16:01:47 #startmeeting RELENG (2019-05-30) 16:01:47 Meeting started Wed May 29 16:01:47 2019 UTC. 16:01:47 This meeting is logged and archived in a public location. 16:01:47 The chair is mboddu. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:47 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:01:47 The meeting name has been set to 'releng_(2019-05-30)' 16:01:48 #meetingname releng 16:01:48 The meeting name has been set to 'releng' 16:01:48 #chair nirik tyll sharkcz masta pbrobinson pingou puiterwijk maxamillion mboddu dustymabe ksinny jednorozec 16:01:48 #topic init process 16:01:48 Current chairs: dustymabe jednorozec ksinny masta maxamillion mboddu nirik pbrobinson pingou puiterwijk sharkcz tyll 16:02:25 .hello2 16:02:29 bcotton: bcotton 'Ben Cotton' 16:04:15 morning 16:04:16 * mboddu waves at bcotton 16:04:25 nirik: Morning 16:04:50 I haven't prepared for this meeting at all, so, lets see 16:07:17 So, lets get to some old tickets 16:07:21 #topic #7400 Rethink how we handle networking config with createImage (oz / ImageFactory) 16:07:28 #link https://pagure.io/releng/issue/7400 16:07:32 * mboddu reading back 16:08:21 * nirik reads, not sure I ever saw this one 16:08:39 oh wait, yeah 16:09:32 I remember seeing this, but didn't get to it 16:10:38 so right now, the oz images install with a ensa263152761 whatever interface 16:10:56 which gets written to a ifcfg file by NM 16:11:15 so we have to nuke it in the kickstart to allow for someone using that image to get their own interface 16:11:42 if we just set everything to use eth0 it's easier to nuke 16:13:20 so, I'm in favor of making these changes... 16:14:17 nirik: So, set to use eth0 and rm -f /etc/sysconfig/network-scripts/ifcfg-eth0? 16:14:27 yeah 16:14:34 Thats a change to koji 16:14:54 I can file a ticket against koji 16:15:37 Or say we agree to it and ping mikem on the ticket 16:16:22 sure, or adamw might do a PR if he's in the mood 16:19:11 #info mboddu commented on the ticket to check with mikem and adamw and if anyone can make the necessary changes to koji 16:20:59 #topic #7802 Several EPEL7 package conflicts with RHEL/CentOS 7.5 16:21:05 #link https://pagure.io/releng/issue/7802 16:21:11 I dont think its an issue anymore 16:21:16 nirik: Can I close the ticket? 16:22:41 I've not looked... well, we should probibly look at those packages and see if they are still/ever causing problems. 16:23:27 perhaps we could give reporter more time to chime in 16:23:47 and ask them... what packages in EPEL7 are causing problems specifically. 16:24:03 since those are the only ones we can do anything about... by removing them. 16:24:13 nirik: Okay, I will ping him on the ticket again 16:24:26 the lists are a bit confusing to me. 16:24:34 Yes, exactly 16:24:52 I just want: here's epel7 packages that overlap with any rhel repos we care about... so we can say 'yes, thats ok' or 'no, we should remove X' 16:25:47 nirik: I know there is no communication involved when packages are moved to RHEL and we have to remove them in EPEL 16:26:28 Cant we ask RHEL maintainers to send us the list whenever they do so, or once a week/month? 16:26:57 Or at the least when there is a release? 16:28:10 we have a list of packages in rhel 16:28:37 https://infrastructure.fedoraproject.org/repo/json/pkg_el7.json 16:28:42 https://infrastructure.fedoraproject.org/repo/json/pkg_el6.json 16:29:44 nirik: Yes, I know them, but they contain all of them, not anything newly added 16:30:07 Or we can write our own thing to compare list of packages in epel vs those json files 16:30:26 sure. 16:30:41 but we need to whitelist the limited arch ones... 16:30:50 Yes 16:30:55 do we know how often that changes? because if it's infreqent, I can talk to my counterparts who work on RHEL to notify me/someone 16:32:03 it changes at point releases... 16:32:21 I'm not sure notifying us is great. 16:32:22 bcotton: ^ mostly at point releases 16:32:34 nirik: Why? 16:32:52 there is supposed to be a step when adding a new package to rhel to file a bug on the epel package if it exists and getting the maintainer to retire it. 16:33:05 because releng doesn't scale, and we should get this as automated as possible. 16:33:10 manual workflows suck. 16:33:26 agreed. if there's something that's supposed to be happening that isn't, i can bring that up, too 16:34:16 I suppose the best we might do is a script that runs every day and checks rhel against epel and has a whitelist. When something new comes up we retire it or whitelist it. but thats still manual... so dunno. 16:34:24 That is true, but we can add something like "Talk to epel maintainer....blah blah....if the epel maintainer doesn't respond then talk to RelEng" 16:35:29 I am okay with either one rather then being in the dark unless someone files a ticket 16:38:00 well, sure, but it needs script writing... 