15:36:37 <dgilmore> #startmeeting RELENG (2014-06-23)
15:36:37 <zodbot> Meeting started Mon Jun 23 15:36:37 2014 UTC.  The chair is dgilmore. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:36:37 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:36:44 <dgilmore> #meetingname releng
15:36:44 <zodbot> The meeting name has been set to 'releng'
15:36:44 <dgilmore> #chair dgilmore nirik tyll sharkcz bochecha masta
15:36:44 <zodbot> Current chairs: bochecha dgilmore masta nirik sharkcz tyll
15:36:45 <dgilmore> #topic init process
15:36:58 <nirik> morning
15:37:09 <tyll> hi
15:37:46 <masta> hello
15:39:21 <dgilmore> hey all
15:40:19 <dgilmore> lets get going
15:40:42 <dgilmore> #topic #5870 rawhide signing
15:40:48 <dgilmore> https://fedorahosted.org/rel-eng/ticket/5870
15:40:57 <dgilmore> tyll: what is t he current status?
15:41:04 <dgilmore> nirik: did we get the ports open?
15:41:27 <nirik> yes, the ticket was marked done.
15:41:28 <tyll> after I changed the currint WIP script to sign only one build at a time it did not hang anymore
15:41:37 <nirik> if it's not working, let me know and I can reopen. ;)
15:41:51 <dgilmore> tyll: yeah. thats the only known sure fire way
15:42:05 <tyll> currently a test environment to add code to move builds is missing
15:42:55 <tyll> the signing of one RPM at a time was too slow for the mass rebuild btw., especially to sign e.g. texlive
15:43:14 <tyll> I was not yet able to tst whether I can use secondary sigul now
15:43:42 <dgilmore> tyll: okay
15:43:50 <tyll> nc -v 10.5.124.145 44333 does not work
15:43:59 <dgilmore> tyll: and yeah signing mass rebuilds is likely always going to be too slwo
15:44:03 <dgilmore> slow
15:44:19 <dgilmore> tyll: i think to sign all of rawhide takes about 5 or 6 days
15:44:26 <dgilmore> and the mass rebuild takes 2
15:44:34 <nirik> tyll: hum. Might need external.
15:44:39 <nirik> we can debug out of meeting
15:44:58 <tyll> nirik: ok
15:45:15 <tyll> what do we do about the gating tag?
15:45:40 <tyll> can we setup something on live koji that does not interfer with composing rawhide?
15:46:41 <dgilmore> tyll: we kinda would need to deploy it to test it
15:49:03 <tyll> My idea (not sure if possible) would be to move builds to two tags automatically (f21 and f21-unsigned) for now and let the script only untag builds from f21-unsigned if something was signed, the tagging into f21 would be a no-op then - then for the final version only a minor change is needed
15:49:03 * nirik nods.
15:49:45 <dgilmore> tyll: well koji will only tag into one tag when you build
15:49:59 <dgilmore> you would need to run something external to tag into a second one
15:51:02 <tyll> hm, then it might be good enough to just create the f21-unsigned tag and tag manually builds into this to test the script
15:51:54 <nirik> since rawhide is 1x/day I think we should be ok to switch it one day when we are ready and watch it to make sure it's good.
15:53:30 <dgilmore> yeah
15:53:54 <masta> so make koji point to f21-unsigned for a day?
15:55:10 <dgilmore> at least for a few hours and test
15:55:11 <nirik> well, since we have a day to fix things before the next rawhide compose, we should be able to switch it over someday when we are ready
15:55:23 <tyll> I just looked again at the code, it would already be very helpful if f21-unsigned existed already an I can just tag builds into it manually - the script just watches for tagging fedmsg messages
15:56:22 <dgilmore> tyll: okay
15:56:31 <dgilmore> we can get that setup after the meeting
15:56:48 <dgilmore> tyll: it will be a bit more difficult on secondaries
15:57:01 <tyll> dgilmore: ok
15:57:11 <dgilmore> as the way koji-shadow works it might not be so simple
15:57:49 <tyll> dgilmore: I guess if the script is feature complete for primary koji, it should be easier to adopt to secondaries
15:59:10 <masta> I would imagine we just mirror the tag structure
15:59:12 <dgilmore> tyll: yeah, we may need to work out a new workflow for them
15:59:21 <dgilmore> masta: we do that
15:59:59 <dgilmore> its more down to koji-shadow tags into f21, I guess we could try f21-unsigned used
16:04:35 <dgilmore> okay lest move on
16:04:54 <dgilmore> #topic #5914  Move fedmsg based blocking service to Fedora Infrastructure
16:05:01 <dgilmore> https://fedorahosted.org/rel-eng/ticket/5914
16:05:13 <dgilmore> tyll: what do we need to do here?
