From FedoraProject

Jump to: navigation, search

Infrastructure > Meetings -> 2006-12-14

Meeting of 2006-12-14

*** Time shown in EST

15:05 < mmcgrath> Yoooo
15:06 < mmcgrath> Who's ready for a meeting?
15:06 < jcollie> i got my caffeine standing by!!!
15:06  * dgilmore is here
15:06 < mmcgrath> Role call!
15:06 < teknofile> Aye
15:07 < warren> here, but busy...
15:07 < kschreyack> here, but lurking...
15:07 < mmcgrath> Here we go -
15:07 < lyz> here, but at work
15:08 < mmcgrath> abadger1999: ping?
15:08 < mmcgrath> (re packaging database)
15:08 < abadger1999> I was travelling this week so didn't get much done.
15:09 < abadger1999> I have an updated schema written down on paper but I have to enter it and test it.
15:09 < mmcgrath> Cool.
15:09 < abadger1999> I'll post it to the lit over the weekend.
15:09 < mmcgrath> f13, abadger1999: VCS?  anything else?
15:09 < abadger1999> Anyone know if FAB discussed the fedora-scm sig?
15:10 < mmcgrath> I haven't seen anything beyond "we need one"
15:10 < jcollie> yeah, it'd be nice to get that started
15:10 < mmcgrath> I'll 'restart that' right now.
15:10 < abadger1999> k
15:11 < abadger1999> I think for the VCS to progress we need to get the SIG up and running and start defining what we really want
15:11 < mmcgrath> iWolf: around?
15:11 < jcollie> yeah, it'd be nice to have semi-functional demonstration systems too, but that's an issue for the SIG
15:12 < abadger1999> Exploded sources might be better for working with upstream but might be harder for our contributors to use.
15:12 < mmcgrath> We have pretty funcational systems for mercurial and svn
15:12 < abadger1999> We need the SIG to explore those ideas.
15:12 < jcollie> yeah, if there's going to be a SIG no need to discuss much more here
15:13 < abadger1999> mmcgrath: Right.  but they're both based on the idea that we just want to map cvs -> a better SCM.
15:13 < mmcgrath> I'll send the email out after the meeting to devel, extras and FAB to get a list of interested parties together.
15:13 < mmcgrath> lmacken: ping?
15:13 < abadger1999> Cool.  I think that's what we're waiting to here about.
15:14 < dgilmore> mmcgrath: i think he has a class or something
15:14 < skvidal> hi guys
15:14 < skvidal> sorry for being late
15:14 < dgilmore> hey skvidal
15:14 < mmcgrath> doah, thats right.  he's gone for the next few weeks.
15:14 < mmcgrath> hey seth
15:14 < skvidal> when there's a place in the agenda I want to talk about mirrors
15:15 < mmcgrath> We just got done discussing the priority 1 stuff so have at it.
15:16 < mmcgrath> skvidal: what did you have in mind for mirrors?
15:17 < skvidal> well the subject came up of the infrastructure folks taking over mirror mgmt entirely for fedora
15:17 < skvidal> so it no longer falls on IS to take care of the rsync parts, etc
15:17 < mmcgrath> You're talking about the primary mirror?
15:17 < skvidal> so that fedora-infrastructure is the link to the mirror-admins
15:17 < dgilmore> skvidal: we dont have access to do that
15:17 < skvidal> you're right
15:17 < skvidal> we don't
15:17 < skvidal> I'm talking about more of the social interaction
15:17 < skvidal> and notification
15:17 < dgilmore> ok
15:17 < skvidal> eventually leading to a point where the mirror tool that we're working on
15:18 < mmcgrath> Thats certainly a task we can take on.
15:18 < skvidal> allows mirror-admins to register, etc and that propagates the rsync lists
15:18 < f13> mmcgrath: sorry, I'm in a meeting :/
15:18 < dgilmore> skvidal: mdomsch was asking about that bit earlier
15:18 < mmcgrath> Yeah, someone is actually working on that as we speak.
15:18 < skvidal> yes, we were both in the same call on the board
15:18 < mmcgrath> though he's been quiet on the lists :-/
15:19 < skvidal> right, so what I was thinking if we do that it would be useful to start really making mirror tiering work
15:19 < dgilmore> skvidal: sure
15:20 < dgilmore> would you have the mirrors tier themselves with some guidelines?  or how do you propose it works?
