16:00:47 <jbrooks> #startmeeting CentOS Atomic SIG
16:00:47 <centbot> Meeting started Thu Apr 20 16:00:47 2017 UTC.  The chair is jbrooks. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:47 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:00:53 <jbrooks> hey kbsingh
16:01:32 <jbrooks> #chair jberkus kbsingh jlebon
16:01:32 <centbot> Current chairs: jberkus jbrooks jlebon kbsingh
16:02:17 <jbrooks> #topic downstream release
16:02:30 <jbrooks> There was a new upstream release last week
16:02:48 <jbrooks> And as of yesterday, we have all of the rpms built for centos
16:03:05 <jbrooks> I did a test compose at https://ci.centos.org/artifacts/sig-atomic/downstream/
16:03:24 <jbrooks> It looks good, I'll test a bit more today, I also need to rev the version number in git
16:04:25 <jlebon> nice
16:04:49 <jberkus> so, given that the last release was very late, what's the schedule for this one?
16:04:59 <jberkus> are we going to try to get back on schedule, or just skip a release?
16:05:34 <jbrooks> I'd like to release this one, too
16:05:43 <jbrooks> There are monthly media refreshes for centos
16:06:05 <jbrooks> We've long gone back and forth about whether to update atomic on that schedule or not
16:06:26 <jbrooks> I'd like to see us have a monthly ref and a ref that includes the most recent rpms in the repos
16:06:51 <jbrooks> as described https://github.com/CentOS/sig-atomic-buildscripts/issues/230#issuecomment-280161658
16:07:24 <jbrooks> kbsingh, Any update on the automated release stuff?
16:08:19 <kbsingh> we have a bunch of the things in place
16:08:29 <kbsingh> the ostree repo is now building nightly
16:09:10 <jbrooks> Is it in an accessible place?
16:09:18 <jbrooks> And is it also being signed?
16:09:20 <kbsingh> not being pushed yet
16:10:33 <jbrooks> What's our next step on it?
16:11:43 <jberkus> kbsingh: ?
16:13:58 <jbrooks> OK, jberkus shall we move to talk of goals?
16:14:08 <jberkus> sure
16:14:27 <jbrooks> #topic project goals
16:15:24 <jberkus> let's start with the big question: is Centos Atomic a project we actually expect users for, or is it something we're doing just because RHEL is?
16:15:46 <jberkus> that is, is CAH supposed to be an externally-facing project?
16:16:02 <jbrooks> I definitely expect users for it
16:16:25 <jbrooks> I think in, say, #atomic, we very or more often see ppl asking about centos atomic vs fedora
16:16:44 <jbrooks> In the same way that ppl want to run regular centos
16:17:11 <jbrooks> But, cah is much less of a focus in centos than fah is in fedora
16:18:17 <jberkus> jlebon, walters, kbsingh?  comments?
16:20:24 <kbsingh> sorry guys, line dropped here
16:20:43 <kbsingh> jbrooks: so going back a bit there are 2things that need to get done before the nightly tree's are visible
16:20:55 <kbsingh> (1) there needs to be a process to get from the build host to the buildlogs host
16:20:58 <jlebon> jberkus: yeah, i think it definitely has its place, for the same reason that non-AH centos exists. it's a slower/more stable FAH
16:21:39 <kbsingh> (2) we need a way to trigger the builds, either via ci
16:22:39 <jbrooks> kbsingh, For one, can we rsync once a build is finished?
16:22:39 <kbsingh> for the second aprt - re: attracion of cah - we have ~ 40k IPs that pickup updates for it
16:23:54 <bstinson> jbrooks, kbsingh: putting aside what we have in terms of infra right now in terms of building CAH. what's the ideal process here?
