18:00:06 <nirik> #startmeeting Fedora Infrastructure Ops Daily Standup Meeting 18:00:06 <zodbot> Meeting started Wed Apr 8 18:00:06 2020 UTC. 18:00:06 <zodbot> This meeting is logged and archived in a public location. 18:00:06 <zodbot> The chair is nirik. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:06 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:06 <zodbot> The meeting name has been set to 'fedora_infrastructure_ops_daily_standup_meeting' 18:00:06 <nirik> #chair cverna mboddu nirik smooge 18:00:06 <zodbot> Current chairs: cverna mboddu nirik smooge 18:00:06 <nirik> #info meeting is 30 minutes MAX. At the end of 30, its stops 18:00:06 <nirik> #info agenda is at https://board.net/p/fedora-infra-daily 18:00:06 <nirik> #topic Tickets needing review 18:00:07 <nirik> #info https://pagure.io/fedora-infrastructure/issues?status=Open&priority=1 18:00:14 <nirik> who all wants to stand up? 18:00:15 <mboddu> Hello 18:00:27 <mboddu> Then I cant type easily :) 18:00:29 <nirik> nothing in infra needs review 18:00:42 <smooge> here 18:00:55 <nirik> a few new releng ones 18:00:58 <mboddu> Yes 18:01:00 <mboddu> .releng 9391 18:01:01 <zodbot> mboddu: Issue #9391: rsync the directory tree to mm-backend01 - releng - Pagure.io - https://pagure.io/releng/issue/9391 18:01:27 <nirik> yeah, we can do that... 18:01:40 <mboddu> nirik: I want to discuss the plan here 18:01:45 <nirik> ok 18:02:18 <mboddu> Once MM is setup, we will move the repodata to the /pub/ somewhere? (I am prefering /pub/alt/) 18:02:24 <jcline> Okay, how much does the idea of having an alternate to the lookaside cache that's just a git repository horrify people initially? 18:02:28 <nirik> no. I don't think we want that 18:02:56 <nirik> mboddu: I think that would confuse people a lot. 18:03:29 * jcline notices a meeting is happening and scuttles away to plot 18:03:53 <nirik> mboddu: so, this is going to be composed in odcs moving forward now? 18:04:00 <mboddu> nirik: Yes, thats the plan 18:04:23 <mboddu> Currently its served from sundries01 /srv/web/... 18:04:28 <nirik> is it all set, ie, the next one will be there? also, does odcs support signed repodata? 18:04:47 <nirik> it's not. It's set there, but it rsyncs to all the proxies and is served from them. 18:05:00 <mboddu> nirik: Nope it doesn't I will sign it manually 18:06:00 <nirik> we could just mount the odcs private dir on mm-backend01? but that doesn't work right yet. 18:06:21 <nirik> I was thinking we could just rsync from sundries01 to mm-backend01 like we do the proxies... 18:06:33 <mboddu> Yes, also with DC move, we will loose the odcs backend 18:07:04 <nirik> but the storage is on netapp/nfs? or I guess not yet on that backend. 18:07:07 <mboddu> So, we will be using the same old way until everything is setup in the new DC 18:07:37 <mboddu> Maybe, we could still the older backend, its just couple of builds 18:07:40 <nirik> yeah, lots of moving parts here. 18:07:43 <mboddu> still use* 18:08:01 <mboddu> Thats for sure :) 18:08:17 <mboddu> On that, it reminded me, I need to add the autosign for -openh264 tags 18:08:46 <nirik> that may or may not work without kojipkgs changes. 18:09:19 <mboddu> As in? 18:09:30 <nirik> so, how about short term: we just manually sync the repodata to /srv/cisco-h264/repodata/ or whatever on mm-backend01... 18:09:45 <nirik> so anytime it changes we just update that to get mm to update. 18:09:56 <mboddu> If autosign doesn't work, I have to use the sigulsign_unsigned.py 18:09:56 <nirik> and after the move we come up with a better plan. 18:10:13 <mboddu> Okay 18:10:25 <nirik> we block all access to the rpms there... so people cannot download them from koji. we need to make sure autosign is allowed. 18:10:59 <mboddu> "that may or may not work without kojipkgs changes." which one are you talking about? signing? 18:11:17 <nirik> signing. 18:11:25 <nirik> since we block all access to the rpms... 18:11:43 <nirik> but looks like we do allow the sign bridge. 18:11:46 <nirik> so I guess it's ok 18:11:51 <mboddu> We could give it a try 18:11:59 <mboddu> ^^ Yeah, I thought so, 18:12:13 <nirik> anyhow, does that sound ok? manual for now to get it done and then worry about automating after the move? 18:12:24 <mboddu> Sure 18:12:36 * mboddu checks one thing 18:13:58 <nirik> any other tickets or pull requests or reviews or such? 18:14:15 <smooge> I forgot about the autosign ticket 18:14:34 <nirik> which one? 18:14:35 <smooge> sorry autocloud 18:14:37 <smooge> https://pagure.io/fedora-infrastructure/issue/8708 18:14:52 <smooge> is that a moonshot or a mustang? 18:15:05 <nirik> mustang 18:15:18 * mboddu will wait as I have one other thing to discuss on openh264 18:15:20 <nirik> needs to be made a buildhw. 