18:00:06 #startmeeting Infrastructure (2018-01-04) 18:00:06 Meeting started Thu Jan 4 18:00:06 2018 UTC. The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:06 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:06 The meeting name has been set to 'infrastructure_(2018-01-04)' 18:00:06 #meetingname infrastructure 18:00:06 The meeting name has been set to 'infrastructure' 18:00:06 #topic aloha 18:00:06 #chair smooge relrod nirik dgilmore threebean pingou puiterwijk pbrobinson maxamillion 18:00:06 Current chairs: dgilmore maxamillion nirik pbrobinson pingou puiterwijk relrod smooge threebean 18:00:11 Good day everyone 18:00:23 hello 18:00:25 Good Morning Everyone 18:00:27 morning 18:00:36 hello 18:01:30 morning and happy new year :) 18:02:01 ok next up 18:02:13 #topic New folks introductions 18:02:14 #info This is a place where people who are interested in Fedora Infrastructure can introduce themselves 18:02:22 I don't expect many new people this week 18:03:40 and I am right :) 18:03:55 #topic announcements and information 18:03:55 #info Bodhi-3.1.0 mostly deployed (backend01 still needs to be upgraded) 18:03:55 #info Mass updates/reboots due to Meltdown CVE-2017-5754 18:03:55 #info Expect more in the coming time as this gets dealt with 18:03:55 #info Happy New Year 18:04:12 Any other announcements? We had a 2 week 'break' 18:05:21 hello 18:05:28 hello clime how are you? 18:05:38 fine, thx, sry for being a bit late 18:05:46 oh on bodhi 3.1.0... 18:06:00 Bodhi 3.1.0 is fully deployed now actually 18:06:01 I reinstalled stg with f27, so 3.1.0 should be running there now. 18:06:07 cool 18:06:31 #info Bodhi-3.1.0 fully operational. Onward to Alderan 18:06:40 oh right, it's in prod... not sure what version is in stg... 18:06:47 but I see that it's also not working. ;) 18:07:00 Yeah, staging is a very testing release 18:07:11 #info mdapi re-installed w/ f27 in stg 18:07:17 nirik: it looks good to me there btw :) 18:07:42 #info puiterwijk completed operation redeploy proxies to F27 18:08:04 pingou: I did prod yesterday too. ;) 18:08:18 nirik: cool :) 18:08:22 nirik++ 18:08:34 .title https://bodhi.stg.fedoraproject.org/ 18:08:39 nirik: timed out 18:08:43 we need zodbot to say "You are out of cookies.. get more cookies for 0.00001 BTC " 18:09:04 ok anything else for announcements? 18:09:20 I have been trying to make sure we drop all the f25 machiens... 18:09:32 Do you have a deadline? 18:09:35 there's not too many left 18:09:59 like Jan 20? 18:10:06 soon yeah... 18:10:25 before devconf 18:10:38 smooge: December 12, 2017 18:10:40 #info Project Annihilate Infrastructure Fedora 25 systems is coming to a close 18:11:04 (ideally. We will most likely miss that though, unfortunately) 18:11:11 * smooge watched too many Star Wars movies lately 18:11:22 remember: You only have to invent a time machine once. ;) 18:11:55 ok in this case... 18:12:07 #topic Ticket cleanup 18:12:07 #info none this week. 18:12:07 #topic Upgrading/Migrating Jenkins 18:12:07 #info: Working with Pagure-CI triggers 18:12:08 #info A Pagure SCM Plugin for Jenkins 18:12:21 bstinson: you around? 18:12:35 yes! 18:12:55 * pingou happy to see this topic :) 18:13:19 so i did some investigation over the holidays, and i think the first thing we need to do is to talk about a couple of minor changes to pagure-ci to support newer jenkins versions 18:13:37 namely we need to add a crumb and a CSRF token to the POST requests when adding webhooks 18:14:06 that will allow the existing workflow in the docs to keep working 18:14:50 i was going to explore a patch this week but suddenly kernel patches... 18:15:03 cool. So this is another option? or just a mod to the existing jenkins option? 18:16:33 nirik: this is a change in pagure to change the format of the build triggers slightly 18:16:42 bstinson: do you know if older jenkins will also work? 18:16:53 or do we need to document a minimum jenkins version supported 18:17:16 sure, but now we have: "Type of CI service" and a list with "Jenkins" Do we add a "CentOS CI' there or just leave it at jenkins and just change the urls? 18:17:31 nirik: so there are 2 prongs here 18:17:45 nirik: I think the later, since this is a req for newer jenkins anyway 18:17:46 1.) not breaking existing projects 18:17:49 and 18:18:12 2.) Onboarding new projects 18:18:31 right. and easing migration. 18:19:13 the idea is that we'll stand up a jenkins instance (in the CentOS CI infra) specifically for the miscellaneous projects that don't need their own tenant 18:20:03 to answer pingou's question, we need to break compat for versions less than 2.0 or special-case that in the webhooks 18:20:36 bstinson: we could rename the CI services: Jenkins-1.x and Jenkins-2.x 18:20:44 that would work as well 18:21:24 RE: our previous topic... our jenkins master is f25. ;) So I want to kill it as soon as we can... 18:22:45 ^^ 18:23:12 so we could just mass move everyone to the new instance and ask them to opt out or request their own instance if they require it? 18:23:28 yep, so standing up an instance is easy, i can probably do the "catch-all" one tonight 18:24:12 that way we have something to look at 18:24:39 Sounds good. I guess we will need to tweak urls in pagure side to point to the new instance... along with the changes for 2.0 18:25:09 cool. are there any other items on this topic? 