ReleaseEngineering/Meetings/2008-nov-03

From FedoraProject

< ReleaseEngineering | Meetings
Revision as of 05:23, 4 November 2008 by Poelstra (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Contents

Fedora Release Engineering Meeting :: Monday 2008-11-03

Preview Release

  • on track for release on tomorrow (Tuesday)
  • schedule CVS outage for Thursday evening in order to mass-branch for F-10
  • have some really bad split media splitting going on
    • needing more than one disc for a minimal install
    • needing 5 disks for a default minus office install, etc.
    • f13 to look at where packages are going and why they aren't being split right

Fedora 11 Schedule

  • https://fedorahosted.org/rel-eng/ticket/843 (discussion to date)
    • Includes three proposals
  • attendees would like to see a different proposed scenario to compare to a regular "may day" schedule
    1. May Day schedule should GA on 2009-05-05 (not 2009-04-28)
    2. second scenario should be F10 GA date + six month schedule (2009-05-05 + ~30 days)

IRC Transcript

f13 ping: notting jeremy spot rdieter lmacken wwoods poelcat warren 10:01
jeremy hi 10:01
rdieter here 10:01
* poelcat here 10:02
* notting is here 10:02
* wwoods here mostly 10:03
* warren here 10:03
f13 sorry, got slightly distracted by a shiney 10:09
-!- f13 changed the topic of #fedora-meeting to: Fedora Releng - Preview 10:09
f13 Preview is being mirrored currently, and should be in good shape. Lets run through the remaining tickets. 10:09
f13 .rel 865 10:11
zodbot f13: #865 (Compose Fedora 10 Preview "Spins" Live images) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/865 10:11
f13 This I'm going to be working on today. They won't likely be done and synced by release time tomorrow, there are just too many of them, so they'll go up at some point after that. 10:11
f13 the bandwidth between PHX and the torrent server doesn't help here either. 10:11
f13 .rel 866 10:11
zodbot f13: #866 (Give legal heads up on Preview release) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/866 10:11
f13 This I'll also be doing today, it's quick and easy. 10:12
f13 .rel 867 10:12
zodbot f13: #867 (Stage Fedora 10 Preview content for mirrors / torrents) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/867 10:12
f13 this is all done except for the spins to torrent part. I shoudl probably make that a different ticket next time. 10:12
f13 .868 10:12
f13 er 10:12
f13 .rel 868 10:13
zodbot f13: #868 (Create torrent configs for Fedora 10 Preview) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/868 10:13
f13 again, on my plate for today, should be relatively painless. All the torrent content (minus spins) is on alt. as well, for the people that help with seeding the torrents. 10:13
f13 .rel 869 10:13
zodbot f13: #869 (Release Fedora 10 Preview) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/869 10:13
f13 that's for tomorrow morning. I'll be up nice and early to release those bits. I assume we'll hang in #fedora-admin during the release hours to make sure things go well. 10:14
f13 .rel 870 10:14
zodbot f13: #870 (Branch for F11) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/870 10:14
wwoods woo! fresh, hot bits! 10:14
f13 this is the big ticket item for this week 10:14
f13 abadger1999: and I have been working together on it, to make the outage as short as possible. 10:15
f13 he added logic to pkgdb so that pkgdb can query koji and figure out what needs to be branched and then do all it needs to do 10:15
f13 we should only need to have an outage for the actual branching, and that outage is only so that we don't mail-bomb the world. 10:15
f13 abadger1999: can you confirm? 10:15
abadger1999 f13: We can do the mas branch in pkgdb without email, so no need for an outage during that portion. 10:16
f13 abadger1999: you can turn off the cvs commit mail? 10:16
abadger1999 f13: I'm not sure if we want to have an outage during the cvs branch creation. 10:16
f13 via pkgdb? 10:16
abadger1999 No, not the cvs commit mail. 10:16
