18:00:36 #startmeeting Infrastructure (2017-09-28) 18:00:36 Meeting started Thu Sep 28 18:00:36 2017 UTC. The chair is relrod. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:36 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:36 The meeting name has been set to 'infrastructure_(2017-09-28)' 18:00:36 #meetingname infrastructure 18:00:36 The meeting name has been set to 'infrastructure' 18:00:36 #topic aloha 18:00:36 #chair smooge relrod nirik abadger1999 dgilmore threebean pingou puiterwijk pbrobinson maxamillion 18:00:36 Current chairs: abadger1999 dgilmore maxamillion nirik pbrobinson pingou puiterwijk relrod smooge threebean 18:00:48 Hai 18:00:50 Hi people, happy Thursday 18:00:50 morning 18:00:55 Hello 18:00:59 hi 18:01:02 .hello 18:01:02 netcronin: (hello ) -- Alias for "hellomynameis $1". 18:01:03 .hello2 18:01:06 bowlofeggs: bowlofeggs 'Randy Barlow' 18:01:13 Happy Fedora 27 Beta GO day 18:01:20 \o/ 18:01:25 nd33d 18:01:27 * cverna waves to the crowd 18:01:36 the crowd waves back 18:02:16 hello 18:02:28 hi 18:02:47 #topic New folks introductions 18:02:48 #info This is a place where people who are interested in Fedora Infrastructure can introduce themselves 18:03:00 * bowlofeggs offers smooge a warm la croix 18:03:01 Any newcomers who would like to introduce themselves? 18:04:06 Don't be shy 18:04:19 My name Bennie. Live and work in Pretoria, South Africa. Interests are networking, systems administration and development/packaginh 18:04:35 welcome benniej! 18:04:39 welcome benniej 18:04:39 cool, welcome! 18:04:41 Thnx 18:04:43 welcome benniej 18:05:19 * threebean waves 18:06:00 welcome 18:06:01 benniej: hang around in #fedora-admin after the meeting and we can help you get started/get into the fi-apprentice group so you can start looking around. 18:06:19 awesome thnx 18:06:51 okay, moving on to announcements 18:06:56 #topic announcements and information 18:06:56 #info PHX2 Colo Trip, Dec 4th - 9th 18:06:56 #info Infrastructure will be down during that trip. 18:06:56 #info beta freeze still going on. 18:06:57 #info Moving .stg. ips is ongoing this week - smooge 18:06:58 #info .stg. vh 16 and vh20 moved over .. broke lots of things 18:06:59 #info fedora-packages, looking at a way to replace xappy (dead upstream) - cverna 18:07:28 smooge: so whats left for stg? vh11? or more? 18:07:40 and cverna: whats the packages status currently? :) 18:08:04 we have multiple boxes 18:08:52 nirik, vh11 vh04 18:08:57 about 30 guests 18:09:05 and god knows how many hard coded ips 18:09:24 ok. happy to help, and thanks for working on it 18:09:29 To be able to get packages to rhel7 we need to get rid of xappy which is used by the indexer to populate teh xapian db 18:09:35 happy to help as well 18:09:46 Same. If I get done with homework this weekend I'll hop on. 18:09:59 I just uploaded another patch to the ticket with the IPs I changed. 18:10:00 the ones I have skipped on purpose at the moment is proxy01, koji01, pkgs01, and pgpdr 18:10:21 cverna: yeah. ;( Another bug got filed today... the 'current releases' isn't updating. 18:10:21 I will try to use plain xapian bindings since the rest of the app works with them. 18:10:38 nirik: I have seen 18:11:12 * bowlofeggs switches bodhi's configs to use IP addresses for other services instead of domain names… 18:11:20 I ll try to have some ready for unfreeze time 18:11:33 nirik, relrod netcronin if you want to also stage the patches for the host_vars for vh04 that would be good. A master playbook then needs to be run on all of staging and then a test to make sure ansible did what it said it would do 18:11:56 s/some/something working 18:12:27 cverna: sounds great. Thanks for working on this. So likely it's not worth the el6 version in stg now? since we hopefully will have el7 soon? 18:12:37 smooge: Is vh04 an actual hostname? 