Infrastructure/Meetings/2007-02-15

Meeting of 2007-02-15

*** Time shown in EST

15:08 < mmcgrath> Lets get the meeting started
15:08 < dgilmore> mmcgrath: ill run with it
15:08 < warren> I wont be able to make 7PM EST, but I'm probably not useful right now in this topic anyway.
15:08 < dgilmore> ill just tell you when i need tb's of disk space
15:09 < mmcgrath> Role Call - Who's here
15:09 < abadger1999> yo
15:09 < jcollie> here
15:10  * iWolf is here
15:10 < warren> here
15:10 < mikem23> aye!
15:10 < mmcgrath> Alrighty ---------- Package Database (abadger1999)
15:10 < mmcgrath> Whats the word?
15:11 < abadger1999> How do we want to tie in with koji :-)
15:11  * daMaestro is here
15:11  * f13 is watching
15:11  * dgilmore is here
15:11 -!- mdomsch [n=Matt_Dom@cpe-72-177-104-34.austin.res.rr.com]  has joined #fedora-admin
15:11 < mmcgrath> abadger1999: do you have an idea of how that will go?  I still haven't really seen koji
15:11 < abadger1999> I posted http://www.fedoraproject.org/wiki/Infrastructure/PackageDatabase/KojiIntegration
15:11 < mdomsch> sorry I'm late
15:12 < abadger1999> Initial look at how koji and he PackageDB overlap.
15:12 < warren> schema is one thing, but folks really need to think about how we want the workflow to look and enable
15:12 < mmcgrath> mdomsch: no problem
15:12 < warren> the schema must be able to support what we desire
15:12 < warren> schema without workflow is in danger of being only theoretical?
15:13 < abadger1999> warren: More likely, schema without workflow == different ideas in different people's heads.
15:13 < warren> PackageDB isn't a blocker for F7.  Maybe we should just do minimal work on  koji to get it running, and having something concrete to use, we'll have a better idea of how to integrate it to create PackageDB?
15:13 < abadger1999> The packagedb schema will eable the workflow I'm envisioning.
15:14 < abadger1999> But it doesn't enable all that koji can do.
15:14 < abadger1999> And vice versa
15:14 < abadger1999> I'd really like to see packagedb for f7
15:14 < warren> Hm...
15:14 < dgilmore> warren: thats the plan
15:14 < abadger1999> As the whole owners.list mess is going to drive packagers away.
15:14 < warren> Yes, owners.list and lack of automation in other stuff is a mess.
15:14 < mmcgrath> abadg1er1999: +1
15:15 < abadger1999> I would like to have packagedb and koji separate dbs and processes for F7.
15:15 -!- neuro_ [n=neuro@neuro.me.uk]  has quit [Remote closed the connection] 
15:15 < warren> If you're serious about this, this sounds like something I should be focusing on myself.
15:15 < warren> I was thinking that PackageDB wouldn't be feasible for F7.
15:15 < f13> aside from initialCC, is koji db missing anything for replacing owners.list?
15:15 < abadger1999> The packagers would primarily work through the packagedb to create and manage acls, ownership, etc.
15:15 < f13> oh I guess package description and all that.
15:15 < abadger1999> f13: Yes.
15:15 < warren> We want to be careful not to block new buildsystem on packageDB, if this will take too long.
15:16 < abadger1999> It also misses FAS tie in and such.
15:16 < warren> abadger1999, another issue...
15:16 < f13> abadger1999: the script we use internally actually looks at the rpm to get the description and such for the bugzilla entry
15:16 < abadger1999> warren: The big question is really, how do we want the two to interact.
15:16 < f13> abadger1999: the FAS tie?  Wouldnt that be the "owner" in koji?
15:16 < warren> isn't the fedora-cvs procedure relatively straight forward and usable for the short-term?  the hard confusing and upsetting part has been reviews, which would be even more difficult to fix with packagedb.
15:17 < warren> hm
15:18 < dgilmore> warren: we need minimal talking betwen buildsys and packagedb in phase 1
15:18 < warren> Perhaps an organic growth model would be good here.