16:06:47 <tyll> I need to clean up the script a little and make it into a service. Also we need to write down on which host it should run (releng04?).
16:06:55 <nirik> yeah, probibly releng04.
16:07:10 <dgilmore> i think releng04 yeah
16:07:22 <dgilmore> just need to decide what key to use for koji
16:07:24 <tyll> The latest changes I did were adjusting it to pkgdb2, so it is at least working again (it was not when we first discussed it)
16:07:28 <dgilmore> maybe masher
16:07:36 <dgilmore> or make a generic releng one
16:07:45 <dgilmore> okay cool
16:08:28 <tyll> btw. should it also block packages for EPEL? Currently it does not iirc
16:08:47 <dgilmore> well epel is kinda tricky
16:09:06 <dgilmore> some packages need blocked in the epel tags, but unblocked in the -build ones
16:09:23 <dgilmore> as the package has moved to rhel
16:09:39 <dgilmore> in which case we likely should actually yuntag all builds
16:09:41 <dgilmore> untag
16:09:58 <dgilmore> to make sure that rhel and not epel builds are used in the buildroots
16:10:04 <tyll> we could block them if they are retired and ask to open a ticket if it just moved to RHEL
16:11:47 <dgilmore> id rather not
16:11:53 <dgilmore> it could break building things
16:12:06 <nirik> could we always block in epel and unblock in build?
16:12:18 <nirik> if it's blocked in epel none of the old builds should get used right?
16:12:19 <dgilmore> I guess we could always untag and unblock in the -build tag
16:12:40 <dgilmore> nirik: they still show up as latest in the -build tag
16:12:52 <dgilmore> and ive never checked the actual repo to see whats used
16:13:35 <nirik> hum, ok.
16:14:26 <dgilmore> in which case we likely should actually yuntag all builds build is used
16:14:29 <dgilmore> gahh
16:14:34 <dgilmore> i suspect that the epel build is used
16:14:43 <nirik> humf. ;(
16:15:02 <dgilmore> anything else here?
16:15:05 <nirik> if we untag them all they would also get garbage collected right? I guess thats ok...
16:15:19 <dgilmore> we should work out the best course of action for epel
16:15:24 <dgilmore> nirik: right they will
16:15:57 <tyll> if it is clear what needs to be done for EPEL I'll implement it
16:16:18 <tyll> but for now I continue to skip it
16:16:18 <nirik> well, it's clear what we want, but not sure how to do it. ;)
16:16:29 <dgilmore> tyll: well I think what needs doing is all builds untagged
16:16:35 <dgilmore> perhaps tagged into a holding tag of sorts
16:16:37 <nirik> I guess we could always block in epel, unblock in build and untag all builds
16:16:54 <dgilmore> the package blocked in say epel-7
16:17:02 <dgilmore> then unblocked in epel-7-build
16:18:22 <tyll> should/can it be unblocked in epel-7-build if it did not move to RHEL?
16:18:29 <dgilmore> lets move on and talk about this after
16:18:41 <dgilmore> tyll: it will not hurt
16:18:47 <nirik> well, it wouldn't cause problems... just be extra
16:18:54 <dgilmore> as there will be nothing to satisfy the dep
16:19:54 <tyll> ok
16:20:39 <dgilmore> lets move on
16:20:42 <dgilmore> #topic Secondary Architectures updates
16:20:43 <dgilmore> #topic Secondary Architectures update - ppc
16:20:56 <dgilmore> masta: your up
16:21:01 <dgilmore> you're
16:21:46 <dgilmore> I guess masta is not here
16:22:12 <dgilmore> so ppc64le and ppc64 are both now being built for rawhide in the main ppc hub
16:22:21 <dgilmore> mash needs updated
16:22:36 <dgilmore> and we may need to do some work to add the new arch to mirrormanager
16:22:44 <dgilmore> they have started the mass rebuild
16:22:48 <nirik> cool.
16:23:07 <dgilmore> hopefully within a week the le hub can be gone
16:23:25 <dgilmore> its still there with a builder in case some one off builds are needed
16:23:26 <nirik> excellent.
16:23:50 <masta> sry, was running back from the kitchen with coffee
16:24:53 <masta> so I'm still getting used to dwa's work flow, I've learned some things about releng in generally that were not known to me before, so it's good times
16:25:25 <masta> that's about it!
