/srv/irclogs.ubuntu.com/2012/04/12/#juju-dev.txt

TheMuemorning09:33
wrtpTheMue: yo!09:39
niemeyerHello all!12:48
andrewsmedinaniemeyer: hi12:53
niemeyerandrewsmedina: Heya12:54
=== chuck_ is now known as zul
=== chuck_ is now known as zul
niemeyerTheMue: ping14:29
TheMueniemeyer: pong14:30
niemeyerTheMue: Heya14:30
TheMueniemeyer: hi14:30
niemeyerTheMue: How're things going there?  Good progress on the watchers stuff?14:30
TheMueniemeyer: now yes, had two hard days hunting a bug14:31
niemeyerTheMue: Oh?14:31
TheMueniemeyer: yeah, didn't wanted as i wanted it to do14:32
niemeyerTheMue: Cool.. glad you're unblocked14:33
TheMueniemeyer: first i had to find out that a watch of GetW() won't fire later if a non-existing node is created (like the watch of ExistsW())14:33
niemeyerTheMue: Yeah, it fails immediately I believe14:34
niemeyerTheMue: So the watch is never established14:34
mthaddonniemeyer: have you seen there are some questions outstanding on the charmstore RT?14:36
niemeyermthaddon: Nope14:36
niemeyermthaddon: I missed those14:36
TheMueniemeyer: and in the tests i had to put those changes i'm watching in a goroutine. you'll see, i'll check in today14:36
niemeyermthaddon: I'm working on stats, btw14:36
niemeyermthaddon: Almost ready, but need some reviews14:36
mthaddonniemeyer: ok, thx - if you could take a look and get back to me (via the RT or here) on the questions when you get a chance that'd be great14:43
niemeyermthaddon: Definitely14:44
niemeyermthaddon: Done15:07
niemeyerI'm stepping out for lunch, biab15:07
flacostehi SpamapS, are you uploading a new juju version before final freeze (today)? the archive version still doesn't support constraints15:43
hazmatfwereade_, i'm looking over the orchestra constraints branch,d o you have a moment to chat regrading15:46
fwereade_hazmat, ofc15:46
hazmatmostly i'm curious how reliable the cobbler profile data and arch data is15:46
fwereade_hazmat, it's installed as part of orchestra and so should be solid; but it is true that I have no way of dealing with an installation that someone's messed around with enough15:48
fwereade_hazmat, OTOH setting up distros/profiles is a bit headachey and IMO one of the really nice things about orchestra is that it does that for you so you don't have to worry about it15:50
flacostefwereade_, hazmat: shouldn't we just remove the orchestra provider, now that we have maas?15:50
flacosteDaviey: wdyt? ^^^15:50
flacosteDaviey: what's the plan with the packages in the archives?15:51
flacosteleave it universe?15:51
hazmatflacoste, from a juju perspective, maas for 12.04 has significantly less features15:53
flacostehazmat: really, how so?15:53
hazmatflacoste, because it can only address machines by name15:53
hazmatie. poor constraints support15:53
fwereade_flacoste, I expect we will want to retire the orchestra provider sooner rather than later though15:53
Davieyhmm15:57
DavieyWhat can the juju prover for orchetra do?15:57
DavieyThere isn't a problem leaving it in i guess15:58
DavieyI mean, it's not exposed to the user, right?15:58
flacostehazmat: but you still have to implement the orchestra constraints support, dropping it would means less maintenance :-)15:59
hazmatit can use a specified orchestra class as a constraint15:59
hazmatflacoste, the orchestra class support for constraints is already in15:59
hazmatthere's additional branch in review which i'm looking at that adds arch/series15:59
fwereade_hazmat, sadly, it *doesn't* add arch, mainly because I am an idiot :(16:01
fwereade_hazmat, it matches existing arch16:01
fwereade_hazmat, that branch is a result of me suddenly realising that ubuntu-series is critical, and totally unaddressed in orchestra16:02
Davieyfwereade_: *please* tell me you aren't doing development for the orchestra provider at the moment.16:03
flacosteDaviey, hazmat, fwereade_: yeah, that's my point about reducing maintenance16:04
flacoste:-)16:04
flacostenot doing work is great!16:05
flacostethe zen of non-doing16:05
hazmatDaviey, flacoste  having a useable baremetal solution that works with juju's expectation of a provider is important16:05
hazmateven without cosntraints, right now neither maas nor orchestra does  that  because their ignoring series16:06
hazmatwhich is critical to being able to deploy a charm as intended because the charm tied to a series16:06
flacosteright, but maas only support precise16:06
hazmatand we have zero precise charms atm16:07
hazmatthat's supposed to change this week16:07
