17:02:34 #startmeeting fedora_atomic_wg 17:02:34 Meeting started Wed Dec 14 17:02:34 2016 UTC. The chair is jberkus. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:02:34 Useful Commands: #action #agreed #halp #info #idea #link #topic. 17:02:34 The meeting name has been set to 'fedora_atomic_wg' 17:02:55 #topic Roll Call 17:02:59 .hello jberkus 17:03:00 jberkus: jberkus 'Josh Berkus' 17:03:02 .hello trishnag 17:03:03 trishnag: trishnag 'Trishna Guha' 17:03:06 .hello jasonbrooks 17:03:09 jbrooks: jasonbrooks 'Jason Brooks' 17:03:29 .hello maxamillion 17:03:30 maxamillion: maxamillion 'Adam Miller' 17:03:50 .hello dustymabe 17:03:51 dustymabe: dustymabe 'Dusty Mabe' 17:04:10 i currently am on two video calls and an IRC meeting 17:04:13 :) 17:04:32 .hello tflink 17:04:32 #chair dustymabe trishnag jbrooks maxamillion 17:04:32 Current chairs: dustymabe jberkus jbrooks maxamillion trishnag 17:04:33 tflink: tflink 'Tim Flink' 17:04:40 #chair dustymabe trishnag jbrooks maxamillion tflink 17:04:40 Current chairs: dustymabe jberkus jbrooks maxamillion tflink trishnag 17:04:43 .hello roshi 17:04:44 roshi: roshi 'Mike Ruckman' 17:04:49 #chair dustymabe trishnag jbrooks maxamillion tflink roshi 17:04:49 Current chairs: dustymabe jberkus jbrooks maxamillion roshi tflink trishnag 17:04:50 roshi!!!!!!!!!!!!! 17:04:58 roshi!!!!! 17:04:59 :D how's it going? 17:05:02 roshi: are you back??? 17:05:11 almost 17:05:16 roshi: oh man, so close ;) 17:05:21 you have been missed! 17:05:21 just checking in, and it was good timing for the meeting :) 17:05:35 glad to hear it :) I've missed you all as well 17:05:49 * jberkus doesn't know roshi 17:06:13 same here jberkus :) 17:06:17 roshi: we look forward to your return :) 17:06:21 jberkus: all you need to know is that roshi is awesome, he helps out with cloud/atomic WG and will be back soon 17:06:24 .hello bowlofeggs 17:06:25 hi roshi 17:06:25 bowlofeggs: bowlofeggs 'Randy Barlow' 17:06:30 #chair dustymabe trishnag jbrooks maxamillion tflink roshi bowlofeggs 17:06:30 Current chairs: bowlofeggs dustymabe jberkus jbrooks maxamillion roshi tflink trishnag 17:06:38 roshi is a long time awesome fedora community member who has been on sabatical 17:06:44 #info roshi is awesome, he helps out with cloud/atomic WG and will be back soon 17:06:58 .hello walters 17:06:59 walters: walters 'Colin Walters' 17:07:06 #topic FDLIBS 17:07:13 #chair dustymabe trishnag jbrooks maxamillion tflink roshi bowlofeggs walters 17:07:13 Current chairs: bowlofeggs dustymabe jberkus jbrooks maxamillion roshi tflink trishnag walters 17:07:23 hey jberkus and trishnag :) 17:07:36 I'm a QA guy that's been on a bit of a hiatus 17:07:37 I wanted to open by congratuating maxamillion and the team on getting the layered image build service launched 17:07:44 roshi: QA++ 17:07:52 we could use you 17:07:56 thanks for the kind words dustymabe maxamillion :D 17:07:59 jberkus: thanks :D 17:08:17 roshi: well deserved, looking forward to your triumphant return! :) 17:08:38 I also want to enlist the members of this group in (a) building containers for the service, and (b) taking a hard look at the guidelines/requirements 17:08:43 since we'll be enforcing them 17:09:11 maxamillion: anything else on FDLIBS? 17:09:31 I've got my first request in: https://bugzilla.redhat.com/show_bug.cgi?id=1404421 17:09:51 yeah maxamillion++ for getting the build service all the way to prod! 17:09:51 bowlofeggs: Karma for maxamillion changed to 9 (for the f25 release cycle): https://badges.fedoraproject.org/tags/cookie/any 17:09:58 yah, and matt submitted one too 17:10:23 we're going to need a process for approving contianers; maxamillion, is there currently a specified workflow for that? I'm not an RPM packager ... 17:11:19 :D 17:11:32 jberkus: yes, it was in the announcement email I sent out 17:11:59 well, what the announcement said was "same as RPMs" .... we'll need people to volunteer to take roles 17:12:07 maxamillion++ 17:12:08 trishnag: Karma for maxamillion changed to 10 (for the f25 release cycle): https://badges.fedoraproject.org/tags/cookie/any 17:12:44 \o/ 17:12:50 jberkus: what do you mean? 17:12:57 oh, I see ... I was just looking at the guidelines 17:13:02 there's another page for the review 17:13:11 once I get up to speed again, I'll vet containers 17:13:13 jberkus: correct 17:13:13 https://fedoraproject.org/wiki/Container:Review_Process 17:13:29 maxamillion: so we need people to be reviewers 17:13:40 maxamillion: what Fedora project perms does a reviewer need? 17:13:59 packager, iirc 17:14:11 how many of our group have that? 17:14:41 * misc has 17:14:53 roshi: yay 17:15:07 * maxamillion has 17:15:11 * roshi doesn't remember if he's a full packager or not 17:15:57 misc is helping me (thanks misc!) 17:16:27 maxamillion: there doesn't appear to be a separate queue for containers needing review 17:16:30 is that correct? 