17:00:10 #startmeeting RELENG (2018-08-30) 17:00:10 Meeting started Thu Aug 30 17:00:10 2018 UTC. 17:00:10 This meeting is logged and archived in a public location. 17:00:10 The chair is mboddu. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:00:10 Useful Commands: #action #agreed #halp #info #idea #link #topic. 17:00:10 The meeting name has been set to 'releng_(2018-08-30)' 17:00:10 #meetingname releng 17:00:10 The meeting name has been set to 'releng' 17:00:10 #chair nirik tyll sharkcz masta pbrobinson pingou puiterwijk maxamillion mboddu Kellin dustymabe 17:00:10 Current chairs: Kellin dustymabe masta maxamillion mboddu nirik pbrobinson pingou puiterwijk sharkcz tyll 17:00:10 #topic init process 17:00:41 * nirik waves 17:00:53 * sharkcz is here today as well 17:01:01 * bcotton lurks 17:01:15 Hello everyone 17:01:29 Since sharkcz is here, lets start with alt arch updates 17:01:38 #topic Alternate Architecture Updates 17:02:16 so we have the ppc qemu vs HTM issue (https://pagure.io/releng/issue/7703) on the image builders 17:02:18 sharkcz, nirik : Any update on https://pagure.io/releng/issue/7703 17:02:53 yep. see linked bugzilla bug 17:03:36 maybe there can be something easily done on the infra side, but the virt guys discuss a proper solution 17:04:26 I asked if there's a config option for now... 17:05:15 nirik: So, the plan is to go with HTM=no option? 17:05:30 I do not yet know if thats a valid workaround or not. 17:05:52 * pbrobinson is here for a little bit btw 17:06:08 technically disabling HTM only might cause little performance drop (invisible I would say) 17:06:24 * mboddu waves at pbrobinson :) 17:07:20 So, the ball is in the virt guys court? Or someone has to make that decision? 17:07:34 I'd like to hear from them... 17:07:41 ack 17:07:50 if we don't we could also always revert back, but then we loose the security update 17:08:23 nirik: Yeah, I dont like ending up in that situation 17:09:41 #info nirik debugged the issue and contacted virt team. We are currently waiting on their feedback. More info @ https://pagure.io/releng/issue/7703 and https://bugzilla.redhat.com/show_bug.cgi?id=1614948 17:09:49 Okay, moving on 17:10:06 I have only one topic to discuss 17:10:13 another issue is https://bugzilla.redhat.com/show_bug.cgi?id=1623547 17:10:40 sharkcz: Oh sorry, go ahead 17:10:47 ok 17:11:11 so something changed and we can't install in VM with 2GB mem 17:11:27 thats not been possible for a while I don't think? 17:12:04 in F-29 probably, but worked in F-28 17:12:25 the system simply goes crazy 17:13:10 IIRC I tried F-29/rawhide nightly some time ago (month or 2?) and it was fine 17:13:38 * mboddu also thinks he tried F28 on VM with 2gigs memory and it worked 17:14:36 I suspect the zswap/zram thing, so will look closely on it 17:15:05 I know pwhalen has been having some issues here 17:17:10 which reminds me one thing - do we have an archive of nightly composes? like 1 per week for the last month, one per month for older? 17:17:33 no 17:17:51 we have the last 14 days of composes ish... 17:17:52 sharkcz: No, we will only keep 2 weeks of composes 17:18:12 right, so no way to check older stuff 17:18:36 it would be handful especially when nightlies are failing 17:18:40 sharkcz: I guess not 17:19:36 well, koji might have older I guess... 17:20:02 * sharkcz thinks about filling a RFE to implement such policy 17:20:36 nirik: This reminds me, can we create a koji inheritance of the base tag when a nightly compose is started, so that we can reproduce a compose for a given date? 17:20:40 it would possibly take up a lot of space 17:20:48 Sorry, off the topic, but just got the idea 17:21:06 nirik: Its inheritance, so it should not take the space 17:21:24 s/the space/any space/ 17:21:27 dunno, would that take a while/db resources? 17:21:45 it would be <10 more compose, maybe 5 17:22:38 we talked at one point about saving 'known good' ones that passed testing... 17:22:40 nirik: AFAIK, cloned tags will take a lot of resources but not inherited tags, but we can check 17:22:55 nirik: or that 17:23:05 I'm not sure its related to sharkcz's issue, but container images on armhfp are failing in f29, seems to be oom. The vm locks up at fetching squashfs/treeinfo. Using nfs mounted squashfs works as expected 17:23:43 * mboddu takes a note and will investigate later 17:24:08 ^ about the koji inheritance, not the anaconda failure :) 17:25:32 sharkcz: Do you think https://bugzilla.redhat.com/show_bug.cgi?id=1622248 is related to your issue? 17:26:40 mboddu: doesn't seem so, Fedora-29-20180815.n.1 failed same way as Fedora-29-20180827.n.0 17:26:45 with OOM 17:27:17 sharkcz: Okay, I guess then we have 2 issues it seems 17:27:28 at least. ;) 17:28:10 nirik: I am trying to be optimistic here :D 17:29:06 Anyway, at least the 2nd one seems to have a fix 17:30:35 #info There seems to be an issue with anaconda when installing on a VM with 2 gigs of mem and booting container images on armhfp 17:30:42 that should be all alt-arch from me 17:30:52 sharkcz: Thanks :) 17:31:09 Moving on... 17:31:13 #topic #7718 The FTBFS cleanup policy is not happening 17:31:22 #link https://pagure.io/releng/issue/7718 17:32:28 well, it's not been clear to me who is responsible for that... 17:32:39 but I suppose it could be releng... 17:32:55 nirik: Yeah, but we never had a script for it, at least afaik 17:33:21 I want to bring it up and see if anyone can confirm it? If not, can anyone help with it? 17:33:34 Or else, I might have to create one 17:34:41 pbrobinson: ^ do you know? 17:35:14 about the FTBFS cleanup policy? 17:35:40 pbrobinson: No, a script that does it? 17:36:00 I believe it's a cron job that runs on the koji hub 17:36:25 I don't think we have ever had any scripting for beyond filing them... 17:36:29 oh, no idea about the weekly reminder 17:36:59 pbrobinson, nirik : Okay, let me add it to my list of things to do, thanks 17:37:26 #info We have to work on the tooling and once its done, we will enable it 17:37:43 Okay, thats all I got 17:37:49 #topic Open Floor 17:38:01 Anybody has anything to share? 17:40:57 * nirik has nothing off hand 17:41:44 Okay, I will give 5 min before I close the shop 17:42:14 And, anyone interested in security keys, https://store.google.com/product/titan_security_key_kit?utm_source=gblog&utm_medium=google_oo&utm_campaign=GS101145&utm_term=sk-launch 17:42:20 nirik: ^ 17:42:28 yeah, saw... 17:43:18 nirik: Not bad for $60 17:44:43 Okay, thanks everyone for joining 17:44:45 #endmeeting