[12:02] Sounds like the rosetta instance that's currently being used fits into "test", so that leaves us needing to get the dev LP up and running, and then on the next test deployment, include soyuz and malone along with the next ver of rosetta === kiko-fud is now known as kiko [12:05] BradB: the Rosetta alpha server is not suitable for testing other things [12:06] BradB: it is on its own code branch, is not regularly updated, and has modifications in place so that Rosetta is the only application available [12:07] the test apps shouldn't get regularly updated [12:07] what;s the diff between the dev and the test lp? [12:08] dev is the one that gets updated every day with the latest and greatest changes (or breakages) [12:08] bradb: i don't think people work well in too many branches [12:08] right [12:08] test is the one that users test, and should actually expect to work relatively well, and change relatively infrequently. [12:08] i'd like to be moving to production as frequently as is sane [12:08] weekly, or fortnightly [12:09] so there's not really much difference between dev and test [12:09] and i don't think it's worth having it [12:09] not at this stage [12:09] when lp is huge with tens of thousands of users, yes [12:09] but right now we want it to evolve FAST [12:09] so prefer to keep it lighter weight [12:10] ok, so first step, we want testing LP that is running all three apps, right? what happens to the rosetta alpha? [12:10] dev is the one that gets updated every N minutes, not every day [12:11] I think we should update soon the rosetta alpha version with some bug fixes [12:11] the main point of the dev one is so that lu and others can see changes that developers make hour-by-hour [12:12] daf: did you merged the patchsets to fix the language_country language selection? [12:12] rather than waiting a day to see some contentious change take effect [12:13] SteveA: you can see the hour-by-hour status on your own dev box [12:13] yes [12:13] rosetta.wh.h.c [12:13] but can lu? [12:13] but we should hjust blow that db away every half hour too [12:13] oh, you mean the dev server on mawson [12:13] I thought you were proposing to do away with that [12:13] the dev box is a daily build, with data we care about enough to backup regularly [12:14] no [12:14] i'm tired, so i may not be explaining this clearly [12:14] heres what i would suggest we have: [12:15] (a) crack of the hour (rosetta.wh.h.c equivalent) that nobody uses except to poke people at a brand new feature, blows away its own code abd db every half hour [12:15] (b) dev / dogfood box, updated daily to latest db and code schema, with specific merges by the team leads to fix critical issues that come up [12:15] (c) staging server ONLY FOR THE HOUR OR TWO BEFORE A PRODUCTION UPLOAD [12:15] (d) production server [12:16] stub will focus mostly on (b) [12:16] then (c) and (d) during a push to production [12:16] we should push to production weekly or fortnightly [12:16] time based releases [12:16] use tests to ensure the whole system is always deployable [12:16] phew [12:17] how's that sound [12:17] sabdfl: I suggest you write up your proposal to the mailing list [12:17] for the benefit of those not present [12:17] sounds good to me. I imagine staging will be hit by automated tests, not people so much. [12:18] timeforbed [12:18] SteveA agreed [12:18] staging is a place to bring in a dump of the production db, run the db update scripts, run the tests, and see if it all looked to work. then repeat on production server [12:19] it's a defined process, not a place for arbitrary work [12:19] BradB: any success on mawson? [12:20] sabdfl: i stopped deploying, based on the current discussion (which started somewhat earlier) [12:20] does this proposal sound sane to you? [12:21] it makes sense [12:21] we bring up malone. we get kinnison ad cprov to bring up soyuz. [12:21] three seems like enough, but if you really want to maintain a forth, i won't object. [12:21] and when daf is ready they migrate the rosetta alpha to the dogfood platform [12:21] fourth even [12:22] r.w.h.c is on autopilot [12:22] all our work is focused on the dev [12:22] staging is a pritine environment for stub to blow away, test, blow away again [12:22] production is production [12:22] this is less fragmented than having a rosetta "test" and a malone "test" and a soyus "test" [12:23] isn't this more complicated than what i suggested of just dev, test, and live? [12:24] (one dev LP where anything goes, one test LP that should be fairly sane, and one live LP where lives hang in the balance.) [12:24] it's the same, just addin the pristine staging server we always planned on [12:24] so dogfood = dev [12:24] same proposal [12:25] keep going :-) === carlos is tired [12:26] daf: please, send me an email with a "go" to the merge if you are able to test it today so I can do the merge tomorrow morning. [12:26] good night === BradB starts build-config again, on patch-659. [12:27] I can have that up and running in however much CPU time it takes, less the appropriate Apache configuration. [12:27] cool [12:27] elmo will sort out apache config in a tick [12:28] basically, no reason not for it to be identical to the current rosetta.wh.h.c [12:28] client cert for vpn effect [12:28] we should probably get rid of the rosetta.w.h.c domain [12:28] as an alias for mawson [12:28] daf: right, just run it on mawson, on a different port? [12:29] it's purely to show off brand spanking new features [12:29] rosetta.w.h.c is the same machine as mawson.ul.o [12:29] diff port [12:29] diff db [12:29] you're thinking of rosetta.sf.o [12:29] (I suspect) [12:30] no... [12:30] that's the "alpha" [12:30] ok [12:30] https://mawson.ubuntulinux.org is exactly the same as https://rosetta.warthogs.hbd.com [12:30] rosetta.wh.h.c is the "update every half hour to crack of the minute db and code" [12:30] NOW it is [12:30] right [12:30] but? [12:30] but bradb is bringing up the dev / dogfood server on mawson too [12:31] or not? [12:31] i am [12:31] right [12:31] waiting on the build-config [12:31] sabdfl: are we going to use rosetta.w.h.c for something difference? [12:31] so... r.w.h.c is just for kicks, its just a "public" snapshot of your and my dev box [12:32] daf: please read what i wrote [12:32] the dev server will only be updated once a day [12:32] I'm saying I think we should stop using that domain [12:32] to avoid confusion [12:32] *relatively* stable [12:32] daf: agreed [12:32] ok, then we were agreed all along :) [12:32] let's call them the bleeder, the feeder, the staging and the breeder :-) === sabdfl ducks and weaves [12:33] the bleeder should take none of our time [12:34] it just happens in the background [12:34] the feeder is where we really test [12:34] it's running a relatively full db [12:37] and it's hosting guests working on rosetta beta code [12:37] and it's hosting our own launchpad bug tracking instance of malone [12:37] staging is just for stub and lifeless [12:37] production is production [12:37] ok, bleeder, feeder and breeder are terrible names :-) [12:37] but bleeder, dailydev and production might be better === debonzi_ [~debonzi@200-206-134-238.async.com.br] has joined #launchpad === debonzi_ is now known as debonzi === BradB goes to get food while build-config continues chugging === BradB is now known as BradB|food === BradB|food is now known as BradB === debonzi [~debonzi@200-206-134-238.async.com.br] has joined #launchpad === debonzi [~debonzi@200-206-134-238.async.com.br] has joined #launchpad [03:15] daf: How bad will it be if I blow away rosetta's DB on mawson? [03:27] daf or elmo_: ping [03:28] I need someone to stop rosetta so that I can create malone_alpha [03:30] BradB: Ignore what I said in email about the header rewriting - just changing the destination address in the envelope (what is happening now) is fine. That said, I'd already implemented the (optional) full header rewriting before I realised it so I'll check it in anyway. [03:33] Should I document this sort of thing on the Wiki or in a README.email ? [03:35] Mmm.... Malone is starting to look like a real system :-) [03:35] wiki or README, heads or tails, hm [03:35] I'd say a README [03:37] Mmm... I prefer that too... keep it in the same RCS. [03:37] yep [03:37] as for malone: once edit notification works (which i almost was able to wrap my brain around earlier), the 80%'ll be done. [03:39] I guess that email thing is pretty cool. Where