18:12:42 virthost04 18:12:44 smooge: ok. grep the /etc/hosts files on all stg hosts for hostnames of all boxes changed, too. 18:12:46 I think 18:12:54 relrod, I am working on that 18:12:56 Okay, I'll try. 18:13:09 that is my patch while netcronin does the host_vars 18:13:21 smooge: ok :) 18:13:39 there is also changes in postgres database files and changes in iptables that will need to be dealt with 18:13:57 yep 18:14:04 nirik: the latest packages will not work on el6 anymore :P , I have remove the python-appstream dependencies 18:14:05 I can move pgbdr... 18:14:16 cverna: ok, fair enough 18:15:00 Okay, any other comments on the announcements, or things people would *like* to announce? 18:15:11 we got a beta maybe? 18:15:27 oh, yeah 18:15:31 * cverna will break his teeth on xapian over the week end :) 18:15:36 #info Beta GO! woot! 18:15:44 \o/ 18:16:19 alright 18:16:25 #topic Moving apps into OpenShift 18:16:37 I think this is leftover from last week... ;) 18:16:48 no one has had cycles to work on it yet 18:17:15 yeah. I'd like to start on some of our smaller apps, in stg (fedocal comes to mind. Possibly elections). I will start on that next week 18:18:00 doing some small apps would be nice to help us develop some guidelines 18:18:06 * relrod nods 18:18:15 but eventually i'd like to try bodhi in it, once others have gotten the infra feet wet :) 18:18:19 * threebean nods 18:18:31 i can't see myself having time to focus on that myself in the near future 18:18:43 side note: we still need to update our prod setup after freeze sometime 18:19:06 mjia is working on putting out an update to waiverdb and greenwave this week. 18:19:14 which might also involve just reinstalling it, because we may need to shuffle things around for some new drives. 18:19:19 getting a feel for what sticking points there are in updating a service. 18:19:26 yeah. I can do that. We'll need to announce an outage, but if it goes as smoothly as the stg one did, it shouldn't be very hard at all 18:19:45 (of course now I probably jinxed it :P) 18:20:49 threebean: sounds good. 18:20:55 bowlofeggs: Well, you gave me a good segue for the Bodhi stuff (I'll go back up to Ticket cleanup after) 18:20:58 #topic Can Bodhi move to be hosted on Fedora instead of Fedora and EL 7? 18:20:59 #info The version of Pungi in EPEL 7 is quite old, and is causing some difficulties on https://github.com/fedora-infra/bodhi/pull/1697 18:21:06 bowlofeggs: all yours 18:21:50 heyo! 18:22:05 ok, so there is a pull request to add pungi mashing to bodhi that's open right now 18:22:09 linked in that #info above 18:22:33 unfortunately, it is using internal pungi APIs that are only available on newer versions of pungi 18:22:45 and the version of pungi available on EL 7 doesn't have those APIs 18:22:47 well, bodhi-backend01 is already fedora. :) 18:22:51 indeed! 18:23:02 but, bodhi currently fully supports EL 7 and Fedora 18:23:12 (keep in mind that there are other orgs that use bodhi too) 18:23:30 and bodhi's tests currently all pass and have 90% coverage on EL 7 and fedora 18:23:33 bowlofeggs: so, how much promises do we have that the APIs they're using are going to continue to exist? If it's internal, that would be not promised 18:23:46 puiterwijk: i don't think we have promises 18:23:51 is there any chance that pungi is ok on epel7, they just didn't build it yet? 18:23:58 nirik: I believe that's the case. 18:24:01 puiterwijk: but, i think mcurlej is working to get an API added to pungi that bodhi could use 18:24:03 nirik: "in theory" 18:24:42 the tests kinda bother me, because bodhi's CI currently measures test coverage, but only runs on EL 7 today 18:24:57 bowlofeggs: and I'm going to assume that we're waiting with going to prod until we get stuff to use non-internal APIs? 18:25:06 i kinda just feel conflicted about various options i can see, and wanted feedback from others 18:25:22 Anyway. Given that there's already Bodhi boxes on Fedora, I guess that that part is "fine" 18:25:42 puiterwijk: i have a feeling that that current plan won't allow time for that to happen 18:25:44 I have no objection to just switching all bodhi hosts to fedora if desired... but imho if we can keep supporting epel7 without great burden we should try to 18:25:58 yeah i feel conflicted about switching it all to fedora 18:25:58 Agreed to nirik. 