f13 that's what we had outages for 10:16
abadger1999 So we would want to have an outage for that portion. 10:16
f13 we wanted to kill the commit mail, but not let commits hide during that time. 10:16
f13 nod 10:17
abadger1999 It's still a two step process but now the pkgdb half doesn't send email. 10:17
abadger1999 And should be faster. 10:17
f13 goot. 10:17
f13 abadger1999: what day did we decide on donig it? 10:17
abadger1999 We should do it towards the end of the week. No specific day yet. 10:17
abadger1999 there's a bunch of new branching code and I'd like nirik and other people processing cvs requests to pound on it for a few days before we hit it with mass branching. 10:18
f13 ok, I'm good with Thurs or Fri evening. 10:18
nirik abadger1999: is that code live now? 10:18
abadger1999 nirik: Not yet. I'm waiting for the infra freeze to end. 10:19
nirik abadger1999: ok. Let me know when it's in and I would be happy to test. 10:19
f13 Proposal: schedule CVS outage for Thursday evening in order to mass-branch for F-10 10:19
abadger1999 +1 10:20
notting +1 10:20
jeremy +1 10:21
f13 depending on how fast it goes, (and we shoudl time it), I think we should schedule this earlier in the cycle next time. 10:22
f13 right now, we're going to wind up with a lot of builds in 'dist-f10-updates-candidate' that aren't inherited into dist-f11 and confusing maintainers. 10:22
f13 of course, I dont' think we want to do a wholesale tag over of -candidate to dist-f11, but maybe that's the answer for this time. 10:23
f13 IMHO branching should occur at the same time we enter the final freeze where builds don't just "go" somewhere automatically 10:23
f13 alright, that's enough + votes, I'll announce the outage. 10:25
f13 since abadger1999 and I are west coast, we can do it relatively later in the evening, abadger1999 what time would you aim to start the CVS outage? 10:25
abadger1999 f13: Up to you. I can do it anytime on Thursday evening. 10:26
abadger1999 I need to time a complete run of the pkgdb side on app1.stg today as well. 10:27
abadger1999 cvs server we don't know how long... just that it's much faster since notting made the changes to what information we add to the modules file last release. 10:27
f13 yeah, it was incredibly fast once those changes were made 10:28
f13 I recall doing 30+ packages in like 3 seconds 10:28
f13 or something rediculous like that 10:28
f13 abadger1999: say 7pm Pacific? 0300 UTC Sat 10:30
abadger1999 Sounds good. 10:30
f13 ok. 10:31
abadger1999 0300UTC Sat would be Friday evening Pacific. 10:31
f13 right 10:31
abadger1999 Okay. 10:31
f13 er whoops 10:31
f13 sorry 10:31
f13 0300 UTC Fri for Thursday Pacific 10:31
abadger1999 Okay:-) 10:31
f13 alright, other things that I'm looking at. 10:32
f13 We've got some really bad split media splitting going on. 10:32
f13 needing more than one disc for a minimal install, needing 5 disks for a default minus office install, etc.. I need to look at where packages are going and why they aren't being split right. 10:32
f13 also 10:32
f13 Even after jeremy's patch for filtering available groups in anaconda, we're still getting af ew groups with just one or two packages in them. This is mostly due to packages being in multiple groups, or packages being pulled in via deps of another package in another group 10:33
f13 this leads to some confusing UI. I'm going to be reviewing the "available" groups in the install spin and trying to do some comps tweaks to make this better. 10:33
f13 those are my two main concerns this week 10:34
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Open Floor 10:34
f13 anybody got anything else? 10:34
poelcat f13: there was so feedback on the schedule in the ticket 10:35
poelcat schedule for F11 that is 10:35
f13 "so" == "no" ? 10:36
f13 .rel 843 10:36
zodbot f13: #843 (Draft Fedora 11 Schedule) - Fedora Release Engineering - Trac - https://fedorahosted.org/rel-eng/ticket/843 10:36
f13 I gave some feedback based on what I just experienced trying to get PR done 2 days after the final freeze 10:37