15:18 < warren> start with koji with minimal modifications to get it running, using our existing owners.list and FAS
15:18 < mmcgrath> abadger1999: Think you can do what you need to by F7?
15:18 -!- lutter [i=dlutter@nat/redhat/x-da84fb76bd0b3cc4]  has quit [Client Quit] 
15:18 < mmcgrath> or does that depend on koji?
15:18 < warren> Having that, will have something tangible to spark imaginations for PackageDB, which can grow alongside it, depending on koji at first.
15:19 < abadger1999> It depends on what workflow we want -- exactly as warren said.
15:19 < dgilmore> warren: thatshow i saw it  with ties into the packagedb as it came up
15:19 < abadger1999> Maybe we should keep them completely separate and packagers will have to use two separate interfaces -- packagedb for owners.list stuff and koji for building.
15:19 < dgilmore> with the packagedb and buildsysdb becoming one over time
15:19 < warren> I suspect that anything that grows to depend on koji will have redundant and poorly normalized (confusing and bad design) databases.  We can take what we learned from this Ph
ase 1, and reimplement it together for Phase 2 after Fedora 7.
15:20 < warren> abadger1999, separate but with same authentication shouldn't be too bad.
15:20 < abadger1999> The area where I see the packagedb starting to get into koji's space is that we can define acls on who can build a package.
15:20 < dgilmore> warren: all should authenticate from fas
15:20 < abadger1999> (Duplicate information on what collection a package belongs to as well.  But that's not as much of a problem in the short term.)
15:21 < warren> Do we like the phased approach?
15:21 < warren> Phase 1: Get koji up, use FAS
15:21 < warren> Grow packagedb to depend on both
15:21 < dgilmore> warren: yes
15:21 < warren> Perhaps let packagedb takeover owners.list if it becomes ready.
15:21 < warren> Phase 2: Maybe after Fedora 7, make it less of a spaghetti mess.
15:22 < mmcgrath> abadger1999: is that something like what you had in mind for this?
15:22 < abadger1999> Pretty much.  The only question I have is should I enable build acls in the packageDB.
15:22 < abadger1999> It's something we want.
15:23 < abadger1999> So, for instance, we can have "limited comaintainers" who can commit but not build.
15:23 < f13> that starts to get quite complicated :/
15:23 < abadger1999> But that requires the packagedb to talk with koji.
15:23 < f13> I'd rather keep things simple
15:23 < abadger1999> It can be simple if the packagedb invokes koji to build.
15:23 < abadger1999> Rather than packagers going straight to koji.
15:23 < mmcgrath> f13: where do you see the build acl's living?
15:24 < f13> mmcgrath: well, somewhere that koji can read from (:
15:24 < mmcgrath> I think thats what toshio is saying, that they'd both be pulling from the same datasource.
15:25 < f13> all ACLs should live in the same place and be adjustable by the same interface
15:25 < mmcgrath> unless I'm confused?
15:25 < f13> I really think buildACLs is an After F7 feature
15:25 < abadger1999> How about we make buildACLs the first priority of Phase 2?  Then we can finish up the easy points of the PackageDB and getting koji running.
15:26 < abadger1999> Where they're separate.
15:26 < warren> We really don't need buildACL's, and IMHO cvs ACL's is a bit paranoid.  In practice I suspect the vast majority of packages to just remove their ACL.
15:26 < mmcgrath> abadger1999: sounds good to me.
15:26 < abadger1999> And if we have time, we can work on it for F7 if not, it'll come in when we integrate everything for F8.
15:26 < warren> what is more important is push ACL, which is enforced at the update system interface.
15:26 < warren> which will probably need to tie into packagedb eventually...
15:26 < mmcgrath> You can always get a prototype together for F7 and let us decide if it should wait.
15:26 < warren> has anyone looked into the schema of lmacken's update system?
15:27 < mmcgrath> I haven't.
15:27 < abadger1999> Not yet.