18:26:11 on one hand, it works today and dropping support feels "bad" 18:26:17 especially since i don't know how others deploy it 18:26:26 does pungi even have a published api? 18:26:30 but on the other hand, it does make this difficult 18:26:32 nirik: it does not 18:26:35 nirik: nope 18:26:48 nirik: but i think martin is trying to add one too 18:26:52 Which means they make no promises as to whether the APIs bodhi is going to call exist in 2 weeks 18:27:02 threebean: is that still the plan? for martin to get this into pungi as a supported API? 18:27:11 puiterwijk: correct 18:27:16 he's working on it, yup. 18:27:18 sure. 18:27:50 so, for infra, i guess we could carry our own pungi package that's newer than EL 7 has? but then we have to stay on top of updates... 18:28:06 and that might not be great for third party bodhi users, depending on how they deploy (unknown) 18:28:11 I think first step is to ask lseldar if he can build for epel7 18:28:17 * threebean nods 18:28:21 yeah that's an option too 18:28:30 i guess that would be nicest 18:28:53 if I had to bet, I'd say the hurdle is going to be python-kobo. ;) 18:29:01 hum... 18:29:07 epel7 currently has pungi3 18:29:16 ie, ancient 18:30:07 anotehr crazy idea i had was trying to get some CI setup that can merge two coverage reports into an "uber coverage" report 18:30:17 and then have a container to run the tests in F25+ too 18:30:31 and then make sure we hit the new masher lines in at least one of the test runs 18:30:48 and then skip tests related to pungi when running on EL 7 18:30:53 but that seems pretty complicated 18:31:08 and also still drops support for EL 7 for mashing, which again i don't know if others use today 18:31:22 well, only for rpm mashing atm. 18:31:28 meh just drop it 18:31:58 oh, side note: bodhi-backend01* are f25, should sometime take them to f26. 18:32:06 +1 18:32:34 so i'd love to hear more ideas if anyone has them, but it sounds like asking lsedlar about getting newer pungi in EL 7 might be a good place to start? 18:32:52 or, do we think just moving all of bodhi to F26 is better? 18:32:58 or something else? 18:33:34 perhaps dgilmore knows: is epel7 supposed to stay on pungi3 for some reason, or can it go to 4? 18:33:35 bowlofeggs: I don't think a new pungi would be in policy. Since the pungi3 -> pungi4 upgrade is probably pretty huge, so would need an EPEL exception 18:34:08 puiterwijk: sure, but pungi3 isn't maintained anymore and I bet few people use it... so it could be done 18:34:12 (I'm assuming that they've changed input-output) 18:34:43 they are... completely different. 18:34:58 there is a policy that can be used to make backwards incompatible upgrades in EPEL that could be investigated, but i did it once and it does take some time (reasonably so) 18:35:19 Right. It just takes a while 18:35:35 nirik: we could move it to 4 18:35:42 so one problem is that we want to deploy this on the monday after the beta (so, like, 10-11 days?) 18:36:05 nirik: we had some patches from scientific linux in the 3.x branch 18:36:15 well, i should say that the modularity folks want to deploy it then. i would love to have more time to do things more cleanly than this. 18:36:18 but at this point I think moving it makes sense 18:36:57 well, if it's just a tide over we could have a pungi4 in infra tags 18:37:03 moving it sounds good, but i doubt we can do that via the exception policy inside of 10 days 18:37:07 I had attempted a pungi4 build for epel 18:37:08 (until the epel one is out) 18:37:09 https://koji.fedoraproject.org/koji/buildinfo?buildID=859976 18:37:23 there was something missing 18:37:38 nirik: true 18:37:44 possibly the python-kobo threebean mentioned 18:38:39 nirik: bowlofeggs: we will have to run pungi on fedora 18:38:51 in order to get rich deps etc 18:38:53 sure. 18:39:01 yeah we do plan to keep doing that 18:39:01 so perhaps we should just let el7 go 18:39:26 nirik: as far as pushing updates in