16:25:03 <kbsingh> its not grown a lot. the initial graph was pretty good, but the growth is ~ 100usr/week sort of ball partk at the moment ( just looked at last 6 weeks )
16:25:11 <jbrooks> bstinson, we compose trees as new rpms become available, and make that tree available in an ostree ref for ppl who want to run it, and then we cut a monthly update ref and media each month
16:25:44 <kbsingh> bstinson: the buildscript we have in git has been very helpful to get people bootstrapped up - i think there are about as many users who build their own, as ones who use our builds
16:26:15 <bstinson> any need to make that faster? or do storage in a different way to make the process less painful?
16:26:40 <jbrooks> There's a lot that could happen around promotion of cah, but we need to nail down the basics
16:26:56 <jberkus> like doing releases?
16:27:04 <bstinson> (i'm coming at this considering the CAHC case, and the potential for other CI tenants that need to build/compose hosts)
16:27:39 <jbrooks> What slows us down is the "official release" stuff
16:27:57 <jbrooks> the signing, the placement on the official release servers
16:28:12 <jberkus> bstinson: CAHC?
16:28:27 <jbrooks> continuous
16:28:28 <bstinson> jberkus: continuous
16:28:32 <jberkus> ah, ok
16:29:31 <jbrooks> continuous isn't signed, and it is distributed from the ci artifacts location
16:29:41 <jbrooks> If we did downstream that way, we'd be much faster
16:29:46 <jbrooks> But less legit
16:30:26 <jbrooks> jberkus, The biggest thing is that we have a very limited contributor group for cah
16:30:56 <jbrooks> And where fedora has a releng team to do releases, we don't
16:31:32 <jberkus> sure, but this starts from goals
16:31:44 <jberkus> that is, do we want CAH to be a thing?  do we want adoption?
16:32:18 <jberkus> because so far the only person to say "yes" enthusiastically has been jbrooks
16:34:43 <jberkus> ... I'll take that as a "no" from everyone else.
16:36:25 <kbsingh> no to what ?
16:36:28 <jbrooks> jberkus, What about you, do you think we should just drop centos atomic?
16:36:53 <jbrooks> kbsingh, He just asked if any of us, other than me, think cah should even exist
16:37:02 <kbsingh> i think it should
16:37:22 <kbsingh> the centos userbase wont touch the fedora builds, we've already established that
16:37:36 <kbsingh> so the question becomes do we care about the centos userbase into this process / technology ?
16:38:10 <kbsingh> my own usercase is pretty limited to being a host for openshift
16:38:22 <kbsingh> but we can switch that out to centos linux proper
16:38:43 <jberkus> yes.  it's not whether CAH is going to get built, it's about whether it's just "there" and we build-and-forget
16:39:22 <jbrooks> That's sort of how centos is -- it gets built, people use it
16:40:12 <kbsingh> jberkus: as oposed to?
16:40:38 <jberkus> jbrooks: yeah, but if a CentOS main release was 6 weeks behind, that would be the cause of some panic, no?
16:41:26 <jberkus> kbsingh: as opposed to trying to promote it and build a sub-community within CentOS around CAH.
16:41:36 <kbsingh> been there, do e that:)
16:41:50 <kbsingh> jberkus: sure; am all for it.
16:42:05 <kbsingh> so is the only real question around speed of release?
16:42:07 <jbrooks> jberkus, I do think that more ppl use regular centos + docker than use atomic
16:42:25 <jberkus> yah, there's also the container pipeline
16:42:50 <jberkus> kbsingh: well, regular releases are a baseline for an OS.  if you don't have that, you don't have anything
16:43:33 <kbsingh> agree
16:45:00 <jberkus> aside from that, though, if CAH is going to be a one-person project, we're not going to do anything more than just make releases
16:45:49 <kbsingh> if we want to line up with the upstream releases, then we need to hit the 5 weekly cadence
16:46:06 <jberkus> 5 week?  what's that from?
