."
17:19:19 <bdbaddog> yes.
17:19:42 <tardyp> right. I was looking at GKE yesterday, and actually it should work reliably.
17:20:00 <tardyp> the GKE console says that the premptible nodes will only survive for 24hours
17:20:08 <tardyp> that looks perfectly fine for us.
17:20:22 <tardyp> so that is cool as this decreases the cost by 4
17:20:33 <bdbaddog> indeed.
17:20:47 <tardyp> Also what I think I will do is to use higher spec VMs line 8 Cpus
17:20:53 <tardyp> and run several containers in them
17:20:57 <bdbaddog> It it turns out to have too many issues then can change?
17:21:05 <tardyp> this will allow to share the download of container image time
17:21:19 <tardyp> yes we can easily change the node pool configuration
17:21:22 <bdbaddog> Also is it possible to set up latent worker such that it tries preemptable and if that doesn't comlete in some time to go for regular vm?
17:21:46 <tardyp> not really.
17:22:06 <tardyp> actually this is managed automatically by GKE using their node pool concept.
17:22:41 <tardyp> then the worker will ask for a container which takes one CPU and xGB of memory, and kubernetes will automatically use an existing VM or spawn a new one
17:22:47 <tardyp> from one of its pool
17:23:26 <tardyp> if the worker is killed in the middle of a build because of VM crash or shutdown, then buildbot will automatically retry the build
17:23:47 <tardyp> and a new container will be sent to another VM.
17:24:01 <tardyp> I think this should work without too many issue
17:24:02 <bdbaddog> gotcha.
17:24:17 <tardyp> minus the eventual bugs.
17:25:00 <tardyp> #topic discourse instance
17:25:09 <tardyp> skelly: still wanting to work on this?
17:25:22 <tardyp> shall we keep it in the agenda?
17:26:57 <tardyp> #topic debian package
17:27:11 <tardyp> rjarry is working on super duper package for debian 10
17:27:29 <tardyp> with automatic non upgrade from buildbot 0.8 :)
17:28:16 <tardyp> but automatic upgrade from 1.x and that is cool.
17:28:27 <tardyp> anything else?
17:28:49 <tardyp> guess not.
17:28:51 <tardyp> #endmeeting