fedora/epel, yeah 18:39:28 to be honest, the webob stuff in EL 7 does also cause me some pain from time to time 18:39:56 i've always been able to work around those pains, but it has slowed me down before on more than one occasion 18:40:16 so moving bodhi to pure fedora might make those pains lessen for me (but cause other pains, like more frequent upgrades) 18:40:17 there are reasons why we can not fully support RHEL for our use cases 18:40:57 well, I guess the question is: are any of the other bodhi users using on el7 only? 18:41:16 so i personally think it would make my life a little nicer if bodhi was pure fedora, but i also know that would be more work for systems people so i want to make sure that wouldn't bother anyone 18:41:42 nirik: yeah i also have that question and i really don't know the answer 18:42:25 on infra end I don't mind switching them all to fedora. The web frontends seem kinda overkill to be fedora, but it should be fine. 18:42:27 bowlofeggs: that side of things is less of a problem, I'd say, given that we have multiple Fedora machines that we need to rebuild regularly anyway 18:42:39 ah ok 18:42:51 we should just make everything rawhide 18:43:02 rollin rollin rollin... 18:43:12 well, it would make my life a little easier and i think it could help with the problem at hand too 18:43:19 any opposed? 18:43:28 Not opposed to making them Fedora, no 18:44:40 ok, well thanks :) 18:44:51 i think that satisfies what i wanted to talk about 18:45:03 okay 18:45:19 So I'll move back up to: 18:45:21 #topic Ticket cleanup 18:46:15 I'm not really sure what this is supposed to be/who added it, but -- is anyone working on tickets that they need to discuss or need help with? 18:47:10 well, I was going to pick some old ones to discuss and move forward... 18:47:14 but then I didn't. ;) 18:47:22 I can look real quick 18:48:09 .title https://pagure.io/fedora-infrastructure/issue/2079 18:48:09 nirik: Issue #2079: querying the list of all the aliases that a given FAS id is on - fedora-infrastructure - Pagure 18:48:23 So, this has been around about 7 years. 18:48:36 can we think of any way to actually implement it? 18:49:11 I guess a grep of /etc/aliases /etc/postfix/* on bastion01 might... 18:49:13 nirik: well, "so that I can request to be removed from ones that I don't need to be on." - people can do a git grep through the ansible repo 18:50:10 right. So, perhaps we close and add a howto clone ansible repo, git grep and look for aliases specifically? 18:51:28 any objections? anyone want to do that? (I can if no one wants to) 18:52:09 ok, shall we do another or call it good for today? 18:52:24 nirik: let's close all the tickets? :) 18:52:47 sure, by fixing/addressing them. ;) 18:52:58 here's another old one: 18:53:15 .title https://pagure.io/fedora-infrastructure/issue/2888 18:53:15 nirik: Issue #2888: Implement HTTP Strict Transport Security - fedora-infrastructure - Pagure 18:53:39 That one's moved along step by step 18:53:44 the only thing left here I think is to redirect http/https pkgs to src via iptables... I am not sure how well that will work 18:54:04 Let's not do it via iptables, but via apache. 18:54:08 That is simple enough to do. 18:54:34 it can't be a redirect tho... that will fail 18:54:57 well, http can be... but not https 18:55:26 and upload.cgi still needs to work 18:56:07 anyhow, we are low on time, can discuss more outside meeting 18:56:38 yeah, only a few minutes left, so... I'm going to merge Apprentice office hours and regular open floor into one: 18:56:47 #topic Open Floor and Apprentice Questions 18:57:08 Anyone have anything they'd like to discuss, or apprentices have any quick questions they'd like to ask? 18:57:45 All good here 18:57:52 Same. Working with the stg IP move. 18:58:04 netcronin: *nod* - thanks for your help with all that! 18:58:07 Feel free to ping me on any issues you guys need help with that you think apprentices could tackle. 18:58:14 relrod: no problem. ): 18:58:16 :) 19:00:26 okay, well, thanks all! 19:00:30 #endmeeting