hazmatbut come 12.10 we either expect folks to move core  infrastructure pieces off an LTS release for a newer version of maas?16:08
fwereade_Daviey, sorry to say I am, simply because we have this provider that (1) still exists and (2) doesn't work right16:08
Davieyfwereade_: wow.. if it doesn't ork.. drop it.16:09
DavieyReally.. It's a *total* waste of time working o Orchestra.16:09
DavieyWould it help if i removed orchestra fro the archive?16:10
flacostefwereade_: i agree16:10
flacosteDaviey: we are getting the following error when bootstrapping:16:10
flacosteSSH authorized/public key not found.16:10
flacoste2012-04-12 12:09:14,139 ERROR SSH authorized/public key not found.16:10
fwereade_Daviey, flacoste: I would personally have no problem with that, but it feels like this is a very late stage to be doing this16:10
hazmatwoah.. it does work16:10
Davieyflacoste: fixed16:11
flacosteDaviey: how did you fix that?16:11
flacostei see16:11
flacosteyou just created an ssh key16:11
Davieyright16:11
flacosteDaviey: pgraner says hold on your fire :-)16:12
flacostebut thanks!16:12
* Daviey lets go of the fire.16:12
flacosteDaviey: +1 to dropping orchestra from the archive :-)16:12
flacosteDaviey: better to be honest to our users about what's happening here16:12
flacostehazmat: yes, it's expected that people who want to use 12.10 will use more recent version of juju and maas16:13
fwereade_Daviey, flacoste: my understanding from Budapest was that we'd *maybe* drop orchestra if we had maas doing everything we wanted by 12.04; when I came back to the python to resurrect the constraints work a couple of weeks ago, it was still around and maas was still a little up in the air16:13
fwereade_Daviey, flacoste: I guess I should have made a bunch of noise about it then16:13
flacosteDaviey: Invalid host for SSH forwarding: ssh: Could not resolve hostname node-00e081d1b147.local: Name or service not known16:13
Davieyflacoste: that just means it's not switched on/installed.16:14
flacosteDaviey: did we set it up with IPMI?16:15
fwereade_Daviey, flacoste: I think I may have been unclear: the specific thing that doesn't work with orchestra is series selection (that is, deploying oneiric/foobar may or may not get an oneiric machine on which to run fobar)16:17
flacostefwereade_: right, but that's still work that has to happen for an unsupported piece of software16:18
flacostei'd say remove it from juju and leave the community to support it if they need it16:18
hazmatplease don't remove orchestra from the archive.. for 12.04 its the only bare-metal provider that will work well with juju both for constraints and soon series.16:19
hazmati understand that orchestra is dead, but maas doesn't fufill the needs of users who want to use it with juju at the moment imo16:20
hazmatfuture dev will resolve that i'm sure, but that's not about 12.04.16:20
flacostehazmat: maas works well enough for precise16:25
flacostekeeping orchestra around just confuses thing16:25
fwereade_Daviey, flacoste: that's a decision I would have been comfortable supporting *if* we'd had a public discussion about it a month or so ago, and determined that maas was in a position to entirely replace it; but (1) we didn't, and (2) it isn't16:25
flacostei'd rather handle flak for going without 1 and 216:26
flacostethan handle the confusion for the next 5 years16:26
fwereade_flacoste, ok, it works well enough for precise; but it seems to me to be bad form to entirely drop the ability to provision pre-precise machines16:27
flacostewhy?16:27
flacostethey can stick with 11.10 if they like it16:27
Davieythis is flippin' crazy16:28
fwereade_flacoste, largely because I imagine people to have been using it to deploy pre-precise machines and I'm not sure a supposed upgrade should take features away from people16:28
flacostefwereade_: we remove features all the time16:29
flacosteespecially ones that have few users16:30
flacosteotherwise you just grow a pile of unmaintainable mess16:30
flacosteit's not like we are forcing upgrades on anyone here16:30
fwereade_flacoste, fair point; I think there's a disconnect somewhere between this and the "we have users now, we can't just keep breaking their installations on upgrade" attitude I have recently developed16:32
fwereade_flacoste, does anyone have any numbers on what providers people are using?16:33
flacostefwereade_: you should know this better than I :-)16:33
flacostemaybe jorge would know16:33