15:20 < skvidal> okay. I just wanted to make sure this came up b/c it would be great to remove the pain from IS of dealing with mirrors and hopefully make our mirror admins happier st the same time
15:20 < mmcgrath> skvidal: would you like to put together some requirements?
15:20 < skvidal> sure, I can come up with something
15:20 < mmcgrath> rockin
15:20 < skvidal> I'll email the list
15:21 < mmcgrath> thanks.
15:21 < jcollie> personally, i think it would be nifty to have some squid caching accelerators scattered about the net... i killed my private rsync mirror because it was just too much trouble to maintain
15:21 < mmcgrath> Ok, so next item is the config management stuff.
15:21 < mmcgrath> dgilmore: have you had a chance to look at cfengine?
15:21 < iWolf> I'm here now, spaced the starting time of the meeting
15:22 < mmcgrath> iWolf: no problem, any update on the db box?
15:22 < dgilmore> mmcgrath: not yet
15:22 < iWolf> Sounds like the old cvs-int box is ready to be rebuilt (still needs firmware updates on the hd's).
15:22 < iWolf> I need KVM access to it, and not sure what tool you guys are using for that.
15:22 < mmcgrath> thats right, I keep forgetting about the firmware on those boxes.
15:23 < mmcgrath> does the console not work?
15:23 < iWolf> sounds like I can get started on it, just hold off on pushing it to production till they get the firmware updated.
15:23 < mmcgrath> K, that'll probably involve coordination with Stacy.
15:24 < iWolf> I had trouble with the console the last time I tried it (awhile back).  I will try again tonight and ping you if I still have troubles.
15:24 < mmcgrath> Sounds good.
15:24 < mmcgrath> warren: any word on the smtp server?
15:24 < warren> mmcgrath, ack
15:24 < warren> ..
15:24 < warren> I suck
15:24 < warren> well, sqlgrey is almost done with review now
15:24 < mmcgrath> I suck is "open source" for "I'm busy" :-D
15:24 < warren> any objection to using mysql with sqlgrey?
15:24 < warren> or want to stick to sqlite?
15:25 < warren> sqlite might be easier to back up
15:25 < warren> OTOH.... backing that up is not so important
15:25 < jcollie> mysql would scale better
15:25 < warren> yeah, msyql would scale better
15:25 < iWolf> warren: I've always used mysql with it.
15:25 < mmcgrath> I'm fine with either, we have a MySQL server so might as well use it.
15:25 < abadger1999> mmcgrath: +1
15:26 -!- glezos [n=glezosd@fedora/glezos]  has quit [Remote closed the connection] 
15:26 < mmcgrath> plus we can cluster smtp in the future should we need to.
15:26 < mmcgrath> so kim0 and paulobanon aren't able to come to the meetings very often because of work but they've been working quite a bit on the proxy caching servers....
15:27 < mmcgrath> Asside from coordination does anyone have any concerns / suggestions regarding the load test that is to occur?
15:27 < mmcgrath> I have a feeling we will need multiple tests to get good results but a first run through is key to see exactly what we're dealing within our environment.
15:28 < mmcgrath> Anyone?
15:28 < abadger1999> I think we all agree :-)
15:28 < jcollie> sounds good to me
15:29 < abadger1999> Do it once, reevaluate, do it again.
15:29 < iWolf> mmcgrath: I agree.
15:29 < mmcgrath> rock
15:29 < iWolf> the first time is the starting point to see what we need to tweak - either in our testing methodology or what we have set up.
15:29 < mmcgrath> ok, aside from project hosting thats all I've got and f13's in a meeting.
15:29 < mmcgrath> <nod>
15:30 < mmcgrath> So the floor is open, who's got what?
15:30 < jcollie> did the mod_python version of the mirrorlist script go public?
15:30 -!- JSchmitt [n=pclinux@fedora/JSchmitt]  has quit [Remote closed the connection] 
15:31 < mmcgrath> Yes it did but we had a problem with it that I've been meaning to look at.
15:31 < mmcgrath> It ran fine for hours then all the sudden started freaking out seemingly for no reason.