poelcat earlier in the ticket there were some objections to the drafted schedules 10:38
jeremy I very strongly object to rel-eng proposing a drastically different schedule than our standard for f11 10:39
jeremy if we have suggestions about how to lay out milestones within the constraints of the every six month mandate that we're under, that's reasonable 10:39
jeremy more than that, rel-eng feels very much like the wrong place to have the discussion 10:40
jwb mandate from whom 10:40
* notting isn't even sure that the schedule discussion started in rel-eng 10:41
poelcat it started on the releng list 10:41
jeremy jwb: was originally the board iirc 10:41
poelcat by stickster i think 10:41
jwb i'm pretty sure if rel-eng and FESCo went to the board and said "this needs to be changed", they would at least entertain it 10:42
jeremy jwb: I think that such a disucssion should happen in the open, not in some rel-eng trac ticket that no one watches 10:43
jeremy any prior "we need to have a different schedule" discussions have begun either on fedora-devel or fedora-advisory-board 10:43
f13 jeremy: what's wrong with releng itself using the trac ticket to come upw ith our suggestion, to propose to the community? 10:43
f13 jeremy: I'd rather not have a 20 day long bikeshed argument before we even get to a suggested alternative to the existing schedule. 10:43
jwb ok, sure. i was just pointing out that the 'mandate' can be changed 10:44
f13 jeremy: the alternative schedule being discussed in that ticket can /be/ the starting point for the "we need something different for F11" discussion on F-A-B/Fedora-devel-whatever 10:44
jeremy f13: fine. send it to f-a-b 10:44
jeremy f13: let's have the discussion there. 10:44
jeremy you call it bikeshedding 10:45
jeremy I call it being inclusive 10:45
f13 jeremy: we're going to, once we as rel-eng agree that the proposed alternative works for us 10:45
poelcat of the three schedules proposed in the ticket... is there one that folks would like me to tweak/adjust to get closer to the desired end product? 10:45
f13 starting a thread with "we need change" and not having anything to propose is rather.. pointless. 10:45
jeremy f13: if there is a suggestion coming from the set of people that make up rel-eng, then we stifle real discussion 10:46
f13 poelcat: personally I'm in favor of the http://poelstra.fedorapeople.org/schedules/f-11/f-11-all-tasks.html as being proposed as our suggested "change" to various lists for discussion. 10:46
f13 jeremy: so, suggestions are only useful if they come from individuals? 10:47
f13 discussing a suggestion beforehand isn't useful? 10:47
f13 I'm not allowed to ask a few of my peers to sanity check something before I post it to a wider audience? 10:47
jeremy f13: sanity checking is one thing 10:47
jeremy f13: coming to a conclusion and presenting it is entirely different. 10:48
jeremy because then the argument is "well, this has already been agreed on by X, Y and Z" 10:48
f13 "coming to a conclusion" 10:48
spot FWIW, I think I'm more inclined towards http://poelstra.fedorapeople.org/schedules/f-11/f-11-may-day.html 10:48
f13 yes, the releng team is hell bent on railroading this schedule into place. 10:49
f13 end of story. 10:49
stickster spot: And I'm thinking about something in between :-D 10:49
notting there was http://poelstra.fedorapeople.org/schedules/f-11/f-11-standard.html which was somewhere in between (although not halfway) 10:50
spot stickster: i think that we need at least 6 months 10:50
f13 er, may-day doesn't give us 6 months 10:50
spot notting: thats an 8 month schedule 10:50
f13 spot: the may-day schedule is an extremely short schedule for F11 10:50
f13 how about this though. 10:50
spot f13: yeah, its not quite perfect 10:51
f13 I don't necessarily see any glaring problems with all three options. 10:51
f13 I propose that all three be passed along to the relavent lists for discussion 10:51
spot i'd like to propose a fourth option 10:51