15:27 < mmcgrath> abadger1999: lets make that the plan then, leave out the buildACL's for now but focus on the rest.
15:27 < warren> abadger1999, if you happen to make something that WORKS rapidly and convince everyone of it, then we may just jump onboard.  but I think people are more concerned now about getting the minimal stuff online in order to get the job done, and avoid blocking on something that will take too long.
15:28 < abadger1999> warren: Right.
15:28 < abadger1999> That sounds like a plan.
15:28 < warren> the phased approach (evolving over time) might work
15:28 < mmcgrath> Sounds good, abadger1999: anything else or should we move on?
15:28 < warren> get the minimal online, then evolve it to suck less with packagedb, show that it works.
15:29 < warren> Phase 1: minimal koji with FAS
15:29 < warren> Phase 2: packagedb improvments to make it suck less
15:29 < abadger1999> BTW: I havne't published this URL yet, but I have a development snap running on https://admin.fedoraproject.org/pkgdb-dev
15:29 < warren> Phase 3: unspaghettification
15:29 < abadger1999> Which I reserve the right to break while admin.fp.o/pkgdb is supposed to work for people to show others.
15:30 < mmcgrath> sounds good, moving on.
15:30 < mmcgrath> VCS choice.
15:30 < f13> choice for what?
15:31 < mmcgrath> I vote that we should move this down to priority 2 for a while.  The fact is CVS is here, it works and people aren't exactly beating down our doors to change it.
15:31 < f13> yeah, bench ituntil we get reports from SCM Sig
15:31 < mmcgrath> plus we have enough other stuff going on at the moment.
15:32 < abadger1999> +10
15:32 < mmcgrath> heh, ok.  moving it down
15:32 < mmcgrath> Upgrading systems ----------------------- (iWolf) ?
15:32 < iWolf> here!
15:33 < mmcgrath> still waiting on the drives on the old CVS box?
15:33 < mmcgrath> mgalgoci: ping
15:33 < iWolf> I have completely sucked of late.  Between day job and baby on the way my time has been scarce.
15:33 < iWolf> Hence, why I've been so quiet of late.
15:33 < iWolf> So no real progress to report.  :(
15:33 < mmcgrath> iWolf: no worries.  It happens.  If you'd like We can always move that task to someone else till you get some more free time?
15:34 < iWolf> If someone has time to get to it, my feelings won't be hurt.
15:34 < mmcgrath> <nod>
15:34 < mmcgrath> luke's not here so thats out
15:34 < mmcgrath> wiki upgrade ------------------------------- (mmcgrath)
15:34 < mmcgrath> so paulobanon and kim0 have both done a wiki upgrade and have informed me that it will be pretty painless.
15:34 < mmcgrath> I've been testing on my own and I think it will be.
15:35 < mmcgrath> here's the idea, make the wiki read-only.
15:35 < mmcgrath> copy the data over to the app server that will take it over.
15:35 < mmcgrath> test
15:35 -!- lutte1 [i=dlutter@nat/redhat/x-25b99838c4522506]  has quit [Read error: 110 (Connection timed out)] 
15:35 < mmcgrath> change DNS.
15:35 < mmcgrath> wiki upgrade done.
15:35 < mmcgrath> There will be odds and ends we need to fix but I'm pretty sure they are small.
15:36 < mmcgrath> I'm trying to get us some common storage on the back end so we can cluster the wiki but I've not been given any timeline as to when that may happen or even that we have the resources for it now.
15:36 < mmcgrath> the old fpserv will become a live backup of the cvs server as well as being our primary DNS server.
15:36 < mmcgrath> Config management ----------------------------- (mmcgrath)
15:37 < mmcgrath> I've been testing puppet, I'm trying to make sure it is and stays sane.  I'll be sending things to the list to let everyone know when proxy1/2 are converted.
15:37 < mmcgrath> Metrics --------------------------------------- (Metrics)
15:37 < mmcgrath> Metrics --------------------------------------- (mmcgrath)
15:37 < mmcgrath> Smolt is up and ready, I've been adding some queries, breaking some things, this and that.
