17:00:59 #startmeeting RELENG (2018-03-01) 17:00:59 Meeting started Thu Mar 1 17:00:59 2018 UTC. The chair is mboddu. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:00:59 Useful Commands: #action #agreed #halp #info #idea #link #topic. 17:00:59 The meeting name has been set to 'releng_(2018-03-01)' 17:01:00 #meetingname releng 17:01:00 #chair dgilmore nirik tyll sharkcz masta pbrobinson pingou puiterwijk maxamillion mboddu Kellin 17:01:00 #topic init process 17:01:00 The meeting name has been set to 'releng' 17:01:00 Current chairs: Kellin dgilmore masta maxamillion mboddu nirik pbrobinson pingou puiterwijk sharkcz tyll 17:02:00 * nirik is barely here. 17:02:41 hi 17:02:43 .hello kellin 17:02:44 Kellin: kellin 'Kellin' 17:04:15 .hello2 17:04:16 dustymabe: dustymabe 'Dusty Mabe' 17:05:04 Okay, lets get started 17:05:16 #topic #7301 Remove non-namespaced symlinks to rpms/ 17:05:25 #link https://pagure.io/releng/issue/7301 17:06:10 This is just a cleanup, we should do it. ;) 17:06:16 * puiterwijk is here and in favor 17:06:22 (which is why I suggested it :) ) 17:06:24 I am +1 to this change 17:06:46 puiterwijk: What needs to be done and who should do it, is my question? 17:06:55 The next steps if it were approved would be: 1. write up a short email to devel-announce, and 2. remove the symlinks 17:07:01 I'm fine with doing both 17:07:19 +1 on the change, and puiterwijk dropping a line to devel 17:07:44 puiterwijk: Okay, just to make sure, if people coming from EOL'd versions which has old url, they will get a warning and once adjusted everything should work fine, right? 17:07:55 mboddu: yes 17:08:09 puiterwijk: Okay, thats all I need to know 17:08:54 Also, do we need anyone else approval before I say we should do it now? 17:09:01 I don't think so. 17:09:07 Okay, thanks for the info 17:09:08 I'd say that if releng agrees with it, we're good 17:09:30 If people still have a fedpkg old enough that it won't work... It's time to update 17:10:18 Well, sending an email to devel@ will ping the people who wants to talk about it, so we are good 17:10:54 #info puiterwijk will send an email to devel@ about the change and once its done, he will work on removing the symlinks 17:11:27 * dgilmore has never updated old checkouts 17:11:34 I suspect most people have not 17:11:53 #info people coming from old fedora versions will get a warning about the url and once its updated everything should work fine 17:12:15 mboddu: get a warning how? 17:12:27 i use "git pull --rebase" 17:12:32 that will just be broken 17:12:49 unless I am missing what the topic is 17:12:57 [puiterwijk@workstation koji (master % u=)]$ git pull 17:12:58 * dgilmore assumes distgit 17:12:59 WARNING: 'koji' is an alias for 'rpms/koji' 17:13:01 Already up to date. 17:13:12 dgilmore: ^ we have been sending a server-side warning for a long while now 17:13:17 puiterwijk: okay 17:13:47 puiterwijk: and that warning will change to what when the symlinks are removed? 17:14:18 puiterwijk: I guess we just need to make sure we send plenty of warning to devel-announce witha cut off date 17:14:35 dgilmore: the remote warning wil ldisappear and say "could not find". "fedpkg pull" will tell the git set-url command 17:14:45 And yes, I was planning to do devel-announce with cut-off date 17:14:50 puiterwijk: that will be bad 17:15:02 I doubt people will think to switch to fedpkg pull 17:15:08 needs communication 17:15:13 and lots of it 17:16:15 anyway carry on 17:16:17 Note that it's been >1 year. So only things that people cloned before that would stop working 17:16:18 * mboddu uses fedpkg pull, but I understand people using git 17:16:36 * puiterwijk uses git pull 17:17:18 puiterwijk: Make sure your cut-off date is good enough and probably few emails between now and then 17:17:29 we have made a simililar change in the past 17:17:38 a script was provided to update repos 17:18:22 dgilmore: right now we have "fedpkg pull" which will tell you what to update things to 17:19:14 Anyway, conclusion: needs prior announcement and needs cut-off date. 17:19:26 Unless anyone has further issues to bring up, I think I can manage that 17:20:10 puiterwijk: +1 17:22:02 dgilmore: Your last statement seems incomplete, do you want to add anything before I move to next topic? 