fwereade_flacoste, my main worry is that *anyone* wanting to use juju for bare metal will have been using orchestra, and that if we trash that and offer only a less-featureful alternative we'll be chasing them away16:34
flacostefwereade_: it's not less featureful16:35
flacostestop that16:35
flacostethese "more" features don't exist yet16:35
flacosteconstraints -> that's not a feature they can use16:35
flacosteit requires more development to support16:35
flacosteit's more featureful than the existing setup16:35
fwereade_flacoste, depending on exactly what machines you have set to use juju, it could be argued either way16:36
fwereade_s/to use/to be used by/16:36
fwereade_flacoste, if you currently have a couple of dozen oneiric machines running oneiric charms, it's fine; and (1) that's what any users will have been using, and (2) maas can't do that16:37
hazmatconstraints support for orchestra is already in trunk16:38
niemeyerwrtp: ping16:38
wrtpniemeyer: pong16:39
niemeyerwrtp: Heya16:39
niemeyerwrtp: How're things going there?16:39
fwereade_flacoste, hazmat: indeed, and has been for a while; this is more a bug than a feature, because I just wasn't handling every applicable constraint correctly16:39
wrtpniemeyer: pretty slow getting back up to speed yesterday, but more or less back in the groove now, doing the tests for the ssh forwarding stuff16:39
wrtpniemeyer: not entirely convinced that my approach to testing error paths was the correct one. i think perhaps i *should* try to provoke all the desired misbehaviours from ssh.16:40
wrtpniemeyer: am intending to get around to looking at your reviews soon!16:41
wrtpniemeyer: how about you?16:42
niemeyerwrtp: Super16:45
niemeyerwrtp: I'm good as well.. have been a bit slow early this week too, but also getting up to speed16:46
niemeyerwrtp: Happy about the stats stuff that's coming into the store16:46
wrtpniemeyer: cool. it'll be nice to see some feedback.16:47
wrtpniemeyer: (from the store traffic, that is)16:47
wrtpniemeyer: are we still planning a meeting on mondays?16:47
niemeyerwrtp: Yeah16:48
niemeyerwrtp: We've lacked the last couple of meetings16:48
niemeyerwrtp: But we should definitely do them from now on16:48
wrtpniemeyer: sounds good.16:48
wrtpniemeyer: perhaps we could do them when you first come on line (given you're in the latest time zone)? i usually have to finish fairly promptly on monday evenings.16:49
fwereade_wrtp, +116:50
niemeyerwrtp: Isn't the timing we agreed to good?he timing we've agreed to sounds16:51
niemeyerErm16:51
niemeyerwrtp: Isn't the timing we agreed to good?16:51
wrtpniemeyer: i don't remember agreeing to a timing...16:52
wrtpniemeyer: perhaps i'm missing a calendar entry16:52
niemeyerwrtp: 14UTC was the one we settled on16:53
wrtpniemeyer: that sounds fine to me.16:54
wrtpniemeyer: i'll put an entry in my calendar16:56
niemeyerwrtp: Cheers16:57
wrtpniemeyer, fwereade_: i've made a calendar for us and added the meetings to it. i dunno if that's the best way to do it.17:08
niemeyerwrtp: We can try it out and see17:10
wrtpniemeyer: at least it makes it easy for us to amend the dates and times easily and be reminded of the change...17:11
hazmatbcsaller, ping17:55
bcsallerhazmat: hey17:55
hazmatbcsaller, are we ready to get the last subs branches in..17:55
bcsalleryou approved it yesterday17:56
hazmatbcsaller, right.. but is it merged?17:56
bcsalleryes17:56
hazmatsweet!17:56
bcsaller:)17:56
wrtpniemeyer, fwereade: i'm off. see you tomorrow.18:02
niemeyerwrtp: Have a good EOD18:03
wrtpniemeyer: will do, ta!18:03
hazmatbcsaller, the change to lxc broke the tests18:15
hazmatactually the env.defaults thing that i thought got removed..18:15
bcsallerugh, those didn't run on my last try (and yes that should have been removed), checking18:16
hazmatbcsaller, i'm on it18:21
hazmati'm rolling up the fix into my next merge18:22
hazmatits a one liner18:22
bcsallerhazmat: thanks18:23
hazmatbcsaller, i was realizing though that the subordinate state change from last week is still going to cause issue with extant envs because of the topology change, even with the transparent upgrade... actually because of the transparent upgrade18:24
hazmatbecause there are older agents running around18:24
bcsallerw/o that upgrade code...18:25
hazmatits actually worse than not upgrading it18:25
bcsallerhmm18:25
hazmatbecause all of the agents are going to fail simulatenously without any user warning18:25
hazmati think we need to disable that migration18:25