16:46:13 <kbsingh> 6
16:46:18 <kbsingh> sorry, am on my phone here
16:46:33 <kbsingh> to get to 6 weekly releases, with a 1 week offset is a good target to aim for
16:47:27 <jbrooks> If we do the two refs, the rolling one and the monthly one, then users can either track upstream by ~1 week behind, or they can just go monthly
16:47:41 <jbrooks> And rebase between the two
16:47:45 <jbrooks> if desired
16:48:30 <kbsingh> if we are ok to go from mirror.c.o to buildlogs.c.o ( i think we decided that was ok ) we can add a third nightly tag as well
16:48:48 <kbsingh> and we can cut media at some specific interval and call that a release
16:49:12 <jbrooks> Well I think nightly can basically be the updates ref -- just all the updated rpms currently avail in the repos
16:49:40 <kbsingh> this is all doable, but i dont have any time between now and 8th of may
16:50:09 <kbsingh> however, this is just all up on github, so anyone can do the wor
16:50:10 <kbsingh> k
16:50:30 <kbsingh> and 1 of us can setup the cron at the right place to run it push it
16:50:59 <jberkus> kbsingh: don't you have to personally sign everything, though?
16:52:20 <kbsingh> no we have a workaround for that now
16:53:06 <kbsingh> we have a key specific to the ostree repo, that isnt used for rpms etc, that can be planted in the same place as the cron
16:53:38 <walters> kbsingh, i think part of the FAH messaging should be running CentOS containers
16:53:55 <walters> that's actually a major part of the idea
16:54:29 <walters> when we talk about "centos userbase"
16:54:31 <kbsingh> you'll get the fedora userbase with that, but not much in the centos space.
16:54:49 <kbsingh> we can try it :)
16:55:08 <jberkus> there's also potentially fedora-containers-on-CentOS for upstreams which aren't available in CentOS yet
16:55:23 <walters> yes
16:55:45 <walters> anyways i'm not quite in favor of turning off CAH myself, at least not right now
16:56:32 <jberkus> but ... is anyone going to work on advocating CAH or container pipeline other than jbrooks?
16:57:08 <kbsingh> i am always happy to help whenever i can, with as much time i can
16:57:20 <jberkus> there's several things we could be doing to build up CAH and CP, but there doesn't seem to be anyone who wants to/can do them
16:57:30 <walters> jberkus, it's linked prominently from the website right now
16:57:38 <jberkus> which?
16:57:42 <walters> http://www.projectatomic.io/download/
16:59:32 <walters> kbsingh, as far as use cases, I'll say personally I'm never going to use yum on a baremetal machine ever again
16:59:36 <kbsingh> is there a list of these things todo ?
16:59:38 <walters> now that rpm-ostree has package layering
16:59:42 <jberkus> walters: yeah, but I'm about to switch places for CAH and FAH again
16:59:52 <walters> and I definitely see usecases for non-OpenShift hosts
17:00:04 <walters> rpm-ostree is quite good for "bare metal pets/elephants"
17:00:05 <kbsingh> walters: i think thats ok - its going to take a while for that mind set to perculate through to the user base
17:00:13 <walters> 100% agreed on that =)
17:00:16 <jberkus> kbsingh: well, that's goal-setting.  I have ideas and opportunities, but I'm an advisor.
17:00:20 <kbsingh> walters: eg. less than 5% of the centos userbase currently uses docker, in any shape
17:00:52 <kbsingh> from my side - the best thing we can do is demonstrate the wins from this process, into that other 95%
17:01:20 <walters> jberkus, i wouldn't disagree with doing that
17:01:55 <jberkus> for example: educating the CentOS community about Atomic and how it works; promoting the container pipeline and getting community contributions; working with the IoT team to promote Atomic for IoT, including alternate-arch builds
17:02:59 <jbrooks> We'll need to attract more contributors
17:04:06 <jbrooks> The biggest difference I see between fedora atomic and centos atomic, activitywise, is more RHers are engaged w/ fedora
17:04:10 <kbsingh> yeah
17:05:53 <jbrooks> All right, we're past the top of the hour
17:05:54 <jberkus> jbrooks: and far more Fedora people are engaged with Atomic, as a % of Fedora people
17:06:25 <jbrooks> jberkus, I guess?