spot which is essentially the may day schedule out to May 31 10:51
* notting would like to apologize for his eyes glazing over from one too many schedule discussions and not posting much input 10:51
spot give it a proper 6 months 10:51
notting spot: we can't do that until we ship f10 on time ;) 10:52
spot but leave the feature freeze date as is in feb 10:52
jeremy (... so, why is it fair that this discussion only have input from the people that are here / pay attention to rel-eng as opposed to having it on f-a-b or fedora-devel-list? that other input is valuable and we should not be looking to avoid it. it's not "bikeshedding") 10:53
f13 jeremy: because what we're doing right now is discussing what the schedules would look like with a suggested adjustment to the F11 cycle. 10:54
f13 jeremy: what rel-eng would suggest the community review 10:54
notting for example, the may-day one has two weeks between alpha GA and beta freeze, which, well... 10:54
f13 the adjustment really isn't that hard to get to, adjust F11 to the same manner that F10 was delayed 10:54
f13 although, wait a sec. 10:55
notting poelcat: refresher - what's the genesis of the 'regular' schedule being 230 and not 180 days? 10:56
poelcat notting: link? 10:57
f13 poelcat: also, I thought "regular" and "regular with extra freezes" was supposed to be the same length in time, but with some dates picked out to freeze key software earlier 10:57
notting poelcat: http://poelstra.fedorapeople.org/schedules/f-11/f-11-combined-scenarios.html 10:57
poelcat notting: okay 10:57
* poelcat looks just to be sure 10:58
f13 yeah, regular with extra freezes seems to have a few extra days at the tail end of Beta that are somewhat unaccounted for 10:58
notting f13: and at the start of alpha. or i'm missing something. 10:59
f13 crap, I should have noticed this before, sorry poelcat 10:59
poelcat notting: i (wrongly/rightly) pushed things forward ~1.5 weeks to compensate for holidays 10:59
poelcat f13: no worries... it is all changeable :) 10:59
notting poelcat: that's still doesn't account for 50 days 10:59
poelcat + 1 month for f10 slip 11:00
* jeremy thinks there are multiple things trying to be worked at the same time and thus make the discussion hopelessly complex. 1) end date) 2) changes to how some of the internal freezes work out 11:00
* poelcat is trying schedule w/ "work days" not "calendar days"... a separate debate :) 11:01
notting poelcat: ??? that month is taken care of by the later start date 11:01
jeremy poelcat: every day is a work day in the community :-) 11:01
notting honestly, i think you chop a month off the pre-alpha in either of the later two schedules, and you have something eminently reasonable 11:03
f13 so lets take a step back. 11:03
f13 for the sake of community discussion, I think we can make things simple by talking about 2 scenarios. 11:04
f13 1) Shoot for May Day release, disregarding the delay in F10's schedule. 11:04
f13 2) Pad F11's schedule by the month we delayed F10's schedule. 11:04
f13 Those two options should be pretty simple for people to grasp. 11:05
poelcat we have #1 already, correct? 11:05
notting f13: just to have clarity, 'shift', not 'pad' 11:05
f13 in addition to that, as a side discussion, regardless of end date I want to explore adjusting the freeze dates slightly, move feature freeze a week back from beta freeze, and ask for a freeze of critical software a bit before that even. 11:05
f13 notting: ah right, 'shift' 11:06
f13 poelcat: I think we'll confuse people too much by throwing full schedules at them, and wind up derailing those conversations like this one is on the details of the individual inter-schedule dates.  :/ 11:06
poelcat so scenarios #2 you want a GA of ~May 1 + 30 days ? 11:06
notting poelcat: ignore GA, start by taking a 'normal' six month schedule, and just have it start at the start point? 11:07
notting GA-as-output, not input 11:07
poelcat notting: people seem to disagree on what a "normal" 6 month schedule is 11:07