16:38:55 nirik: I am okay with it 16:39:03 I will create a ticket for it 16:39:15 so, yeah... lets put the generic solution in a ticket... 16:39:25 and in this one ask reporter for specfic details 16:40:02 #info mboddu will ping the maintainer about the specific epel packages that that are causing issues 16:40:53 #info mboddu will create a ticket to create a script that will compose epel packages with rhel package list json file and report back with results 16:41:17 ack 16:42:19 I think there the is-it-rhel project that pingou maintains 16:42:28 not sure if that solves the issue 16:42:33 cverna: yeah, thats the thing that makes those json files. ;) 16:42:39 (I thought) 16:42:46 ah ok :) 16:42:55 https://pagure.io/is-it-in-rhel 16:43:18 so we need a is-it-epel and compare both :P 16:43:19 ah no, thats a interface to it. ;) 16:43:25 yeah 16:43:38 or is-it-in-both-epel-and-rhel 16:43:47 :) 16:43:52 haha :) 16:44:17 at least for epel8 we won't have to deal with the limited arch stuff. 16:45:07 Yes 16:47:03 #topic #7763 define and implement retention policy for nightly composes 16:47:09 #link https://pagure.io/releng/issue/7763 16:47:28 I am not sure if we need to change any policy here 16:47:33 14 days seems good to me 16:47:53 and when we really need one, we can use koji event to clone the tag and create a compose 16:48:07 But that means, we have to keep track of those koji events 16:48:12 cverna: ^ fpdc :D 16:49:13 yeah. that all sounds good to me. 16:50:22 cverna: So, can you add that to the requirements doc? 16:50:28 for fpdc 16:51:43 #info We will not change the policy but we will keep track of the koji event id (probably in fpdc) from which we can clone a koji tag and run a compose from it if and when its really needed. 16:53:03 #topic Open Floor 16:53:09 I got 2 things 16:53:18 I wanted to share my progress on builders this week... 16:53:26 #info F28 went EOL yesterday 16:54:26 #info FLOCK CFP selection starts on Jun 1st, so please submit your talks 16:54:29 nirik: Go ahead 16:54:59 so, our builders currently are mostly fedora29, some f28 (the ppc64 ones) and some f27 (armv7) 16:55:31 I'm hoping to move armv7 to f30 and all the f29's to f30 and drop the ppc64 ones 16:55:41 +1 16:55:48 I'll need testing on armv7 to try and find a stable os/kernel combo. 16:56:27 also, oz and ImageFactory in rawhide have moved to python3, and I am going to try and move them for us too in f30-infra for now. 16:56:29 nirik, or f29 w/ 4.20 GA , that seems the most stable 16:56:45 which means that koji also has to be moved to python3 16:57:02 pwhalen: ok. I was hoping f30 with 4.20 would work, but I haven't gotten to testing yet. ;( 16:57:28 I had a question about the workstation repo, do we still plan on moving workstation to use Everything? 16:57:30 I've got oz/imagefactory python3 builds in infratags... also koji, but it needs a few more patches. 16:57:54 pwhalen: I thought so.. yes. 16:58:33 I'm not sure the timeframe for builders... I am going to work on it more this week, but I am out next week and then traveling the week after, so it might not be until the week after that. 16:59:46 * nirik looks for when the mass rebuild is. 17:00:04 2019-07-24... ok, cool. 17:00:27 Yeah, we have time 17:00:29 also, I am continuing to push about moving our s390x builders to kvm... there's some progress on it... 17:00:39 Oh finally 17:00:51 thats all I had 17:02:08 nirik, ok. I've seen 4.20 GA with f30 userspace pause and the 'prefetch abort' error. F29 looked much better for some reason. 17:02:35 bummer, ok. I was hoping to move them all to f30... but ok, can deal with 29 I guess 17:02:41 me too 17:03:08 but hopefully we can get to the bottom of it and solve it upstream once and for all. 17:03:10 I really hope it can be solved and not worked around 17:03:13 agreed! 17:03:33 * nirik wonders why no one else hit this. I guess not too many armv7 vm's around? 17:03:39 #info nirk is looking into upgrading builders to F30 and killing ppc64 and also working on moving s390x builders to kvm 17:03:59 #info nirik is working on - Oz and ImageFactory in rawhide moved to python3 and trying to move them to f30-infra and if everything works out, then moving to python3 koji. 17:04:22 I did disable ppc64 in bodhi pungi for epel6/7. we need to remove it from koji tags before we retire those machines. 17:06:32 nirik: I guess epel7-build and dist-6E-epel-build build tags, right? 17:06:37 yeah 17:07:02 nirik: Do you want me to do it now or does it have to wait? 17:07:16 go ahead if you want... or I can do it sometime 17:07:22 Oh we went overboard 17:07:31 nirik: I will do it now and then go to lunch 17:07:34 meeting overboard! 17:07:38 haha :) 17:07:45 Thanks everyone for joining 17:07:48 #endmeeting