17:16:36 I'm also a packager sponsor so if anyone would like to become a packager, please feel free to ping me 17:16:54 roshi: if you don't have full packager, bowlofeggs can help 17:16:54 jberkus: correct, not at this time ... there's a not about this in the doc 17:16:57 he can sponsor 17:17:09 sgtm 17:17:31 maxamillion: hmmm ... realistically we're going to need a way to pull out the containers pending 17:17:40 right now there's 379 new tickets 17:18:35 jberkus: I don't disagree 17:18:45 is that a ticket already? 17:18:53 jberkus: it is not 17:19:04 ok, I'll open one ... which project? 17:19:27 we also need some automation, on what is gonna happen with the first approved package after the approval have been given ? 17:19:33 We could keep track of open container reviews in our pagure, maybe 17:19:51 (like I did all I could (except nitpicking on source code) for https://bugzilla.redhat.com/show_bug.cgi?id=1404421 ) 17:20:15 we still have internet faeries that do things for us, right? 17:20:17 is it really on this WG to gate all container reviews? 17:20:25 dustymabe: according to the current docs, yes 17:20:26 that's where I get all my automation :p 17:20:30 I don't see that as something that is sustainable 17:20:34 misc: what do you mean automation? 17:20:34 at all 17:20:45 misc: this should be a request to pkgdb just like for rpms, but in the docker namespace 17:20:49 or to nominate a group to do reviews 17:20:58 dustymabe: we'd probably have to have a "core" set of containers and that's it 17:21:02 maxamillion: that's a script that do that, no ? 17:21:07 misc: no 17:21:09 I think we should start it though so that we can work out the kinks in the process 17:21:17 and scale it with resources as time goes on 17:21:19 misc: to the best of my knowledge, it hasn't been for some time 17:21:25 yeah, i don't disagree that we should have a heavy hand in the beginning 17:21:38 but think of the reviews that are going to build up over time 17:21:44 dustymabe: no question 17:22:05 TBH I don't really have a lot of time for doing reviews of containers :( 17:22:29 unless they represent some sort of strategic path for us, i.e. the k8s container does 17:22:34 they are much faster than package review 17:22:47 like there is not enough guideline to take time :) 17:23:15 misc: but should you build the container and make sure it works? 17:23:31 yeah, eventually I can see asking Fedora to create a separate container-packager permission 17:23:31 dustymabe: mhh, technically yes 17:23:41 (but now, I didn't :/ ) 17:23:41 but not right now 17:24:14 just because I'm behind, can someone point me to notes when we decided that Fedora should be gating any containers in the first place? 17:24:27 maybe we can already say "let's see how it go with 5 containers" and see maintainance, the process, etc ? 17:24:33 k 17:24:36 I presume we're going to be running a docker registry or something? 17:24:41 roshi: http://www.projectatomic.io/blog/2016/12/announcing-fedora-layered-image-builds/ 17:24:49 roshi: already are 17:24:53 roshi: registry.fedoraproject.org 17:24:56 * roshi heads off to read 17:25:00 we are? 17:25:04 oh snap 17:25:07 did anyone read the announcement I wrote? 17:25:12 ;-) 17:25:17 that's some work in and of itself 17:25:36 like ... not to be rude, but seriously everyone has asked me a question so far that's in the announcement or linked documentation 17:25:40 ok, so let's follow up on this in email and tickets 17:25:43 * roshi is on sabbatical 17:25:52 roshi: right, totally fair 17:25:55 :p 17:26:00 roshi: it's been a busy six months 17:26:07 so it seems :) 17:26:27 please read your email; I have a proposal there for an alteration to the guidelines 17:26:45 #topic Kubenetes in Fedora Atomic 17:27:10 so, are we putting the kube packages back into FAH 25? 