From Fedora Project Wiki

Fedora Release Engineering Meeting :: Monday 2008-04-07

Fedora 9 Readiness

  • Final freeze on Tuesday, April 8, 2008
  • Internet2 issue could affect ability to feed mirrors at F9 GA
  • Ongoing review of blocker bugs until GA


IRC Transcript

f13ping: notting jeremy jwb wwoods warren lmacken poelcat rdieter13:01
* poelcat here13:02
f13(wwoods is out)13:02
* notting is here13:02
* jwb is here13:02
* lmacken is here13:02
* warren here13:02
f13aight, lets get started.13:03
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting - Fedora 9 Final Freeze13:03
f13The final freeze is tomorrow, and as usual I'm going to setup an at job to throw the switch tonight so tomorrow's rawhide is the frozen content.13:03
f13I've got f9-final tag in koji already, and I'm updating it with the latest builds from dist-f913:04
nottingwe're having an early tuesday freeze, not a late tuesday freeze?13:04
f13I'm going to start signing tonight too, if the clone finishes in time.13:04
f13notting: yeah, we did that for Alpha and Beta13:04
f13gives releng an extra day of testing13:04
warrenhow are we feeling about the stability of rawhide?  I had some worrying crashes during the weekend (no compiz involved)13:04
f13warren: I'm not "happy" but I'm not slitting my wrists yet.13:05
poelcattoday's rawhide was good for me http://fedoraproject.org/wiki/JohnPoelstra/RawhideToday13:05
jwbrawhide has been fine for me since beta13:05
jwbwith the exception of hibernate13:06
jlaskawarren: I think it largely depends on what components you're interested in13:06
f13Well, I triggered a blocker bug this morning in anaconda, so that's lame for me13:06
f13and I still have to plow through this upgrade stuff.13:06
warrenjlaska: oh nothing important, just dbus-daemon crashing, X deadlocking, and kernel oopsing13:07
f13upgrades in general are scary.13:07
jwbf13, what upgrade stuff?13:07
jlaskadetails details ;)13:07
f13jwb: finding things that went from multilib to not multilib and getting them blacklisted correctly in anaconda13:07
jwbah13:07
f13jwb: plus finding other upgrade gotchas.13:07
nottingi am concerned about our ability to deliver the preview, in light of mdomsch's discussion about final13:07
f13there are a number of upgrade related bugs on teh blocker list.13:07
jwbnotting, wait... what?13:07
f13notting: preview is just going ot be torrent only13:07
nottingah, ok13:07
jlaskaf13: so dumb question ... once we freeze, the plan is then to widdle down the blocker bug list?13:07
nottingdoom postponed!13:08
f13notting: it's like the regularly scheduled snapshot, just with a differen tname.13:08
nottingok. i remain concerned about final, though. yuck.13:08
f13jlaska: yeah, I think each day we'll probably trim a bit more off the blocker list, and get more toward "we'd actually slip the release for this"13:08
* rdieter here (a little)13:08
f13notting: as do I.13:08
f13for those not in the "know"13:08
poelcatf13: how will we test out jigdo before GA?13:08
f13RH can't currently deliver any content over Internet2 to anybody13:08
f13which means our I2 mirrors get way behind13:09
warrenAl Gore didn't invent it yet.13:09
jwboh13:09
f13and all have ot use phx resources to get content, which is causing resource contention13:09
* jds2001 sits in the back13:09
f13resources which are supposed to be shut off for releases due to not wanting to knock redhat.com off the internet again13:09
f13poelcat: I'm thinking of testing it by hand a bit with some custom mirrormanager urls that I'll ask mdomsch to create for me.13:10
nottingdo we have everything in place needed for preupgrade?13:10
poelcatf13: let me know if you want some help from a 'home user' :)13:10
f13notting: I have no idea on that front.  I think we're awefully late to be landing "features" such as that.13:10
f13poelcat: will do, likely the mirrormanager url will just reference rawhide13:11
f13Part of the final freeze is also the mass CVS branching13:12
f13I'm not entirely sure why the schedule has it on the 9th, I probably asked for it and promptly forgot why13:12
f13but abadger1999 needs some time to try and recover some optimizations on the pkgdb/fas side of things to make the branching go faster.13:13
f13it's still probbaly going to require a cvs outage, and a mail shutoff to keep from mailbombing everybody13:13
abadger1999also note, I think the pure cvs side of things has slowed down as well.13:14
jwbf13, maybe you were still thinking it was going to be a late tues freeze, so you wanted it on the 9th?13:14
f13I had a script at one point to do it, but that was pre fas2, and some other things have changed on pkgdb side, so I'll likely have to rewrite the script.13:14
f13jwb: could be.13:14
nottingabadger1999: it's not like modules ever gets *smaller*13:14
abadger1999Right.13:14
f13abadger1999: curious, if the cvs slowdown is the ACL lookup + the "whom do I mail" lookup13:14
f13which I would think with an outage we could kill both of those13:15
abadger1999Esp. the whom do I mail would speed things up.13:16
* abadger1999 writes a note to disable getnotifylist13:16
warrenabadger1999: if we have a cvs outage, might we use that as an opportunity to finally rename cvsextras?13:16
abadger1999warren: You and f13 should work it out.  I can make it happen whenever everyone wants it.13:17
warrenf13: any preference?13:18
f13warren: do you have a list of changes that would require?  Maybe an open infrastructure ticket?  noted which wiki pages need updating, etc..?13:19