15:32 < f13> I'm back
15:32 < mmcgrath> unfortunately I couldn't "Make" it do it so I'm going to run some tests later this week (I guess  that means tomorow)
15:32 < jcollie> memory or fd leak maybe?
15:32 -!- mdomsch []  has quit [Remote closed the connection] 
15:32 < mmcgrath> It could be.  And it could have been some deal where apache didn't restart right, who knows.
15:32 < mmcgrath> f13: anything to discuss regarding hosting?
15:33 < jcollie> the pitivi project hosting seems to be working well... initially it seemed a little sluggish but recently it's been pretty snappy
15:33 < f13> well, hosting seems to be going reasonably well
15:33 < f13> a few more projects joined, some more coming
15:34 < jcollie> are there guidelines as to what sort of projects can be hosted?
15:34 < f13> still need to solve the raw webspace
15:34 < f13> jcollie: I'm thinking things that will wind up in a Fedora repo
15:34 < mmcgrath> 'raw webspace' ?
15:34 < mmcgrath> You're talking about storage?
15:34 < f13> also, I'd eventually like a better solution to getting trac to see the repos, as cron jobs to rsync is a bit silly
15:34 < f13> mmcgrath: yeah, a place to drop release tarballs, maybe hold a yum repo, etc..
15:35 < dgilmore> f13: do most projects run trac on the scm box?
15:35 < f13> a better repo solution would probably mean revamping cvs-int or moving hg and git off to another host, perhaps all of hosting, hg and git can all live on the same box
15:35 < f13> trac doesn't support CVS so I think we can reasonably leave it out of the picture
15:35 < jcollie> f13, that seems a bit open ended... i would limit it to projects more directly related to fedora, either the infrastructure or a "core"component
15:36 < mmcgrath> f13: I need to get ahold of stacy or some of the other RHatters and see if we can get some netapp space.
15:36 < abadger1999> f13: I think the latter makes the most sense.
15:36 < abadger1999> It makes sense to separate the fedoraproject cvs/git/hg needs from the hosted projects.
15:37 < f13> jcollie: why?
15:37  * mmcgrath brb...
15:37 < f13> jcollie: why can't we invite more upstream projects to call Fedora their home?  Get them more ingrained with Fedora softwar,e fedora releases, etc..?
15:37 < f13> something that Ubuntu has with launchpad, but we could do it with opensource software
15:38 < jcollie> f13, basically i don't know if we should be the next sourceforge or savannah or whatever (i've not really looked at launchpad)
15:38 < f13> abadger1999: I think so to.  Actually the only 'fedoraproject' stuff would really be the packageSCM.  Anything else is really just hosted content.
15:38 < f13> jcollie: there is some strong desire to provide something that isn't as craptastic as SF
15:38 < jcollie> yeah, i hear that
15:38 < abadger1999> f13: nod
15:39 < iWolf> f13: Do current resources allow us to do that though?
15:39 < f13> jcollie: I feel really bad if as a software project your only options were SF, Launchpad, or 108
15:39 < warren> jcollie, launchpad has been a powerful way to get people to work closer to Ubuntu
15:39  * mmcgrath back
15:39 < warren> I'd choose launchpad over 108
15:39 < f13> iWolf: I would think so.  It doesn't take a whole lot
15:39  * warren wonders why it was called 108
15:39 < f13> iWolf: I think we could reasonably host a fair number of projects on a single box with decent storage and backups
15:40 < mmcgrath> I think thats part of the point of 108.
15:40 < warren> (but away from cvs-int where fedora dev happens)
15:40 < f13> and eventually split out some projects to multiple trac boxes, etc...
15:40 < mmcgrath> err
15:40 < mmcgrath> I think we should aim as large as we can with this.
15:40 < mmcgrath> The community will decide what it should be.
15:40 < iWolf> f13: that is probably true, do we have storage estimates and space for backups?
15:40 < f13> iWolf: unfortunately no, it really depends on what projects come our way.  A project like kernle would take significant space.
15:40 < warren> until we have resource commitments for storage and servers, we need to be selective of who we allow on hosted
15:41 < f13> a bunch of projects like pungi, mock, readahead, etc... should all take relatively little space
15:41 < f13> warren: agreed.  Until we have a blessed hosting platform we should be selective of who we allow on hosted (:
15:41 < warren> a single active project like gaim almost singlehandedly brought sourceforge to its knees a few years ago.