17:27:17 I think we should 17:27:38 I agree. What are the blockers on that? 17:27:47 I think we should put kubernetes, etcd and flannel back in 17:28:04 There are ways to run all of them in containers, but we don't have official fedora containers 17:28:13 jbrooks: +1 17:28:30 walters: are there technical blockers to putting them back into the tree? 17:28:34 jberkus: +1 17:28:38 errr 17:28:40 jbrooks: +1 17:28:42 brb 17:29:13 There shouldn't be 17:29:33 jberkus, but then when do we take them out again? 17:29:44 We could do it for f26 17:29:46 walters: 26, *if* we meet the requirements 17:30:03 that is, having another viable method of install, including documenation which covers migration 17:30:09 We should be able to have official containers for the missing components by then 17:30:09 okay 17:30:22 ok 17:30:29 has someone written anything regarding the testing of any of these things? (Sorry to be jumping all around) 17:30:38 Which things? 17:30:41 #action add kubernetes packages back into the base OStree for FAH 25 17:30:59 (things meaning the registry, layerd builds, standards for containers in the registry, etc) 17:31:09 #action jberkus to create tickets for prerequisites for removing kube packages from base ostree 17:31:38 roshi: not sure, take it up on #fedora-cloud after the meeting? 17:32:09 sure thing 17:32:49 * maxamillion is back 17:32:50 #topic ISO Images 17:33:05 https://pagure.io/atomic-wg/issue/185 17:33:48 one thing i'm uncertain about long term is how much fedora-based containers make sense for some of these infra components vs using centos 17:34:08 i understand our mindset coming from the distro world is "all software from one source" 17:34:33 walters, dustymabe: does it look like we'll be able to get the anaconda fix merged this week? 17:34:34 anyways we don't need to discuss it now 17:34:43 jberkus: it is merged upstream 17:34:43 walters: yeah, that's a big topic 17:34:47 and also into F25 17:34:54 jberkus, i'm a proven packager so I could do the build myself but... 17:34:55 waiting on a new rpm build 17:34:58 or, like, busybox-based containers 17:35:00 dustymabe: ah, so it's just waiting on the new 2-week release? 17:35:06 (aside: this whole "individual package fiefdoms" thing is also totally broken) 17:35:30 jberkus: basically we have to get the rpm into fedora 17:35:32 jberkus, before that someone needs to manually initiate a koji build and manually click the stuff in bodhi 17:35:40 and then the media should get built with the new anaconda 17:35:50 who's "someone"? 17:35:59 one of the anaconda maintainers for the first part 17:36:05 walters: I've asked them to do that 17:36:13 ok 17:36:14 I believe they are going to 17:36:15 anyways i've already typed more here in irc than would be required to do the work 17:36:24 dustymabe: can you follow up in a couple days if there's no update? 17:36:39 jberkus: yeah, this is something I'd like in the next two week 17:36:55 which is next week FYI 17:36:59 if we can make it 17:37:03 #action dustymabe to follow-up on merge of UEFI patch into anaconda 17:37:33 speaking of which, walters is it realistic to get kube added back in for the same release? 17:38:11 jberkus: to be clear, we are talking about adding it back in *only* for Fedora 25 17:38:30 so we would not be adding it in to F26, we would leave that like it currently is 17:38:32 right? 17:38:37 dustymabe: correct 17:38:58 jberkus, keep in mind that everything we're talking about in fedora for Atomic Host releases actually doesn't make sense https://pagure.io/releng/issue/6545 17:39:10 specifically in this instance if I change the treefile json, it will ship the next day 17:39:56 walters: indeed 17:40:02 that is something we want to get fixed 17:40:03 anyways, that's probably OK 17:40:14 walters: well, that's a plus when it comes to pushing changes fast 17:40:17 ;-) 17:40:27 once a day isn't fast =) 17:40:43 walters: but yeah, want to have the discussion around a different release process for first meeting of next year? 17:40:55 walters: realistically, nothing is going to get done next week 17:41:08 jberkus: yeah, that discussion is on me 17:41:40 i've got some work to do with releng folks, and then I'm sending out a request for review of the proposed changes 17:41:53 #action dustymabe, walters to discuss new release process at beginning of calendar year 2017 17:42:23 #topic other issues 17:42:39 New PRD: I'm not finished yet, watch your email 17:43:29 other critical issues 17:43:42 Website: are the cloud images available again? 