16:25:40 <masta> not too sure what the status is besides that ;-)
16:26:06 <dgilmore> #topic Secondary Architectures update - s390
16:26:15 <dgilmore> no sharkcz
16:26:27 <dgilmore> #topic Secondary Architectures update - arm
16:26:40 <dgilmore> aarch64 mass rebuild is almost 100% done
16:26:59 <masta> =)
16:27:32 <nirik> excellent
16:27:47 <dgilmore> hopefully soon a 100% fedora will run on actual hardware
16:28:00 <dgilmore> we will need to deal with comosing etc
16:28:15 <masta> yeah
16:28:42 <dgilmore> composing
16:28:49 <dgilmore> its in pretty decent shape
16:29:00 <dgilmore> #topic Open Floor
16:29:10 <dgilmore> anyone got anything to bring up?
16:29:10 <masta> I still use lmc to generate foundation images, but it seems lmc is falling out of favor ... we just need to get something working
16:29:39 <dgilmore> masta: lmc has never really been in favour as its never been useable for releng purposes
16:29:43 <dgilmore> we wanted to use it
16:29:53 <dgilmore> but the anaconda guys designed something we couldnt use
16:29:54 <nirik> dgilmore: only releng thing I can think of... we should switch epel7 to use rhel7 final
16:30:05 <dgilmore> nirik: we need to just do that
16:30:07 <masta> nirik: yes
16:30:09 <nirik> yeah.
16:30:15 <nirik> it's ready, just needs koji repointed.
16:30:33 <dgilmore> nirik: okay. get me the urls and ill do it
16:30:46 <nirik> dgilmore: can do.
16:30:57 <tyll> I wonder if notting is/will be still part of rel-eng, since he moves/moved from Red Hat to Ansible afaik
16:31:04 <masta> I noticed that in Fedora we do not have the koji runroot plugin, have we ever thought about bringing that in?
16:31:17 <dgilmore> tyll: hes really only been honoury for years now
16:31:25 <nirik> tyll: yeah, likely he won't have time...
16:31:26 <dgilmore> masta: its not open
16:31:42 <dgilmore> masta: its a propietory red hat thing and cant be discussed here
16:31:54 <masta> dgilmore: that is unfortunate, it has value I'd say. otherwise we have to make custom tasks I would imagine?
16:31:57 <nirik> we also have the mash in koji plugin thats similar and open.
16:32:27 <masta> dgilmore: my bad, sry...
16:32:30 <dgilmore> masta: there is a version intended to be opened but its not gone through review to do so
16:32:31 * masta zips it
16:33:33 <tyll> to keep things clean, it might be a good idea to remove notting from the wiki and maybe privileged groups
16:33:42 <dgilmore> tyll: i guess we should ask notting if we should remove him
16:33:51 <tyll> dgilmore: sounds good
16:33:57 <dgilmore> im not 100% sure what groups he is in
16:34:11 <tyll> it's a log
16:34:12 <masta> it is a community project, I guess he could choose to stay, right?
16:34:13 <tyll> lot
16:34:43 <dgilmore> tyll: im sure
16:35:07 <dgilmore> masta: it is and he is welcome to stay, though realistically he has not been active in releng in over 5 years i think
16:35:18 <tyll> masta: I am not saying to close his account, but IMHO only active accounts should be in privileged groups to reduce the risk of account break-ins
16:35:27 <dgilmore> the tools he was maintaining he handed over to me
16:35:38 <masta> tyll: that makes sense +1
16:35:44 <nirik> his account is active...
16:35:57 <dgilmore> he is still staying active
16:37:10 <dgilmore> just pinged him on irc and he has indicated he doesnt think he needs to
16:37:50 <dgilmore> ill work to remove him
16:37:55 <tyll> ok
16:38:22 <masta> geez... it's like some kind of ex dues machina, changing of the gods... err.. cycling of rel-engineers
16:38:33 <masta> =(
16:39:52 <dgilmore> okay if noting else
16:39:55 <masta> so there was something about fedora-release packages
16:39:56 <dgilmore> lets wrap up
16:40:04 <masta> err.. sub-packages
16:40:11 <masta> or stand alone
16:40:25 <dgilmore> i need to write up an email
16:40:32 <dgilmore> I talked to mattdm friday on the phone
16:40:46 <dgilmore> release alone
16:41:00 <masta> we can take it to the main chan if we need to wrap, was just struggling to think of the emails over the past week or so... thought of that near the last minute
16:41:05 <dgilmore> thgere is about 4 or 5 ways to do it all
16:41:22 <dgilmore> and we need to work out the one thats the most manageable
16:41:36 <masta> yep
16:41:58 <masta> the yicky part is that they were talking about using alternatives
16:42:26 <masta> which implies they want to clobber each other's files
16:42:35 <masta> sigh
16:42:45 <masta> ok.. will look forward to the email
16:44:50 <dgilmore> okay lest wrap up
16:44:55 <dgilmore> #endmeeting