f13I wouldn't be so keen on breaking cvs at the critical point in our release.  Much more comfortable doing it post-f913:19
warrenyeah, that might be a better idea13:21
f13I'm going to be flipping the fedora-release bit to make it look like F9, as well as turning off rawhide and on the "fedora" and "updates" repos.  We'll pull a mirrormanager trick again and just have those redirect to rawhide.13:21
f13(which should aide in testing out jigdo prior to release day)13:22
warrencool13:22
warrenperfect13:22
f13and the rest of the time is just losing sleep over blocker lists.13:22
f13So, open floor on F9 Final issues.13:22
nottingif we can't push it, we can't ship it.13:23
f13yep13:23
f13I'm probably going to harass people daily about this13:23
nottingfor F7 EOL, do we want to pick a date now, or wait until we're a little more clear on the panic/don't panic state of the blocker list?13:24
f13I'd say panic/don't panic.13:25
f13wait until we have a real RC13:25
warrenDo we consider the upstart fails to shutdown, upstart causing always/frequent unclean unmount issues, possibly related runlevel changes do nothing... blockers?13:25
nottingwarren: the hell?13:25
f13warren: have you been able to find anybody else who sees this?13:26
nottingplease elaborate on your first and second comments with bug numbers13:26
warrenreally nobody else is seeing this? wow.13:26
poelcatrelated to that F7 EOL and the recent ______ on f-dev-list about those darn BugZappers13:26
poelcatare there any compelling reasons to keep EOL bugs open?13:27
poelcatsomeone shut me down if this is too much of a tangent for this meeting :)13:27
f13poelcat: well, the answer to that question is always "yes", but the real question is 'compelling enough to put up with the cruft' and that I think is no.13:27
nottingpoelcat: only on an individual reporter/maintainer's opinion on specific bugs13:27
warrenhttps://bugzilla.redhat.com/show_bug.cgi?id=438444 switching to runlevel 1 does nothing https://bugzilla.redhat.com/show_bug.cgi?id=439699 you see nothing during shutdown (and sometimes it fails to shutdown)  https://bugzilla.redhat.com/show_bug.cgi?id=438449 killall race (might possibly related)13:27
buggbotBug 438444: high, low, ---, Casey Dahlin, NEW , upstart - 'telinit 1' is not doing anything even close to advertised13:27
buggbotBug 439699: low, low, ---, Casey Dahlin, NEW , No shutdown msgs displayed13:27
buggbotBug 438449: low, low, ---, Casey Dahlin, NEW , /etc/rc.d/init.d/killall is racing with other "stops"13:27
warrenI've personally NEVER seen my filesystem not replay journal since we've switched to upstart13:27
nottingpoelcat: does the bugzappers/triage stuff have an ability to only prod bugs once?13:27
nottingdid you file a bug?13:28
poelcatnotting: sortof... we've got that c00l whiteboard tag ;-)13:28
warrenwell, been relying on the existing bugs.13:28
nottingwarren: none of those bugs have anything to do with journal replay13:28
warrennotting: if it fails to cleanly shutdown it might?13:29
nottingwarren: presumably you'd notice that , considering you'd have to hit the power key, etc.13:29
warrennotting: it is inconsistent, but sometimes it powers off very quickly after I hit Shutdown13:30
poelcatnotting: so you're saying some way for a developmer to tag a bug as "don't auto close me" ?13:30
warrensometimes I do have to hold down the power button because X died, I haven no login getty's, and it is just stuck13:30
nottingpoelcat: i mean, once the triage system flags needinfo/whatever, i think it shouldn't try and do it *again* the next time it's chugging through rawhide13:30
jds2001well rawhide will be re-based to F913:31
poelcatnotting: well it shouldn't because the next action would be to ignore it or CLOSE if still needinfo13:31
nottingpoelcat: and if it's moved to 'devel' again...?13:32
warrenin rare occasions I do actually see stuff shutting down13:32
poelcatnotting: then it stays in 'rawhide' until the next GA when it gets rebased to the GA version13:32
jds2001it'll get put in the frying pan for the next cycle i think.13:32
jds2001poelcat: +113:32
nottingpoelcat: right. that's what i mean about only frobbing the bug once13:32
jds2001hmm, is there a compelling reason not to every 6 months?13:33
nottingnot sure. i suppose we can see how it plays out13:33
nottingi suppose my bigger triage concern is 'only one message per bug plz' :)13:34
f13seems like we're drifting here.13:34
jds2001yeah, sorry bout that :(13:34
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Open Floor13:34
nottingf13: apologies.13:34
nottinglmacken: bodhi need any prep for f9?13:36
lmackennotting: I don't believe so13:37
lmackenwhen do we want to start queueing up updates ?13:37
f13our sync scripts probably need updates13:37
lmackenI can create the F9 release in bodhi at any time, and keep it locked13:37
f13lets wait until we've branched13:38
poelcatf13: similar question for bug triage... when should we rebase rawhide-->f9 ?13:38
poelcator should I ask this somewhere else?13:39
f13poelcat: probably elsewhere, but I'd say after we start shipping F10 content in rawhide.13:40
f13which would be after F9 goes gold.13:40
poelcatf13: okay thanks13:42
f13if there's nothing else, I'd like to wrap up and get back to work.13:43
warrennod13:44
* warren looking into upstart issues...13:44
f13thanks all!13:45

Generated by irclog2html.py 2.3 by Marius Gedminas - find it at mg.pov.lt!