17:44:03 jberkus: not sure what that last item is? 17:44:10 can you rephrase your question 17:45:06 ah 17:45:18 so, in the changed to getfedora.org, the cloud base images vanished 17:45:26 I don't see them 17:45:27 they're not on Atomic and they're not on Server 17:45:27 https://getfedora.org/en/atomic/download/ 17:45:30 whoops ... /part'd the wrong chan 17:45:43 and the planned changes for alt.fedoraproject.org for 25 didn't happen 17:45:47 so they're not there either 17:45:58 well cloud images are not even here now https://alt.fedoraproject.org/ 17:46:04 here is the ticket: https://pagure.io/fedora-websites/issue/57 17:46:13 been waiting on the websites team 17:46:28 but it's not their fault that we were bad at communicating the change to them 17:46:30 hmmm 17:46:33 so we have to patiently wait 17:46:46 however, I actually foud a new contributor here in the NYC office that is working on it 17:46:56 dustymabe: oh, that's good 17:47:12 so hoping she can get things working, i'm going to sit down with her later today 17:47:22 ok 17:47:42 dustymabe: FYI, the websites team showed me a plan for alt.fedoraproject.org which was going to have all of the downloads in a searchable grid 17:47:58 jberkus: that's nice 17:48:04 which was the reason I didn't see removing the cloud base images as critical 17:48:23 dustymabe: might be nice to find out what happened to that 17:48:49 jberkus: yeah, no idea 17:49:31 added to issue 17:50:57 anything to discuss with fedora-dockerfiles other than needing to start work on porting all of those to FDLIBS? 17:51:05 https://pagure.io/atomic-wg/issue/180 17:51:20 * jberkus can port containers, since he can't review them 17:51:43 jberkus: you can work on getting *packager* status 17:51:52 it's not too hard, bowlofeggs helped me out quite a bit 17:53:21 ok, 17:53:25 #topic open floor 17:53:29 anything else? 17:54:18 jberkus: overlayfs 17:54:20 * roshi has nothing :) 17:54:25 dustymabe: go ahead 17:54:26 should we propose that all of fedora move to overlayfs? 17:54:38 dustymabe: for what specifically? 17:54:38 roshi has 6+ months of catch-up 17:54:40 * walters votes yes 17:54:53 dustymabe: what's the drawback? 17:54:53 yeah, I have a lot of reading to do... 17:55:03 walters: yeah I'd like that too, the "don't store data in containers" thing does scare me a bit 17:55:06 roshi: feel free to bug me or any of us 17:55:14 roshi: we really need you doing QA 17:55:30 dustymabe: explain? 17:55:34 oh I will :) I'll probably surpass the annoying threshold :p 17:56:09 jberkus: see ML: https://lists.fedoraproject.org/archives/list/cloud@lists.fedoraproject.org/message/SF4GGOU2NHLQRDEBFII3YRT3WYU24RR3/ 17:56:20 the conversation in that mail should give some perspective 17:56:37 roshi: consider me annoyed already :) 17:56:53 lol 17:57:19 so especially if we get that patch upstream in (farther down the ML thread), then I say we merge it and see what other issues we find 17:57:21 dustymabe: so stuff errors out if you try to write to files inside the container? 17:58:13 dustymabe: I think this is an excellent thing to run by our userbase 17:58:29 publicize the tradeoff, get feedback, get people to test real workloads 17:58:29 jberkus: it can based on that one issue 17:58:43 jberkus: yeah I think that's what I'll mention in the blog post 17:58:57 * dustymabe is going to do a blog post on overlayfs on F25, showing people how to set it up 17:59:01 dustymabe: that also potentially adds a requirement to FDLIBS 17:59:16 jberkus: why? 17:59:23 that containers need to call out all writeable dirs as VOLUMES in the container def 17:59:31 *all* of them 18:00:04 is that true though? or just that they should call out the ones they care about? 18:00:12 I don't know 18:00:14 walters: ^^ 18:00:33 dustymabe: "doesn't fully support POSIX sematics" suggests to me that there could be fatal errors 18:00:49 at runtime 18:01:06 like, I don't care about Postgres' tmpstats dir 18:01:10 but I do care if it errors out 18:01:22 ok 18:01:27 jberkus: yeah, I think we need to understand the problem space better 18:01:38 #action dustymabe walters jberkus to explore overlayfs writeability issue 18:01:47 #action dustymabe to blog using overlayfs in Atomic 18:01:58 ok, we're out of time 18:02:08 take further discussion on #fedora-cloud and the mailing lists 18:02:13 #endmeeting