18:25:15 yeah, i can hit up pingou maybe tomorrow (depending on how all these reboots go :) 18:25:44 bstinson: let's do Monday, I'll be offline tomorrow afternoonw 18:25:50 pingou: ack 18:25:58 smooge: last thing is longer-term integration 18:25:58 humm... 18:26:03 bstinson: but if you could drop me an email with how you see both system talk, that would be awesome :) 18:26:21 so some of these projects don't use the pagure.io CI stuff... they use fedmsgs. I guess we will need to convert them... 18:26:31 pingou: certainly, how do you feel about taking a dep on python-jenkins? 18:26:38 bstinson: similarly, if you have a page where the new API is described, maybe I could poke at this Jenkins-2.0 18:26:46 bstinson: it's already there :) 18:27:04 nirik: but these aren't using our jenkins, do they? 18:27:21 nirik: the catch-all instance can be configured with the existing fedmsg plugins 18:27:44 pingou: some are... for example, fedora-comps. 18:28:06 oh 18:28:21 at least I think it's using fedmsg there 18:28:48 there's also auth, but we can migrate the jobs and get them running while we work on configuring that 18:29:31 bstinson: is there some kind of converter for 1.x jobs to 2.x ? 18:30:00 no conversion needed, 2.x will take the configs from 1.x 18:30:09 oh nice. 18:30:15 we'll import, then re-save them to make it a little nicer 18:30:17 except that most of those are defined in jenkins itself 18:30:36 pingou: jenkins exports config.xml that can be directly imported 18:30:40 ok 18:30:45 that we have :) 18:31:10 we can also just take the jenkins filesystem and start 2.x over it as a last resort, but that gets a little hairy 18:31:51 I would say that sounds like a plan C 18:32:49 maybe even a plan F 18:33:09 plan Z 18:33:26 heh basically 18:33:37 or plan 9 from outer space 18:33:37 so what do we need: 18:33:49 1/ a new tenant for all these projects in cico -- bstinson 18:33:57 2/ support for jenkins 2.0+ in pagure -- pingou 18:34:03 2/ fedora slaves -- bstinson 18:34:38 yup good idea :) 18:35:41 4/ a migration plan? 18:35:56 how is jenkins going to report back to pagure? 18:35:59 i didn't see any jobs that do non-normal configs in the EL6/EL7 space, are we ok with doing CentOS for those? 18:36:42 I think so assuming the package/version don't differ too much 18:38:40 ok this all looks good. Anything more on this? 18:39:30 pingou: for reporting back, we'll do the notification plugin (as documented) 18:40:00 the more interesting stuff (fedmsg, conversions to the CI pipeline) can happen later 18:40:26 ok cool 18:40:41 bstinson: let's sync up on Monday, could you send me a meeting invite? 18:40:46 will do 18:40:53 thanks 18:41:34 is there a template to follow for a migration plan? 18:42:06 i can transcribe some of these notes if there's an existing template 18:42:50 not really, but we'll need to get things up, migrate a few projects that volunteer (pagure I'm looking at you) and from there announce things and set a date 18:43:06 bstinson, do you guys have a template? 18:43:15 maybe we could already send an announce for this actually, just stating the will 18:43:41 smooge: i'll gather something up 18:43:58 ok that would be helpful on both sides 18:44:12 to the wiki, infra-list, both? 18:44:22 Let us go with infra-list starting out 18:44:30 cool 18:44:53 after our meeting on Monday, maybe we could also draft a first announcement email :) 18:45:06 that sounds good 18:46:27 #topic enabling default Pagure fedmsg hooks on Fedora DistGit - clime, pingou 18:46:27 #link: https://pagure.io/fedora-infrastructure/issue/6612 18:46:37 ok last item I think we can fit in today 18:46:42 clime pingou? 18:46:43 So we have two git hooks for publishing fedmsg messages 18:46:50 one from dist-git and one from pagure 18:46:55 both having their own format 18:47:00 I will let pingou describe it here 18:47:11 the dist-git one is installed on all main projects by a daily cron job 18:47:19 the pagure one is opt-in from the UI 18:47:41 thing is: the pagure hook is sending pagure.git.receive messages and these are not allowed by our policy 18:47:57 clime wanted to change that and I wanted to run it by you all before :) 18:48:07 I was asking pingou today if he things they can be enabled, ye 18:48:15 *thinks 18:48:19 some reduntncy, but fine with me... 18:48:51 definitely redundant, but that redundancy is basically opt-in :) 18:49:16 it simplifies consuming the messages from multiple pagure instances at once...that's the good thing :) 18:49:29 * nirik nods 18:50:29 ok, so I'll work on this likely tomorrow if not next week early :) 18:50:39 Hmm, if we do that, I'd want to make very sure we blacklist rpms/ etc on pagure.io as namespaces. 