17:23:49 mboddu: lets move on 17:23:49 Okay, lets move on 17:23:52 :) 17:23:57 dustymabe: haha 17:24:53 #topic #7293 Pull Requests for fedora-release need discussion in larger group 17:24:59 #link https://pagure.io/releng/issue/7293 17:25:38 Kellin: Can you start on this, not sure what are your expectations 17:27:12 mboddu: mostly to find out what if any impacts these requests have, and whether it's going to break anything else 17:27:39 puiterwijk: https://pagure.io/fedpkg/blob/11f99d50f5d84f3c7bfefe703c635de358a35517/f/src/pyfedpkg/__init__.py#_92 17:28:19 Kellin: https://pagure.io/fedora-release/pull-request/119 is not true atm, as the repo is upstream for fedora-release, unless we decide to use dist-git 17:28:45 mboddu: correct 17:28:46 main issue will be people mergeing in or commit things that break machines, installs composes etc 17:29:09 mboddu: there was a proposal to do that... close the pagure.io repo and only maintain it in src.fedoraproject.org 17:29:37 nirik: Yes, I remember, but not sure whether we decided anything on that 17:29:49 I don't see a reason to do it 17:30:00 I don't know that we did... 17:30:00 I would be for it, if we had sufficent test suites 17:30:15 * nirik doesn't care much either way. 17:30:46 * mboddu is okay with anything, but I like to have things separate as upstream and fedora, but eh 17:30:53 it wil simplify the process at a cost of a loss of control. 17:31:26 dgilmore: exactly 17:31:31 I am pretty sure that I made my thoughts clear on devel list 17:32:47 https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/message/CIKK4WEF5ACVWZ6EWBKNHSKKKCDTV27C/ 17:32:51 ^ thats the thread 17:33:29 dgilmore: how would test suites work there? 17:33:56 nirik: start with making sure the formating is valid 17:34:11 nirik: extended to make sure we do not break composing 17:34:24 nirik: make sure updates work 17:34:38 all the things that in the past we have managed to get wrong 17:34:46 well, those are all pretty difficult. You would need to make the fedora-release package, compose something with it and test it with openqa or something 17:35:09 it would need something yeah 17:35:38 I mean it would be awesome, but it sounds like a LOT of work. ;) 17:35:46 right I agree 17:35:52 it seems to me like a lot of work for not much gain compared to other stuff 17:36:07 Kellin: I disagree 17:36:07 it doesn't seem to solve a current deficiency as far as I can tell 17:36:14 we should have it regardless 17:36:25 well, either way: if it's in pagure.io or in src.fp.o as upstream its really the same work... just location... 17:36:31 Kellin: it does 17:36:53 nirik: putting it in src.fpo.o allows provenpackagers to update it as well; which is why we would need a test suite to help prevent disasters 17:37:03 because we currenlt frequently need to have expensive loops in getting changes to fedora-release right when mistakes are made 17:37:19 Kellin: sure, althought they can do that now too. 17:37:54 nirik: I had been under the impression from chatting with mboddu that provenpackagers can't commit to it today? 17:38:03 Kellin: that is not true 17:38:19 Kellin: They cant commit to pagure.io where fedora-release is hosted, but they can make patches and do releases 17:38:30 they can... just their changes would be overwritten the next time we do a release from pagure.io and sync it over 17:39:30 So, conclusion or punt? 17:39:48 We need someone to scope the test suite and create a plan for it and get that into Tiaga 17:39:51 Taiga 17:40:01 I see dgilmore's reply in that thread asking for a test suite as part of the move, and no reply to that. ;) 17:40:19 nirik: Yeah, thats the last message on the thread :D 17:40:34 apparently I scared everyone off 17:40:41 dgilmore: haha :) 17:41:23 anyhow, I'd say lets keep the status quo until people wanting to do the move step up to help setup a test suite. 17:41:27 If we could improve the reliability and trust in changes we make, then yep lets do it 17:41:38 get CI tied into PR's 17:41:45 nirik: +1 17:41:55 I would go with nirik here 17:42:02 so - to that point 17:42:06 we could at least have simple-ci build the package as a start of tests. 