16:01:00 #startmeeting RELENG (2019-10-03) 16:01:00 Meeting started Wed Oct 2 16:01:00 2019 UTC. 16:01:00 This meeting is logged and archived in a public location. 16:01:00 The chair is mboddu. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:00 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:01:00 The meeting name has been set to 'releng_(2019-10-03)' 16:01:01 #meetingname releng 16:01:01 #chair nirik sharkcz pbrobinson pingou mboddu dustymabe ksinny jednorozec 16:01:01 #topic init process 16:01:01 The meeting name has been set to 'releng' 16:01:01 Current chairs: dustymabe jednorozec ksinny mboddu nirik pbrobinson pingou sharkcz 16:03:10 morning 16:05:25 * dustymabe waves 16:06:10 Hello guys 16:07:51 I just wanna bring up one issue and then we can go to backlog 16:08:02 #topic #8851 RFE: When the retirement commit is reverted before a package gets blocked, don't block it 16:08:08 #link https://pagure.io/releng/issue/8851 16:08:35 How can we achieve this? 16:08:42 yeah, I think this is perhaps possible... but will require reworking the listener 16:09:04 But what would be the logic? 16:09:07 it could just do those checks... 16:09:26 when it sees a revert of a dead.package commit, check how long ago it was retired/blocked... 16:10:20 But you cannot push if the eol is set to a past date, so the listener cannot see that commit 16:10:43 ah true... 16:11:00 so it would have to be in the eol hook... 16:11:28 so yeah, perhaps this isn't too possible after all 16:11:41 Right 16:11:48 Its not easy as we think it will be 16:13:00 perhaps this could be done similar to the way branch requests/etc are done... with a pagure plugin... 16:14:53 I think, in order to get away from the confusion, we should update fedpkg retire to set the eol date of yesterday's date than today's date 16:15:30 That way, people cannot immediately git revert the retire commit, and then they can file a ticket to unretire it (like we do normally) 16:16:20 It is a special snowflake to retirement case and should be considered as retirement only 16:17:11 hum, yeah... I wonder if we replace pdc if we could add another state. 16:17:31 "active" "retired" "retired_less_than_8_weeks" 16:18:00 and have the eol hook allow reverting the one but not the other... dunno. 16:18:31 True 16:18:56 Or another thing we can do is automate the entire damn thing (at least retired less than 8 week) 16:19:14 automation++ 16:19:14 nirik: Karma for automation changed to 1 (for the current release cycle): https://badges.fedoraproject.org/tags/cookie/any 16:19:18 ha. oops 16:19:32 So, that people can unretire stuff by themselves and if its more than 8 weeks, ask them to file the releng ticket 16:19:48 Its a nice nick to have, we will get more cookies :D 16:20:37 not as much as if someone had the nick 'C' 16:21:20 Hahaha :D 16:21:29 So, what are our options here: 16:21:43 1. Set the eol date to last day in fedpkg retire for now 16:21:44 I don't think there's an easy short term fix... 16:21:51 2. Automate the <8 weeks thing 16:22:14 yeah, 1 would at least avoid some confusion... 16:22:16 Yeah, which is why I am suggesting 1 for now 16:23:42 yeah, lets do one and look to automate when we can 16:23:43 #info When fedpkg retire is called, set the eol date to previous date so that immediate reverts are not allowed. This is a temporary fix until we get to automate the unretirement for less than 8 weeks. 16:24:04 #info mohanboddu will file a fedpkg ticket 16:24:16 Or probably patch it myself 16:24:24 I mean send the PR 16:24:25 is it fedpkg? 16:24:26 i think it's job for pdc-updater not fedpkg 16:24:28 or is it the listener? 16:24:35 right 16:24:39 Oh right 16:24:41 fedpkg / user has no token 16:24:44 #undo 16:24:44 Removing item from minutes: INFO by mboddu at 16:24:04 : mohanboddu will file a fedpkg ticket 16:25:07 Thanks mizdebsk :) 16:25:09 I'm glad our processes are so clear! NOT. 16:25:30 haha :) 16:25:56 #topic Open Floor 16:26:11 #info Next Tue, Oct 8th is the f31 final freeze 16:26:17 so I had 2 things... 16:26:23 * dustymabe has one 16:26:29 can I go real quick? 16:26:36 let me do one, then dusty then my longer one? 16:26:41 +1 16:26:47 this will be quick... 16:26:59 FYI, I have scheduled a koji outage later today (21UTC) 16:27:14 I'm going to update prod to 1.18.0 and also give db-koji01 more memory. 16:27:22 should be down only a few min. 16:27:35 Thats it on that... dustymabe go ahead. 