notting poelcat: take what the mayday schedule used, and add back whatever was taken out ;) 11:08
* poelcat definitely isn't clear what that is... is May/Halloween the "normal 6 month" 11:08
poelcat  ? 11:08
f13 yeah, I"d say shoot for the nearest Tuesday to a May 1 + 30 days. 11:08
* f13 hates that we're piling more work on poelcat after these schedules sat in a ticket for weeks :/ 11:09
poelcat that's okay :) 11:10
poelcat it is an itterative process 11:10
poelcat so here is what I'm going to do... 11:11
poelcat 1) scenario #1 is considered "okay enough for now" ... agreed? 11:11
notting as a proposal? sure! as something sane to do? hell no! 11:11
poelcat right 11:12
f13 poelcat: just for snorts and giggles, move that final day to May 4th (: 11:12
f13 er may 5th 11:12
poelcat okay 11:12
f13 (you know, just so that you have more work to do) 11:12
f13 sooner or later poelcat is going to take the train up here and beat me. 11:12
poelcat 2) create a "better" scenario #2 which mirrors f10 and ends May 5 + ~30 days 11:13
notting (for it to be a valid 'something sane to do', realistically you have to move the alpha back so you have more than 2 weeks of alpha) 11:13
poelcat where "mirrors" is defined loosely in the sense that changes can be made for "lessons learned" in F10 11:13
poelcat notting: yes, most likely the cuts are going to come from the length of devel time 11:14
poelcat before alpha 11:14
notting erm, the main lesson learned during f10 is not applicable to scheduling. *sigh* 11:14
poelcat though i purposefully made it the way it was based on what we usually hit for length of devel time before alpha 11:14
f13 right, for the sake of discussions on lists concentrating on the duration of the schedule in whole is important, not the individual dates within teh schedule. 11:14
poelcat f13: okay so i remove several layers of detail 11:15
poelcat so mostly what we end up w/ is ~ 10 milestones? 11:15
f13 even that can be misleading 11:15
poelcat lol 11:15
* poelcat runs out of ideas :) 11:16
notting so 11:16
notting you have a 6 month schedule 11:16
f13 all I really want to hear from the community at large is whether or not they'd accept a shifting of the F11 schedule or not. 11:16
notting you have a generally accepted length of time for a milestone that is counted from *availability of that milestone, to availability of the next milestone* 11:16
f13 a shifting of the f11 and using F12 to try and bring it back to mayday/haloween 11:16
f13 compatabilty with RHEL 6 is my primary motivation to a full shift of the F11 schedule, capturing as much of the RH attention as possible. 11:17
notting freezes are added to that schedule based on whatever those avail dates are. and when those freezes are are the bits that f13 wants to adjust, iirc. those adjustments should not affect the length of the milestones, necessarily 11:17
notting (unless you get into a 'we're freezing for milestone A+1 two weeks after milestone A') 11:17
notting am i over-simplifying or over-complicating? 11:17
poelcat f13: what do you mean by "capture attention of RH" ? 11:18
f13 poelcat: well, when Red Hat is gearing up to base a RHEL release on a Fedora release, more facilities within Red Hat pay attention to Fedora, and try to improve things within Fedora. Because it's easier than doing them within Red Hat for RHEL 11:18
f13 poelcat: maximizing their time to generally improve Fedora within Fedora seems like a worthwhile goal. 11:19
f13 they had expected a 6~ month schedule for F11 to deal with, and I don't want to short-change that. 11:19
f13 however, if the greater community does not agree, then we're going to go with what the greater community wants. 11:20
jwb be careful 11:20
wwoods that has to be phrased carefully or we're gonna hear a lot of complaints. well. a lot more than normal. 11:20
jwb heh 11:21
f13 well sure, careful but honest 11:21
jwb particularly when "greater community" seems to equate to "the people that bark the loudest and most" 11:21
f13 jwb: by 'greater community' I think I mean a FESCo vote after the requsite arguing on mailing lists 11:21