17:06:40 <jbrooks> I know it's not 100% RH, but it isn't far
17:07:28 <jberkus> ok, if we're ending the meeting, how can I follow up on this without waiting until next week?
17:08:02 <jbrooks> jberkus, Let's get straight on what remains unanswered
17:08:10 <jbrooks> We agree that cah should be a thing
17:08:25 <jbrooks> We agree that we need to nail regular releases
17:09:02 <kbsingh> yeah
17:09:05 <jbrooks> KB identified the next things that need to be done on the release process -- I'll follow up on those items
17:09:13 <kbsingh> lets quantify what that means -  what is a release ?
17:09:15 <jberkus> Right.  And then we need to set goals after releases.
17:09:37 <walters> isn't a release at least a CAH tree commit (and potentially media) after core+extras updates?
17:10:00 <jbrooks> Yeah
17:10:09 <jbrooks> That's how I see it
17:10:59 <jbrooks> jberkus, I think we stand a better chance of attracting people if we work from the project atomic level -- promote things for fedora and centos at once, and just ensure that those projects have timely releases
17:11:19 <jbrooks> All the technology and patterns is the same
17:11:24 <p3ck> bstinson, is it possible to provision an atomic-centos host in duffy?
17:12:00 <kbsingh> so how often do we want media
17:12:10 <kbsingh> and what media do we want
17:12:41 <jberkus> jbrooks: yah, probably.  We also need an answer to "I want to work on CAH, how do I get started?"
17:12:59 <jbrooks> jberkus, like, as a contributor?
17:12:59 <bstinson> p3ck: we have the kickstarts, but we haven't flipped a switch to enable it yet. would that be helpful to you? (taking this to PM so the meeting can continue)
17:13:01 <jberkus> kbsingh: I thought the idea was to have media every 6 weeks?
17:13:21 <jberkus> jbrooks: yes
17:13:22 <jbrooks> This is why I want the second ref
17:13:52 <jbrooks> jberkus, I can writre something up for that -- how to get started contributing to cah
17:14:10 <walters> jbrooks, hm...CAH or CAHC?
17:14:22 <walters> given CAH is defined as a rebuild (right?) what would it mean to contribute?
17:14:33 <walters> contribute to the releng/test infra?
17:14:39 <p3ck> bstinson, well.. I suppose I could do it in libvirt for now
17:14:39 <jbrooks> Well, there are more cloud providers to get on
17:14:42 <jbrooks> for instance
17:14:53 <walters> yeah...that's a whole topic
17:15:06 <kbsingh> yea
17:15:15 <kbsingh> brb, switching to machine
17:15:16 <jberkus> docs (which would be me), container images, automation for stacks, help with porting to other architectures ...
17:15:19 <p3ck> bstinson, thinking about my use case more I think libvirt would work better for me anyway.
17:15:45 <jbrooks> Now, container images, that's not currently part of atomic sig, and I don't know if that's part of any sig?
17:15:49 <jbrooks> is it, kbsingh ?
17:16:14 <jbrooks> on the fedora side, the atomic wg say attn to that
17:16:20 <jbrooks> pays attn
17:16:56 <kbsingh> its not atomic per se
17:16:59 <jberkus> yeah, and I'd say that the CP is really the place were we could build up some community activity AND some crossover from the other AHs
17:18:19 <jbrooks> containerized apps is the app side of atomic -- that's how project atomic positions it
17:18:26 <jbrooks> Anywayu
17:19:03 <jbrooks> I could see cah just sort of melting into the core sig or whoever produces the regular centos -- just throw it on to the pile
17:19:32 <jbrooks> Because, like jberkus is sort of talking about, fedora atomic wg does more stuff here
17:19:39 <kbsingh> yeah
17:20:16 <jbrooks> I've been working w/ fedora and centos on this -- it's very similar, I just want both to exist and be healthy
17:21:38 <jbrooks> OK, we should close for this week
17:21:45 <jbrooks> #endmeeting