16:27:40 https://hackmd.io/5uB7hOJKSjGUt65iLgPnbA#Existing-requests-for-Fedora-Releng 16:27:53 can we make some progess on the final comment in #8811 16:27:59 https://pagure.io/releng/issue/8811#comment-598901 16:28:10 #info nirik updating koji in prod to 1.18.0 and there will be a short outage later today at around 21UTC 16:29:31 dustymabe: I can try and do that this week... friday? 16:29:46 nirik: sounds good to me.. mind adding a comment to the ticket? 16:29:52 sure thing. 16:29:57 * mboddu would like to follow it, if its a simple thing 16:30:35 #info nirik and dustymabe will work on https://pagure.io/releng/issue/8811#comment-598901 on Friday this week 16:31:15 +1 16:31:15 sure, sounds good. following is good. 16:31:21 * dustymabe runs off to run another meeting 16:31:38 Thanks dustymabe for joining 16:31:39 we can even pick a time... 16:31:56 nirik: I got only 1 meeting on Fri, so anytime would work for me 16:32:09 * mboddu feels happy to say that he has only 1 meeting :D 16:32:17 say 10am my time / 1pm east? 16:33:01 nirik: That would work, and I am also in pacific for this week, I am in CA :D 16:33:07 pacific time* 16:34:09 * mboddu waiting for nirik's bigger stuff to discuss 16:34:14 oh really? cool. just touristing around? 16:34:31 can someone toss a meeting invite out on calendar if 10am PDT is ok? 16:34:40 So, the other thing I had was backlog. 16:35:04 Just like for the infra side I marked some old releng tickets backlog and we can try and focus on them one or two at a time and get some done. 16:35:41 my thought was actually to mix these with the infra stuff... to more cross train the teams... but if you prefer we can keep them seperate. 16:35:52 https://board.net/p/fedora-infra-backlog has a list 16:36:28 so do we want to pick one and see if we can make progress by next wed? 16:36:46 nirik: Came here on some personal work - regarding CA visit 16:37:02 * mboddu can also send the invite 16:37:35 nirik: Sounds good to me 16:37:38 * mboddu checks 16:37:54 any of them suit your fancy? 16:38:27 May be 7718 or 7793? 16:38:39 ok. 16:38:42 Or 6746 16:38:45 I like all of them :D 16:38:55 6746 is an old one... 16:39:02 And useful one 16:39:11 Lets take a look at 6746 16:39:23 I can get something done 16:39:31 yep. I think adamw already re-implemented it in openqa... butwe can try and come up with a better/more generic way 16:39:42 Yup 16:40:01 I think the thing to do is schedule an actual time to work on it, and gather everyone interested... we can sort out a plan and all help 16:40:57 so, perhaps send to infra/releng lists we are gonna work on it, pick a time next week and see who shows? 16:41:10 Okay, I will do that 16:41:30 were we going to try and use odcs for this? 16:41:54 We can 16:42:06 Not sure if that an overkill or not 16:42:21 except that just gives a repo right? not images... 16:42:31 but could use odcs to make a repo and from that make images... 16:42:56 My initial plan was, listen to fedmsg, call a pungi script with a config and put it under /mnt/koji/compose/ 16:43:19 we could also use loopabull... 16:44:11 which now that I think of it, needs porting to fedora-messaging 16:44:45 anyhow, we can discuss and decide what the best plan is. 16:44:53 we can also look at adamw's implementation. 16:45:07 Lets take a look at ODCS, we never used it anywhere in fedora infra/releng, may be we can move other stuff to ODCS as well 16:45:28 sure. 16:45:34 ODCS can also take raw pungi configs 16:45:58 thats all I had... although like I said I'd like to see cross training/work between releng and infra. 16:46:01 So, it can create the repo as well images 16:46:11 I would love that as well 16:48:54 nirik: Who is maintaining pdcupdater now? 16:49:02 no one 16:49:37 cverna is working on a pdc replacement, but it's not at all ready yet 16:49:50 Yeah, ^ this I know 16:50:14 All that it needs is adding "- timedelta(1)" to https://github.com/fedora-infra/pdc-updater/blob/develop/pdcupdater/handlers/retirement.py#L192 16:50:17 and pdc is also somewhat broken... it's not been importing composes right for a while 16:50:33 it = our first ticket 16:51:14 so might be easier to just patch that in ansible... but I suppose we could talk to others and see if we can do a new pdc release... 16:52:06 May be, since the last commit on that project is on May 11 2018 16:52:21 * nirik has to go take a dog out... back in a bit. 16:53:25 Okay, thanks for joining nirik 16:53:31 #endmeeting