15:41 < f13> right now everybody hosted understands that its a trial, a proof of concept, and may break at times.
15:41 < iWolf> I think the goal is worthwhile, I just want to be sure we have the infrastructure to reliably support it.
15:42 < f13> iWolf: indeed
15:42 < mmcgrath> In general our infrastructure is very under-utilized (this excludes the wiki)
15:42 < mmcgrath> xen will help us balance and use it better but initial adaptation will be slow to begin with.
15:42 < warren> continue trial while being selective in who we let in?
15:42 < jcollie> +1
15:42 < mmcgrath> +1
15:42 < warren> see how that goes, and ask for more resources
15:43 < iWolf> mmcgrath: yeah, the PHX could be doing much, much more.  We just need to be sure we have storage, reliable backups and that people on theteam know how to support it.
15:43 < iWolf> +1
15:43 < abadger1999> +1
15:43 < mmcgrath> Ok, so f13.  For now, full steam ahead.
15:43 < f13> col.
15:44 < f13> cool.
15:44 < mmcgrath> Alrighty, anyone have anything else to add?
15:44 < f13> if anybody has experience with general web hosting, that would be greatly appreciated.
15:44 < f13> getting people access to drop files but without being a security nightmare
15:44 < f13> shell would be nice, with the scm tools to be able to do checkouts/exports/tarballs but not 100% necessary
15:44 < warren> f13, write it in php, it'll be secure! =)
15:44 < mmcgrath> POW!
15:45 < jcollie> what about making the projects set up a special repo for web content?
15:45 < f13> and then somebody with good web coding could help by making a 'make it happen' page.
15:45 < mmcgrath> f13: I'm pretty experienced, let me know what you need.
15:45 < f13> jcollie: nah, that fails pretty badly
15:45 < f13> jcollie: hard to add a yumrepo
15:45 < jcollie> hmm yeah, that'd be bad
15:46 < f13> we should make it as easy as possible to use it.
15:46 < f13> trac makes things incredibly easy
15:46 < jcollie> you think that webdav works well enough nowadays? or should it be sftp?
15:46 < teknofile> "make it happen" page?
15:46 < jcollie> sftp would be pretty easy since everyone has a SSH key in the accounts system
15:46 < f13> teknofile: probably more for admins to 'launch' a new project.  punch in a project name, and the fedora account to be the admin and it would setup the trac space, setup the repo, add the right groups to account system, setup webspace, etc...
15:47 < f13> jcollie: indeed, thats not a terrible idea.
15:47  * f13 welcomes all ideas (:
15:47 < mmcgrath> Should we continue this on the list?
15:47 < mmcgrath> right now for package uploads in CVS it requires certification auth.
15:48 < f13> mmcgrath: list is fine sure.
15:48  * f13 notes that he doesn't have a lot of time in the next few months to do much more with it, other than add projects as they come up
15:49 < jcollie> the certs are nice, but annoying when you think that CVS is busted because your cert just expired
15:49 < jcollie> another "nice to have" feature would be to have something that would email people in the accounts system just before their cert expires
15:50 < mmcgrath> f13: you going to be around for a bit?  We can continue discussion after the meeting if we want?
15:50  * mmcgrath just likes keeping meetings short.
15:51 < jcollie> to keep meetings short you need to provide lots of soft drinks and prevent anyone from leaving until the meeting is over
15:51 < mmcgrath> heh
15:51 < jcollie> either that or schedule it just before lunch and not provide any snacks :)
15:52 < teknofile> or at 5:00PM on Friday
15:52 < f13> mmcgrath: sure, I'll be here
15:52 < mmcgrath> Cool.
15:52 < jcollie> or you can do like the Queen's Privy Council and require that all meetings be conducted standing up...
15:52 < mmcgrath> Alright, so anyone have anything before we close the meeting?
15:52 < dgilmore> mmcgrath: nothing from me
15:53 < iWolf> nothing from me
15:53 < abadger1999> nothing here
15:53 < jcollie> nothing more than idle chit-chat
15:53 < mmcgrath> Alllllright.  MEETING END ============================================