do bounces go though? [03:40] e.g. if the envelop says From: test@bbnet.ca, and the From: header says steve@canonical.com, I hope it bounces to bbnet.ca :) [04:05] At the moment there is an Errors-To: header being set to 'nobody@example.com' along with a TODO. [04:06] BradB: I opened a bugzilla bug about configuration information via ZCML, as it should be configurable rather than hard coded. [04:11] Feel free to change the default if your mail server doesn't mind everyone spamming it :-) [04:21] We will need to do bounce processing in the future to detect expired email addresses, but I think it is ok to just check them in /dev/null for now (?) [04:23] Merge to rocketfuel@canonical.com/launchpad--devel--0: Header rewriting and email documentation (patch-670) [04:42] BradB: "rosetta's DB"? [04:54] on mawson [04:55] stub: Probably, only reason I was asking about bouncing was perhaps slight over paranoia of real emails getting spammed when tests with expired recipient addresses get run. [04:57] Yup. I think it is very valid paranoia - the email system needs to get this stuff right. [04:58] Errors-To: should cope, or the header rewriting if we want to be doubly sure. [05:43] BradB: you mean the "launchpad_test" DB on mawson? [05:44] BradB: trash that as you like [05:45] is that what you're running on? oh. [05:45] presumably i can't trash it as i like though [05:46] nope, that definitely wouldn't be a good idea ;) [05:46] but i need rosetta stopped momentarily, so that i can create malone_alpha [05:46] daf: can you stop rosetta on mawson momentarily while i do that? [05:47] oh, ok [05:47] when I say you can trash the launchpad_test DB, I mean it [05:47] it doesn't contain any critical information [05:47] other than the schema :) [05:47] the schema cna be recreated also [05:47] there's no guarantee that it'll be the same as what i'm using [05:48] there's a separate DB for the alpha, which actually should not be changed [05:48] oh, ok, well anyway, i'm creating a separate db anyway [05:48] it sucks how Postgres won't let you create new databases while other databases are being used [05:49] yeah, i don't understand why [05:49] ok, stopped [05:49] do I need to stop the Alpha also? [05:50] nope, db recreated, you should be able to restart the thing you just stopped [05:51] huh? [05:51] weird [05:51] the Rosetta alpha was using the DB [05:51] the test LP was using the DB [05:51] but I only needed to stop one of them? [05:53] *shrug* [05:54] eh, how do i start an LP app on this machine? [05:54] if I run it as the launchpad user, i get a perms error on a log file [05:55] if i run it as bradb, i can't access the db [05:56] which log file? [05:57] /home/bradb/launchpad-malone-dogfood/launchpad.log [05:57] if you want to run it as launchpad, you'll probably need to make a copy of the checkout as launchpad [05:58] how are you doing it with rosetta? [05:58] e.g. sudo -u launchpad "cp -a ~bradb/launchpad-malone-dogfood /home/launchpad; cd /home/launchpad/launchpad-malone-dogfood; make run" [05:58] actually, the existing instances are still running as me [05:58] the launchpad user doesn't have a home dir :) [05:59] I should change it, but I always have more urgent stuff to do [05:59] oh, I think the launchpad user has permissions in /srv or someplace like that [05:59] how did you auth to pgsql then? [05:59] um [05:59] if you're running rosetta as daf [05:59] I think the DB has permissions for either me or LP to access it [05:59] both databases [06:00] because you created the new DB as bradb, the launchpad user can't access it [06:00] i created it as launchpad [06:00] oh, hmm === daf shrugs [06:18] daf: can you access https://mawson.ubuntu.com:9020/malone? [06:21] no [06:21] it won't work [06:21] because it is firewalled [06:22] ah, so it was more than just apache config :) [06:22] well [06:22] Apache acts as a proxy [06:23] Apache listens to ports which are allowed through the firewall, and proxies ports which aren't [06:24] proxy on block ports? i don't follow... [06:24] s/block/blocked/ [06:26] Apache listens on port 443 [06:26] and forwards requests it recieves based on the virtual host [06:26] sure [06:27] i'll get james or thom to set me one up tomorrow then [06:29] best