wwoods It totally makes sense to summarize as: well, we slipped a month, so *something* has to be a month shorter. And since large chunks of RHEL engineering are going to be getting involved in F11 improvements 11:21
f13 and maybe a board nod for doing a schedule adjustment. 11:22
jwb f13, then maybe you should just say that 11:22
f13 jwb: agreed 11:22
wwoods it makes sense to shorten the F12 cycle and make *that* release more focused on stabilization/polish than new features 11:22
f13 wwoods: I like where you're going with this 11:22
wwoods (besides, I think we have the beginnings of a star-trek-movie tradition: only the even numbered ones are good) 11:22
f13 anyway, I can take up the task of doing the proposals to lists 11:23
f13 and taking any heat that blows back from it. 11:23
f13 we can re-visit the finer points of our schedules once we have a generally agreed upon end date 11:23
poelcat f13 would we want to increase the "six month" schedule by two more weeks to accomodate the "extra freezes" 11:24
poelcat or do we 11:24
poelcat want to somehow make those work w/in the "6 month" schedule? 11:24
f13 poelcat: no, the "extra freezes" were just dates set within the existing schedule time. 11:24
f13 not extra time added. 11:25
poelcat where "we" == "me" w/ the scheduling tool :) 11:25
poelcat okay, got it 11:25
poelcat so most likely we'll have to take that time from time before alpha... is that okay? 11:25
poelcat if this too detailed for now, close me down :) 11:25
f13 Proposal: jkeating will post to $somewhere asking for a discussion on the F11 schedule, providing two simple scenarios. Shoot for May 1 (May 5th), or shoot for May 1 + the month we lost in F10. 11:26
f13 poelcat: I'm not exactly sure why we need to take "time" away when adding the extra freeze points. 11:26
f13 poelcat: we're A) asking for features to be "done" earlier, but that doesn't stop them from doing more bugfixing. "Done" just means testable. 11:27
f13 and B) we're asking some key software to be "done" (testable) even earlier to provide room for integration work just prior to the beta freeze. 11:27
poelcat fair enough 11:27
* poelcat didn't mean to drag in underlying schedule semantics 11:28
f13 if anybody is still paying attention to IRC, some voting would be appreciated, or else I'll just dictate (: 11:28
poelcat +1 to proposal but change it to "May 4" + 1 month 11:29
f13 sure 11:29
poelcat because normally we would have targetted 2008-04-28 11:29
poelcat 2009 that is 11:29
wwoods well, I'll give it a +1, but I'd prefer we put forth the "keep f11 6-months, shorten f12" proposal as the first choice 11:29
wwoods but that's just my preference in the matter 11:30
f13 wwoods: sure, I think releng as a whole leans more to that. 11:30
f13 we'd rather see a 6 month schedule, and I'll provide rational as we discussed here as to why 11:30
notting isn't may 4 + one month roughly 6 months? 11:33
wwoods yep. I'm just saying the choices should be: (may 4 + 1mo), may 4. Quibbling over ordering. 11:33
f13 notting: it is. I had commented earlier that if we didn't shift the f11 schedule, I'd like the release point to be may 4, rather than april 28th 11:33
wwoods those sorts of things have significant polling effects. stupid election makes me think too much about this stuff. 11:34
f13 I think our "standard" schedules should be "on or after May1/Oct31" 11:34
f13 alright, no further votes coming in, I'll run with it. 11:37
f13 now that we're 40 minutes over, anything else? 11:37
wwoods oh 11:38
wwoods there's no koji tag for preview, right? 11:38
wwoods is there a tree somewhere so I can check to see what packages are in the preview (and thus close bugs etc.) 11:38
f13 wwoods: nope. 11:39
f13 wwoods: you could check the kojipkgs.fedoraproject.org/mash/rawhide-20081031 tree 11:39
f13 that's what I used to compose the PR (minus the fedora-release package) 11:39
f13 alright, ending the meeting, thanks all. 11:40

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!