15:37 < mmcgrath> f13: ping
15:37 < f13> yes?
15:38 < mmcgrath> So do you think smolt will make it into the default install profile?
15:38 < mmcgrath> I heard a lot of +1's and one "aldjflsjf"
15:38 < mmcgrath> from the smolt: revisited thread.
15:38 < f13> mmcgrath: yes.
15:38 < f13> mmcgrath: just mark it 'default' in extras comps
15:38 < f13> I'll try to remember to add it to the manifest
15:39 < mmcgrath> f13: k, so I can just leave that up to you?
15:40 < f13> yes, once you've changed it in comps.
15:41 < mmcgrath> where is the master comps file at?
15:42 < mdomsch> mmcgrath, isn't it in extras cvs in the comps module?
15:42 < mmcgrath> No idea, I've never messed with it.
15:43 < mmcgrath> I'll look into that more
15:43 < mdomsch> comps-fe7.xml
15:43 < mdomsch> .in
15:43 < mmcgrath> msomsch
15:43 < mmcgrath> mdomsch: thanks
15:43 < mmcgrath> Postfix ------------------------------------ (dgilmore)
15:44 < mmcgrath> actually dgilmore is waiting on me for this, I hope to have something about this soon.
15:44 < dgilmore> im in a meeting sorry
15:44 < mmcgrath> no worries
15:44 < mmcgrath> FI NOC --------------------------------------- (mmcgrath)
15:44 < mmcgrath> still on hold but its coming
15:45 < mmcgrath> Hosting --------------------------- (f13)
15:45 < mmcgrath> f13: whats the word?
15:45 < f13> unng
15:45 < f13> well, git:// works.
15:45 < f13> no thanks to me, it was already working
15:46 < f13> I should write up my future plan, but -ETIME
15:46 < f13> so, business as usual
15:46 < mmcgrath> yep, I hear you
15:46 < mmcgrath> Mirror Management -------------------------------------- (mdomsch)
15:46 < mmcgrath> How's TurboGears treating you Matt?
15:46 < mdomsch> like a glove now
15:47 < mdomsch> test instance up for looks
15:47 < mdomsch> gotten some "looks good' comments
15:47 < mdomsch> still have some code to write to export the data once it's all in, but that's coming along
15:47 < mmcgrath> Still looking good to hit the F7 target?
15:47 < mdomsch> mgalgoci, I need publictest7.f.r.c on the rsync ACL list, I sent a note to ftp@ a few days ago
15:48 < mdomsch> mmcgrath, my hope is to let folks kick the tires with Test2
15:48 < mdomsch> and Test3 pretend like we're live
15:48 < mdomsch> so we can be live for F7
15:48 < mdomsch> I think we're on track for that
15:48 < mmcgrath> excellent
15:49 < mmcgrath> Yum DeltaRPM -------------------------------------------- (kim0)
15:49 < mmcgrath> Whats up kim0?
15:50  * mmcgrath thinks kim0 might be AFK.
15:50 < mmcgrath> Open Floor ---------------------------------------------- (anyone)
15:50 < mmcgrath> Anyone have anything else to discuss?
15:50 -!- Harley-D [i=Harley@fedora/Harley-D]  has joined #fedora-admin
15:51 < abadger1999> TurboGears look and feel?
15:51 < mmcgrath> abadger1999: ahh, good question.
15:51 < abadger1999> We have at least three different looks between pkgdb, mirrormanager, and smolt.
15:52 < mmcgrath> Yep, let me ping Mo about that and see if she has something in mind.
15:52 < mmcgrath> abadger1999: what do you think?  Did you have something in mind?
15:52  * mmcgrath is a UI idiot.
15:53 < abadger1999> No.  I'm UI blah as well.
15:53 < mmcgrath> K, I'll wait to see what she says then.
15:53 < mmcgrath> Anyone have anything else?
15:54 < mmcgrath> Alllrighty ------------------------------------------ MEETING END