hazmatits changing the topology format, so really nothing run older code is going to survive.18:26
hazmatits better to just error out on the cli b4 that happens18:26
hazmatbcsaller, thoughts?18:27
bcsallerits odd, because even as far back as reviewing the spec we thought this was a good idea. I'm trying to think how'd we make migration like this work at all if this is the case now18:27
bcsallerwould we force the agents to restart?18:28
hazmatbcsaller, they won't get the new code. restart doesn't help18:28
hazmatwithout consistent code versioning and a notion of code/db upgrade as an explicit op, incompatible state changes are hosers18:28
bcsallerI guess it could support returning the topo in the requested version as the data is still there, but that wouldn't work in all cases18:29
bcsallerand would still require eating this atleast once to get there18:30
hazmatbcsaller, we've got maybe 30m to resolve this, afaics the only thing to do is to disable the transparent topo migration18:30
bcsallerthen thats what we should do I guess18:31
hazmatbcsaller, do you see any alternative?18:31
bcsallerno, some of its a data reorg and some of its new data18:32
bcsallerthe new data doesn't leave us a choice18:32
bcsalleror we could just forward from 1 on read for now (for v2 clients)18:33
bcsallerbut that new data has to live in there18:33
hazmatnew data isn't a problem by itself dependin on the struct (ie a new key in a dict is fine), its the struct change that kills it.18:34
bcsallerthen v2 clients will just refuse to connect to v1 topos for now18:35
bcsallerhazmat: so rather than call migrate you just raid incompat version. Do you want to make that change?18:37
hazmatbcsaller, done18:39
bcsallercool. It seemed like a good plan at the time :-/18:39
hazmatbcsaller, jimbaker, fwereade_  all changes are in, pls consider trunk frozen19:00
niemeyerWoohay!19:13
flacostehazmat: so juju is now complaining on juju deploy mysql19:22
flacosteabout series constraints not being implemented19:22
flacosteError processing 'cs:precise/mysql': entry not found19:22
flacosteand Bad 'ubuntu-series' constraint 'oneiric': MAAS currently only provisions machines running precise19:22
flacosteif i try to do oneiric/mysql19:22
flacostewhat's my work-around until there is a precise mysql charm?19:23
jimbakerhazmat, awesome news!19:24
hazmatflacoste, with maas.. you'd have to pull the charm down to local disk19:35
hazmatand put in a charm repo under a precise series19:35
hazmatthe precise version of that charm does not exist in the store19:36
flacostehazmat: that's what i'm doing now19:39
niemeyerhazmat: Are you querying the cs automatically to feed the web UI in any way?19:46
hazmatniemeyer, not atm. i'm planning to update the ui to mark items not in the charm store when looking at the official series19:53
hazmatniemeyer, it pre-dates the charm store, and there is no query on the charm store outside of querying the thing you want directly by name afaik19:54
niemeyerhazmat: Cool, no worries.. I'm happy to extend with stuff that makes things simpler on the other end19:55
niemeyerhazmat: I'm asking mainly due to stats19:55
niemeyerhazmat: Please don't add any automatic querying without syncing up19:55
niemeyerhazmat: In case you need/want to query one of the existing APIs, I'll hand you a parameter so it's not counted as a legitimate request19:56
hazmatniemeyer, cool, sounds good19:57
hazmatniemeyer, if you have a url to grab stats that would be nice to pull in19:57
niemeyerhazmat: I will have, many of those actually19:57
niemeyerhazmat: Per charm, etc19:58
=== jelmer_ is now known as jelmer
SpamapSjimbaker: https://code.launchpad.net/~jimbaker/juju/relation-hook-commands-spec/+merge/97733 .. you should land this in docs :)22:47
* SpamapS is trying to close out the florence milestone22:47
jimbakerSpamapS, sounds good22:53
SpamapSjimbaker: there are 4 open, Approved bugs left in florence, all for lp:juju/docs, all for you :)22:58
jimbakerSpamapS, yeah, i was planning on combining those together in a more cohesive whole, but it's probably best to do that as a second step22:59
SpamapSjimbaker: I'll push them off to the galapagos milestone if you're not ready to commit them just yet.23:06
jimbakerSpamapS, that probably makes sense23:30
SpamapSBlueprints:23:33
SpamapS1 Implemented23:33
SpamapSBugs:23:33
SpamapS108 Fix Released23:33
SpamapSKapil and William tie for most assigned at 27 :)23:34

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