17:42:20 we ask sly to make a taiga card for the test suite and then we close these with "when this is done, this can happen" 17:42:30 that way it's on the roadmap but not cluttering issues since it's definitely a long-term item not a breakfix 17:42:52 it also would get it somewhere Fedora leadership can prioritize during F29 /F30 /FXX planning sessions 17:42:59 sure. +1 to that. 17:43:17 Kellin: Sure, can you create the taiga ticket? 17:43:19 oh, also, we should make a README for the src.fp.o package that tells people to submit pr's against the upstream project 17:43:36 nirik: yeah 17:43:37 mboddu: sure, add it as an action item 17:44:25 * Kellin thinks it sounds like an epic more than a card with tasks - everyone agree? 17:44:26 does someone want to take that as a task also? 17:44:41 nirik: I can 17:44:49 * mboddu adding to the info actually 17:44:50 mboddu: great. thanks. 17:45:20 Kellin: yeah, likely a initial thing just testing the rpm builds, then something that tests more, then something that tests a compose with it, etc. 17:45:46 #info We are okay with moving fedora-release from pagure.io to src.pagure.io but we want to have the test suits ready 17:46:26 #action Kellin is going to file a taiga ticket with what needs to be done for test suits 17:47:35 as I think about this - one question 17:47:47 #action mboddu will add a README to fedora-release repo on src.fp.o that tells people to submit PR's to upstream fedora-release on pagure.io 17:47:47 if src.fpo.org is the downstream for projects to package into fedora 17:48:08 why are we going to move commit work into src.fpo instead of keeping it in our upstream? 17:48:19 Kellin: src.fp.o is the authoritative location for all sources in fedora 17:48:39 so wouldn't the workflow still be commit to upstream, then merge into src.fpo? 17:48:51 thats what we have today 17:48:56 and we're wanting to invert it 17:49:01 no 17:49:09 wanting to kill the upstream and move it to src 17:49:18 so src.fp.o is the upstream 17:49:22 Kellin: the request is to make src.fp.o be the sole canonical location 17:49:22 and the fedora package 17:49:50 pagure.io/fedora-release will cease to be used 17:50:00 it will reduce some work for us 17:50:49 (work that can by the way be pretty much entirely automated with a standard client-side script) 17:51:11 Kellin: Like not building it manually by us when things gets changed that doesn't need our involvement 17:53:08 Okay moving on 17:53:15 #topic Open Floor 17:53:35 I have one, not sure how big of a topic its gonna be 17:53:38 I have one and I think Kellin has one 17:53:41 Does anyone has anything? 17:53:47 dustymabe: Sure, go ahead 17:54:10 mine is mostly small, but did want to point people towards it if they have not already seen 17:54:15 https://lists.fedoraproject.org/archives/list/rel-eng@lists.fedoraproject.org/thread/637MOWYZWLAL7FKDCLOQEJUIWOB3TBMB/ 17:54:21 related to moving around ostree repos 17:54:26 mboddu: epic https://taiga.fedorainfracloud.org/project/acarter-fedora-docker-atomic-tooling/epic/986; assigned a card to you for the subtask 17:54:49 dgilmore: had a suggestion and I'm working to incorporate his request and submit a new proposal and then me and patrick will make it happen 17:55:33 * nirik is fine with it. Make it so. 17:55:42 +1 17:55:49 dustymabe: I think we keep it simple and have a unified repo 17:55:54 I'll work with puiterwijk to make it happen (new proposal email will be later today) 17:56:07 dgilmore: right. I'm working to incorporate that suggestion in the proposal 17:56:09 dustymabe, so the ostree is composed in one location, and then a ref is pushed to another repo? 17:56:19 masta: yes 17:56:26 that sounds cool 17:56:31 masta: yes. there is a compose ostree repo and a prod repo (where things are published to) 17:56:33 though its composed in one repo and its synced elsewhere 17:56:57 Kellin: want to bring up your topic? 17:57:22 dustymabe: unrelated, we should work with teh compose team to see if we can redo how we do things, so that we can make s390x possible 17:57:30 but thats a bit off topic 17:58:03 dustymabe: sure, so basically wrote up the compose monitor idea per our discussions last week 17:58:09 #info dustymabe is working on a new proposal based on dgilmore comments on https://lists.fedoraproject.org/archives/list/rel-eng@lists.fedoraproject.org/thread/637MOWYZWLAL7FKDCLOQEJUIWOB3TBMB/ 17:58:46 I have not submitted a PR yet - dusty approved of it this morning, this is the working draft: https://pagure.io/fork/kellin/releng/blob/compose-monitor-tooling/f/docs/source/projects/compose_monitoring.rst 17:58:46 dgilmore: yes. let's open a releng ticket for that (if there isn't one already) 17:59:13 dustymabe: I think it likely just needs one for pungi 17:59:44 dgilmore, dustymabe : Dont you think infra should also be involved? 