18:15:20 <smooge> ok .. darn I was hoping for a quick fix 18:15:29 <nirik> and/or we could ship it? 18:15:55 <nirik> mboddu: go ahead? 18:17:09 <mboddu> nirik: Currently we rsync the rpms as well to /srv/web/... on sundries01.phx2.fp.o, but there is a redirect to ciscobinary.openh264.org/... 18:17:29 <nirik> yeah, the redirect is at the proxy layer. 18:17:47 <mboddu> Can we get the redirect working on /srv/cisco-h264/repodata/ or something on mm-backend01 18:17:49 <mboddu> Okay 18:18:29 <nirik> no, because nothing should be able to access it. 18:18:41 <nirik> mm-backend01 is internal only. it has no web access to it that I know of. 18:19:13 <mboddu> Okay, that answers my question, I was wondering there is/will be a web access 18:19:38 <nirik> should not need it at all. it just needs to crawl the directory to read the repodata directly... 18:21:25 <nirik> I had one item... 18:21:59 * cverna is late but around in case 18:22:09 * mboddu still thinking... 18:22:15 <nirik> there are 3 packages that were orphaned recently that I think infra-sig needs to take... 18:22:27 <mboddu> Something seems not right, but I cannot pin it :( 18:22:30 <nirik> reg pam_url python-robosignatory 18:22:44 <nirik> they all FTBFS, so will need that fixed, but we use them all. 18:22:56 <nirik> I'll mail the list about it. 18:23:27 <mboddu> I recently sent a patch to fix sigul 18:23:34 <mboddu> But I could look at them 18:24:00 <nirik> also, there is https://bugzilla.redhat.com/show_bug.cgi?id=1814243 (bodhi does not build with python 3.9) 18:24:31 <cverna> yeah I saw your ping on that bz, I ll answer and will be looking at it 18:24:38 <nirik> cverna: thanks! 18:25:00 <michel_slm[m]> Igor Gnatenko: do I need to be an admin in src.fedoraproject.org to request a branch? is being a committer not enough? 18:25:06 <cverna> yeah reg is :S, go packaging has changed quite a bit and I don't think it will be an easy one to update 18:25:11 <michel_slm[m]> ref: https://pagure.io/releng/fedora-scm-requests/issue/24196 and https://pagure.io/releng/fedora-scm-requests/issue/24195 18:25:20 <nirik> cverna: yeah, I tried, but it was not at all easy 18:25:34 <nirik> michel_slm[m]: yes, you need to be admin 18:25:54 <michel_slm[m]> <nirik "michel_slm: yes, you need to be "> aha thanks 18:26:00 <mboddu> nirik: So, currently I rsync 33/ dir (https://codecs.fedoraproject.org/openh264/33/) to /srv/web/ on sundries01, but with mm-backend01, we only sync repodata and debug/repodata/ or everything? 18:26:07 <cverna> nirik: I ll try to ask some help from the golang sig. jcajka might be able to help us 18:26:13 <mboddu> michel_slm[m] and nirik : Nope, I fixed it in the recent build 18:26:25 <mboddu> michel_slm[m]: I will take a look after the meeting 18:26:40 <michel_slm[m]> mboddu oh being a commiter will be enough? thanks, I'll wait 18:26:44 <mboddu> I guess ignatenkobrain didn't update fedscm-admin 18:27:00 <nirik> mboddu: yeah, we could just sync the repodata (thats all it needs)... or just do all of it to make it easier. It shouldn't matter. 18:27:15 <nirik> ah, sorry for mistaken info... 18:27:29 <mboddu> michel_slm[m]: Yes, I fixed and built it this Monday, so he might not be using the updated version 18:28:02 <nirik> mboddu: idea: have fedscm-admin put it's version in tickets it updates. :) 18:28:34 <mboddu> nirik: If we sync the entire dir, then mm will sync the rpms to mirrors, isn't it considered as we distributing the rpms? 18:29:22 <nirik> mboddu: no, mm-backend01 doesn't distribute any rpms. It only distributes information on mirrors to mirrorlist-servers. 18:30:03 <nirik> so it would be able to make a metalink with the checksums and such from the repodata but it would point people at codecs.fedoraproject.org (proxies) and then they in turn redirect to cisco... 18:30:15 <mboddu> nirik: Okay, sorry, my understanding of how MM works is not very good 18:30:37 <nirik> it's complex. ;( 18:30:38 <mboddu> Thanks for the info 18:31:11 <nirik> basically the frontends just provide the https://admin.fedoraproject.org/mirrormanager web interface. 18:31:15 <mboddu> Indeed it is 18:31:36 <nirik> the crawlers just provide a process to download from mirrors and check contents. 18:31:56 <nirik> and the backend just crawls the nfs volumes for changes and updates the data the mirrorlist servers serve 18:32:22 <mboddu> nirik: I will run it by you when we need to do this, just for being safe 18:32:54 <nirik> yeah, it's good to make sure. 18:33:01 <nirik> ok, if nothing else shall we close? 18:33:03 <mboddu> Okay, thats all I have 18:33:05 <mboddu> Yup 18:33:15 <mboddu> Thanks nirik 18:33:16 <nirik> thanks for coming everyone! 18:33:20 <nirik> #endmeeting