18:51:07 puiterwijk: why is that? 18:51:17 The reason being that if we at some point get a script that acts on rpms/$somename messages, and it forgets to check if it comes from src.fp.o, people can trigger it from pagure.io/rpms/$somename 18:51:47 puiterwijk: the topic is different 18:51:52 but the topic should be different for src.fp.o when compared to pagure.io 18:51:57 org.fedoraproject.prod vs io.pagure.prod 18:51:59 ye, pingou faster as usual 18:52:16 clime: but typos, typos, typos everywhere :D 18:52:24 will fedmsg_meta handle the pagure hook messages correctly? 18:52:40 threebean: it should 18:52:49 Okay. And we're sure that any script will pick that, rather than what I've also seen as: if '$somesuffix' in topic: 18:53:01 threebean: https://github.com/fedora-infra/fedmsg_meta_fedora_infrastructure/blob/develop/fedmsg_meta_fedora_infrastructure/pagure.py#L551 18:53:25 I've seen a few rules in fedmsg_meta_fedora_infra that don't read the entire topic, but rather do things like 'compose.complete' in topic 18:53:57 puiterwijk: unles they are listening to '*' as topic, they will likely list the topics they want to act on 18:54:00 pingou: cool. 18:54:27 puiterwijk: they all do, but fedmsg_meta checks this before hand 18:54:41 using the __name__ and topic_prefix_re 18:54:47 Okay 18:55:02 Just wanted to make sure that that's taken into account... 18:55:19 fedmsg_meta I fairly confident 18:55:26 other scripts... :s 18:55:32 * puiterwijk would hate to find a script that'll rebuild the kernel as soon as $someuser commits to rpms/kernel, only to find it rebuilds it after someone pushes to pagure.io/rpms/kernel... 18:55:41 Yeah, I'm afraid for the other scripts... 18:55:53 Which is why I would much rather just stay safe and just blacklist it for now. 18:56:23 it's easily doable anyway 18:56:26 But if you say it's sufficiently covered, sure. 18:57:03 BLACKLISTED_GROUPS = ['forks', 'group'] 18:57:17 Yeah 18:57:19 -BLACKLISTED_GROUPS = ['forks', 'group'] 18:57:21 +BLACKLISTED_GROUPS = ['forks', 'group', 'rpms'] 18:57:30 pingou: +containers, +tests, etc :) 18:57:50 ... at what point does a white list make sense :) 18:57:51 And I'm not saying that that blacklist should stay there permanently, or that we must do that. I just want it as long as we're not 100% sure 18:58:07 smooge: the list of things to blacklist is small :) 18:58:16 I am not liking the blacklist personally 18:58:18 I think it's like 4 or 5? 18:58:42 rpms, modules, container, tests 18:58:43 clime: why not, and how would you then alleviate my concern, given that the "X in topic" method is used by people? 18:59:25 but it doesn't really depend on me. Well, thing is when you introduce the blacklist, it then stays there like permanently 18:59:27 clime: it just means we can't have a group called "rpms" in pagure.io... what's the deal? 18:59:29 I guess without a blacklist it just becomes that script authors problem. ;) 18:59:36 clime: doesn't have to be permanent 18:59:54 clime: I'm just saying I want this until we have another way to assure that it doesn't happen 19:00:00 * nirik is fine blacklisting those namespaces. I think they might be confusing anyhow. 19:00:04 well, not sure what will be the point when someone decides to clean the problems in the scripts 19:00:18 just mentioning... 19:01:50 ok I don't see us moving on this at the moment. 19:02:03 and I think there is another meeting starting (now/soon?) 19:02:19 Hi all - quick introduction as I know we're running out of time - Interested in helping out with sysadmin tasks and joining apprentice program. FAS username: timjcasey. Have been an Ambassador for quite some time now, will be updating my wiki page later today with more info on my background etc. 19:02:31 hi timc 19:02:37 welcome timc 19:02:38 sorry if I missed the intro section at the start :) Was a bit late. 19:02:50 its 6am here - hard to wake up so early :D 19:03:17 welcome timc 19:03:27 thanks all 19:04:02 welcome timc :) 19:04:12 #topic Open Floor 19:04:22 so I think for now we can enable that topic, and blacklist these groups 19:04:24 mainly you always need thing about the blacklist when you add a new namespace at src.fp.o... 19:04:25 and reconsider later :) 19:04:28 timc, please join #fedora-admin after this and we can set up the account items 19:04:31 sry for late argument 19:04:57 ok lets take the blacklist/whitelist/greylist to infrastructure or proper list 19:05:00 smooge: no probs, already there 19:05:08 are there any open floor items? 19:06:21 * pingou has none 19:06:32 ok in that case... thank you all for coming. 19:06:41 thanks smooge and everyone 19:06:44 #endmeeting