18:00:13 * dustymabe will try to focus on Kellin's topic for now 18:00:29 #info Kellin wrote up a draft on compose monitor tooling available at https://pagure.io/fork/kellin/releng/blob/compose-monitor-tooling/f/docs/source/projects/compose_monitoring.rst 18:00:56 I guess review and folks come back to us in channel during the week 18:01:05 Kellin: why is it not in the wiki? 18:01:26 I think the tracker dustymabe has setup has been handy... a dashboard would be fine, but... I don't know that it's worth putting a bunch of development into. We don't really need anything too fancy IMHO. 18:01:38 dgilmore: AFAIK, its just a draft and will be moved to wiki, Kellin correct me if I am wrong 18:01:47 dgilmore: had a discussion yesterday in #releng; 95% of our docs live in the sphinx docs 18:01:49 but we can gather requirements from the tracker thing we have now 18:02:19 nirik: Yeah, I agree 18:02:31 Kellin: work proposals are not docs though 18:02:58 Kellin: I really do not think that it is approporiate to have in releng docs 18:03:13 but the where and how is a different thing 18:03:14 so is this proposal to make the pungi logs more accessible? 18:03:47 masta: i think the idea is to add sanity to our workflow for debugging pungi issues (i.e. let's have tools that help us do our job better) 18:03:51 masta: well, to track failures, allow comments and info so people know whats going on... 18:03:55 masta: they are already accessible, it's a proposal to assess and analyze failures over time 18:03:55 masta: Well, not exactly, we are tracking the failures and why it happened and what we did to fix it 18:04:00 Kellin: so reading though it, I would like to see some details, it's very wish washy, how do we plan t achieve the stated goals 18:04:18 masta: pungi logs are entirely accesable 18:04:35 so - re-write it. we can close topic and move it to next week; we're already over time anyway. 18:04:36 right, so I was wondering why this proposal was wanted 18:04:46 add an action item for me mboddu 18:04:47 moving rawhide compose time may need to be a devel list discussion + fesco ticket... since people are very used to the existing time. But I'm definitely open to moving it to whenever would help us best. 18:05:07 masta: so we know what we are trying to achieve 18:05:20 masta: and have it turned into a MVP or Epic in taiga 18:05:38 masta: and have the work scoped, prioritised and completed 18:05:41 #action Kellin will rewrite the draft and put it on wiki 18:05:49 #undo 18:05:49 Removing item from minutes: ACTION by mboddu at 18:05:41 : Kellin will rewrite the draft and put it on wiki 18:06:12 #action Kellin will rewrite the draft on compose monitor tooling and put it on wiki 18:06:15 Kellin: perhaps we write it up as a MVP in taiga? 18:06:25 MVP? 18:06:31 dgilmore: I'm MUCH more aligned to that than with doing it in Wiki 18:06:48 * dgilmore notes we really do not have a good way to propose things like this, so it is very much flying by the seat of our pants 18:06:51 dgilmore: we don't really use wiki anymore and Taiga is a better place for work tracking 18:06:55 dgilmore: (we being releng) 18:07:11 mboddu: undo one more time, I'll write up it up in taiga 18:07:15 #undo 18:07:15 Removing item from minutes: ACTION by mboddu at 18:06:12 : Kellin will rewrite the draft on compose monitor tooling and put it on wiki 18:07:17 Kellin: internally MVP's are scoped in google docs, and turned into Jira tickets when approved 18:07:18 nirik: minimum viable product (corporate speak) :) 18:07:32 Kellin: which is kinda what I was looking to replicate 18:07:36 #action Kellin will rewrite the draft on compose monitor tooling and put it on taiga 18:07:39 ok. 18:07:46 so that works for me 18:07:56 I have one other thing, but we're over time. I'll ask about it after lunch in chan 18:08:53 Anyway, lets close the shop now, we are over time 18:09:02 Thanks everyone for joining 18:09:17 thanks mboddu 18:09:20 If anyone has anything more to talk about, lets talk in #fedora-releng 18:09:23 #endmeeting