to send the email now === BradB sends, sleeps === BradB is now known as BradB|zzz === stub [~stub@dsl-246.248.240.220.dsl.comindico.com.au] has joined #launchpad [11:45] SteveA: around? === Kinnison [~dsilvers@81-178-238-108.dsl.pipex.com] has joined #launchpad [11:55] I guess I should be here too [11:57] Merge to rocketfuel@canonical.com/launchpad--devel--0: Tidying lucille ready for renaming views etc (patch-671) [11:57] thanks babe. [11:58] sabdfl: Right; I'm doing the view rename/column rename we discussed on wednesday [11:58] Kinnison: cool, thank you [11:59] sabdfl: The views I have to rename are... BinaryPackageFilesToPublish SourcePackageFilesToPublish PublishedBinaryPackageOverrides and PublishedSourcePackageOverrides [11:59] sabdfl: any suggestions on better names? [11:59] Overrides needs rethinking, and better to do it with grep and %s/x/y/ sooner rather than later [12:00] It was the field names that I thought needed to be the same as the field names they map to [12:00] binarypackagename instead of bpname [12:00] yeah; the column names I'm clear on [12:00] let's use a given name as consistently as possible [12:00] "ToPublish"? [12:02] Those views list packages pending publishing [12:02] I.E. sourcepackagepublishing and packagepublishing records which have status == PendingPublishing [12:04] sabdfl: yep [12:05] hi SteveA [12:06] SteveA: blush, forgotten what i was going to ask [12:10] Anyone mind if I rename the databases, so launchpad_dev will be what is used when you do 'make run' and all the tests use launchpad_ftest? I'm fixing tests atm, and some of them are broken because they talk to the launchpad_test database. [12:10] sabdfl: How about PendingBinaryPackagePublishings and PendingSourcePackagePublishings ? [12:10] I know Daf is for it (from email about a week ago...) [12:10] Kinnison: sounds perfect [12:11] stub: yes please! [12:11] sabdfl: and for the other two... PublishedSourcePackageLocations and PublishedBinaryPackageLocations ? [12:12] the old overrides... is it really just locations? [12:12] isn't it also quite a bit of "we want this, we don't want that"? [12:12] It's not really locations at all [12:12] it's component/section/priority [12:12] ArchiveLocations [12:12] ? [12:12] The 'want this, don't want that' choice is the filelist which is the next bit on my list of things to do [12:13] ok [12:13] So PublishedBinaryPackageArchiveLocations ? [12:13] loooong but true [12:13] Then there'll be Published{Source,Binary}PackageFileList [12:13] hold on [12:13] is there any other view of "published packages"? [12:14] if not, then the rest is spurious [12:14] if this is our primary view of the packages that are published, then just call it "PublishedSourcePackage" === Kinnison ponders [12:14] and the fact that it includes the archive lcoations is nice [12:14] the primary way to see what's published is to look at the sourcepackagepublishing and packagepublishing tables [12:15] we do need to rename the Overrides table [12:15] right [12:15] these views are collating information for lucille to generate her output [12:15] hmm.... so what tables do they pull data from? [12:15] loads [12:16] E.g. 'SourcePackageFilesToPublish' draws from... sourcepackagepublishing sourcepackagerelease sourcepackagereleasefile libraryfilealias distrorelease sourcepackage sourcepackagename and component [12:17] PublishedSourcePackage then [12:17] and PublishedBinaryPackage [12:17] cool! [12:17] simpler better faster gotta love it [12:17] These views are listing *files* not packages [12:18] (assuming the ...ToPublish views are what you're on about now) [12:18] ... i can live with it === elmo_ [~james@82.211.81.249] has joined #launchpad === Kinnison thinks we're talking at cross-purposes [12:18] give me a sec to regroup === Kinnison -> thinking room [12:18] Kinnison: you're right, sorry [12:19] PublishedSourcePackage would be for the alread-published files [12:19] which tables does that draw from? [12:19] same list? === Kinnison returns [12:23] PublishedSourcePackage draws from: SourcePackagePublishing, DistroRelease, SourcePackageRelease, SourcePackage, SourcePackageName, Component and Section [12:25] Columns it provides are: [12:26] id, distroreleasename, sourcepackagename, componentname, sectionname and distribution [12:26] well; theoretically it provides sourcepackagename.name [12:26] What should I call that column? [12:32] sourcepackagenamename ? === Kinnison pokes sabdfl [12:51] Kinnison: pong [12:51] sabdfl: ^^^^^^^^^^ [12:51] just sourcepackagename [12:51] even though that has connotations of being a foreign key onto the sourcepackagename table? [12:52] a bit of a fiddle, that one [12:52] sourcepackagenamereally ;-) [12:52] so PublishedSourcePackage i'm happy with [12:52] I don't mind any of: sourcepackagename sourcepackagenamename or "sourcepackagename.name" [12:53] sourcepackagename [12:53] although sqlobject might barf at the latter [12:53] okay [12:53] what about the name of the other views? [12:53] Pending.? [12:53] PendingSourcePackageFile [12:53] PendingBinaryPackageFile [12:53] ? [12:54] Kinnison: works for me [12:54] Cool [12:54] I'll do that; and then I'll do PublishedSourcePackageFile and PublishedBinaryPackageFile for the new views I need to do too [12:55] and rename Override? [12:55] Override? [12:55] launchpad_dsilvers=> \d override [12:55] Did not find any relation named "override". [12:55] We're renaming PublishedBinaryPackageOverrides to just PublishedBinaryPackage I thought we agreed that up there somewhere? [12:58] Kinnison: yes, that's the view, right? [12:58] yep [12:58] there are no 'override' tables [12:58] but there is an underlying Override table somewhere [12:58] oh [12:58] NupNup :-) [12:58] The overrides *are* the publishing records :-) [12:58] where do you store the info about how the distro wants to publish a given package? [12:58] "when the NEXT package foo comes in, publish it in component/section/priority" [12:59] that's what I understood an override to be [12:59] when foo 1.0-2 comes in, you copy the publishing records from 1.0-1 pointing them at the new package [12:59] Or at least; that's how I was going to do it [01:00] Since that's the obvious way; then if you don't have enough context to do that; the package is clearly NEW and should be queued in the NEW queue instead === Kinnison -> lunch [02:11] Kinnison: i think we need an explicit policy table. when a new package comes in you look at the policy table to decide what to do with it [02:11] we can start with a straight line-per-package approach like the debian overrides system [02:11] but later it needs to be "accept packages that are in this bundle" (gnome / kde / etc) [02:11] and "reject packages that are in that bundle (perl / dev / etc) [02:12] so that derivatives get to shape their distro at a higher level than package-by-package. [02:23] SteveA: kamion just reminded me what i wanted to ask [02:23] wiki / zwiki decision is now urgent [02:23] what's your view on the best course of action? [02:26] What decision exactly do we need to make? Whether to use ZWiki, integrated with the UL plone site, or to use moinmoin not so integrated? [02:36] It seems that there are no tables used on wiki.ubuntu.com === BradB|zzz is now known as BradB [02:38] I did wget -m http://wiki.ubuntu.com/ and grepped the results case-insensitively for " So, we can use ZWiki, and move content across from wiki.ubuntu.com [02:42] elmo_: ping [02:44] BradB: [02:44] SteveA: ?? [02:44] stevea: check out hardware support .. [02:45] elmo_: Two things: [02:46] 1: Will you have a chance to do the Apache config for Malone soonish? [02:46] 2. Do you care if I create another postgres superuser on mawson? (bradb) [02:47] 2: yes, run stuff as launchpad please [02:47] There's a problem with that. [02:47] services should not be tied to one user's account [02:47] why? [02:47] elmo_: ah -- wget didn't get second level pages... [02:48] The problem is that the zope instance needs to write logfiles. the launchpad user can't write logfiles in ~bradb. [02:48] Unless I'm meant to have another dir under /srv [02:51] yeah, create something under /srv [02:52] hmm, there's meant to be a /srv/launchpad or something - if there isn't just let me know [02:52] have you guys agreed on the url stuff then? [02:52] yep, i put it in an email to you guys (i.e. the admins) last night [02:53] SteveA: How recent is lib/canonical/functional.py ? I think it has superceeded lib/canonical/tests/functional.py but I'm not sure. [02:53] yeah, I know, just wanted to make sure it's final [02:53] elmo_: afaic, yes :) [02:54] oh, sorry I should have read that mail [02:54] malone.w.h.c doesn't work - there's like 24 hour+ turn around time on that DNS [02:54] if you pick *.ul.o or *.u.c or *.c.c etc., I can do it.. otherwise.. === carlos [~carlos@69.Red-80-33-181.pooles.rima-tde.net] has joined #launchpad [02:55] yeah, like i say, i'm not picky. m.u.c would be fine. [02:55] SteveA: ok, and the moin support is working, and installed in our copy of zwiki on www.ul.o? [02:56] elmo_, SteveA: the zwiki supports native html for tables [02:56] elmo_: can you create a dir /srv/malone? /srv/rosetta.* doesn't seem like an appropriate place for it. [02:56] if necessary we can munge them over as html tables directly [02:56] BradB: morning! [02:56] hi :) [02:56] sabdfl: that's super user unfriendly for editing [02:57] sabdfl: okay for malone.u.c [02:57] +? [02:57] bradb: re /malone is this for the dev server? it will host ALL the lp componets though [02:57] stub: lib/canonical/functional.py is the newest and best [02:57] elmo_: it's much better for searhcing [02:58] i think we are going to put together a dedicated Plone class for the hardware db [02:58] sabdfl: so launchpad.ubuntu.com? [02:58] BradB: perfect! [02:59] elmo_: Can you make that dir /srv/launchpad instead please? :) [03:00] tramps. this is why I asked if you guys were done "discussing" :P [03:01] i had got the impression that i was just putting up a malone df, but i guess not :) [03:01] sabdfl: the latest released ZWiki is 0.35.0. This is not installed in our plone sites. It has support for moinmoin pages without tables. The preview mode was added one week later, and has been improved since. [03:01] There will be a new wiki release on 1 November containing the preview mode, and any further work done on moinmoin markup. [03:01] If we want to get the preview mode before then, we'll need to check zwiki out of its DARCS repository. [03:02] SteveA: how extensive are the tables in hte hardware section [03:02] what client cert do we want protecting this? the testers one? [03:02] or a 'warthogs' one? [03:02] elmo_ accept either, allow testers into /rosetta but for everything else require warthogs [03:02] sabdfl: there are 11 hardware pages. most have a table on them. [03:03] good thing I checked if there was already a database called "launchpad_alpha" :) === BradB uses one called launchpad_dogfood instead [03:03] I could convert the tables to restructured text format in under 30 minutes in vim [03:04] SteveA: is the ReST table format efficient? moin is terrible [03:04] it is an improvement over moin, but still a bit awkward. Here's an example: [03:04] http://docutils.sourceforge.net/docs/user/rst/quickref.html#tables [03:05] I'm looking at the "simple table" format [03:05] elmo_: Can you stop rosetta too please so that it doesn't hang when I try to create the DB? [03:06] "stop rosetta" ? [03:06] you mean UP it? [03:07] kill it [03:07] daf's running it on mawson [03:07] why would rosetta hang? [03:07] it uses a different db, doesn't it? [03:07] rosetta doesn't hang [03:07] postgres' fucked - I'm going to try UP-ing rosetta to see if it helps [03:07] why would your stuff hang on account of rosetta running? [03:07] me trying to create launchpad_dogfood hangs, because "another process is accessing the database" [03:07] err, UP-ing mawson [03:08] this is why we changed the machine name isn't it :) [03:08] so postgres doesn't like you doing stuff to the databases when a process is using a different database? [03:08] that sucks [03:08] indeed [03:08] makes me want to run several different postgreses [03:09] it seems to me that there must just be a problem with the pgsql config on mawson [03:09] no, you tramps, it's a bug [03:09] maybe it's fixed in 8.0b3? [03:10] sabdfl: I think HTML tables are the way to go, unless people generally use a decent editor to edit them. [03:10] we're not using beta's in production [03:10] and changing major versions between production and devel is a reciepe for madness [03:10] yeah, wasn't suggesting that, but optimistically looking towards the 8.x series :) [03:10]