orospakr | I just tried deploying a charm (using juju-core) that requires the onieric series, but the VMs failed in the "no matching tools" state. trying `juju sync-tools --all` just produced the same message. | 00:10 |
---|---|---|
orospakr | here's how the world looks for me right now: https://gist.github.com/orospakr/b3a1950fbe80c61446b4 | 00:13 |
davecheney | orospakr: we do no support onaric | 00:15 |
davecheney | sorry | 00:15 |
davecheney | you won't find any tools | 00:15 |
davecheney | yeesh | 00:16 |
davecheney | we do not support onieric | 00:16 |
orospakr | huh, okay. thank you. | 00:16 |
orospakr | silly question: when you juju bootstrap with juju-core, the bootstrap VM is running juju-core as well? | 00:17 |
davecheney | the first machine, machine 0 is running the mongodb service (what we call the state server) | 00:22 |
davecheney | it holds the knowledge of everything in that environment | 00:22 |
orospakr | ah, I knew about that. I thought a jujud ran on it as well. | 00:22 |
davecheney | orospakr: well a jujud is running on that machine | 00:23 |
davecheney | i'm not sure which question you are asking | 00:23 |
orospakr | does the first node also actively administer the other nodes, in response to your commands with the juju client, in addition to storing state in mongodb? | 00:26 |
davecheney | orospakr: yes and no | 00:31 |
davecheney | the first node also runs the providisioning agent which is the one that talks to the provider and creates new machines (nodes) if required | 00:31 |
orospakr | ah, that's the thing I'm thinking of. good. :) | 00:34 |
orospakr | so, now I've got this other issue: I can't destroy the wedged service/units/machines, because they're all unaccessible to juju. | 00:36 |
davecheney | orospakr: can you explain what you mean by inaccessible ? | 00:36 |
orospakr | the onieric-based service is in the "dying" state, and the units and machines alike are in the "pending" state. | 00:37 |
orospakr | s/"dying" state/life parameter is "dying"/ | 00:37 |
davecheney | orospakr: have you tried juju terminate-machine ? | 00:40 |
orospakr | aye. [~]$ juju destroy-machine 2 3 4 | 00:40 |
orospakr | error: no machines were destroyed: machine 2 has unit "couchbase/0" assigned; machine 3 has unit "couchbase/1" assigned; machine 4 has unit "couchbase/2" assigned | 00:40 |
orospakr | oof, excuse me. that didn't paste well. | 00:40 |
davecheney | i'd start with destroy-unit, destroy-server, etc | 00:42 |
davecheney | services | 00:42 |
davecheney | but at this point i'd just destroy the environment and start again if that is an option | 00:42 |
orospakr | it is, but I'm trying to stick it out on things like this so I can be confident that I can deal with failures when I do use this in production. ;) | 00:42 |
davecheney | i think destroy-unit and destory-service take a --force | 00:44 |
davecheney | but they will create dangling references | 00:44 |
davecheney | m_3: ping | 01:20 |
davecheney | why did the PA restart ? http://paste.ubuntu.com/5597298/ | 02:37 |
davecheney | m_3-droid: i'm a tit | 04:06 |
davecheney | the problem isn't the HP endpoint, it is the number of concurrent connections to the mongo server running on machine/0 | 04:07 |
davecheney | ├─mongod───828*[{mongod}] | 04:08 |
davecheney | ├─mongod───9*[{mongod}] | 04:08 |
davecheney | oops | 04:08 |
davecheney | https://bugs.launchpad.net/juju-core/+bug/1172110 | 04:08 |
davecheney | shit, we're using 3 mongo connections per machine | 05:05 |
jam | davecheney: because of the different agents? | 05:18 |
davecheney | i suspect so | 05:20 |
davecheney | it isn't exactly 3 x | 05:20 |
davecheney | 450 connections == 156 machines | 05:20 |
davecheney | the mgo drivers' peer probing isn't helping | 05:21 |
* davecheney really hates that feature | 05:21 | |
davecheney | jam: https://codereview.appspot.com/8931044/ | 05:28 |
davecheney | jam: Apr 24 05:30:51 juju-goscale2-machine-0 mongod.37017[9222]: Wed Apr 24 05:30:51 [conn24025] end connection 15.185.174.63:33866 (599 connections now open) | 05:31 |
davecheney | Apr 24 05:30:53 juju-goscale2-machine-0 mongod.37017[9222]: Wed Apr 24 05:30:53 [initandlisten] connection accepted from 15.185.173.39:34560 #24049 (600 connections now open) | 05:31 |
_mup_ | Bug #24049: KDE systemsettings doesn't have a file association component <kde-systemsettings (Ubuntu):Fix Released> <https://launchpad.net/bugs/24049> | 05:31 |
davecheney | ^ this is what you get from mongo | 05:31 |
davecheney | compared to how much we log | 05:31 |
davecheney | it's not a lot | 05:31 |
davecheney | #1 | 05:31 |
davecheney | #1234 | 05:31 |
davecheney | #12345 | 05:31 |
_mup_ | Bug #12345: isdn does not work, fritz avm (pnp?) <isdnutils (Ubuntu):Fix Released by doko> <https://launchpad.net/bugs/12345> | 05:31 |
davecheney | #9999 | 05:31 |
davecheney | #10000 | 05:31 |
_mup_ | Bug #10000: xserver-common: X crashed (signal 7) while scrolling in Mozilla <xorg (Ubuntu):Invalid by daniels> <xorg (Debian):Fix Released> <https://launchpad.net/bugs/10000> | 05:31 |
davecheney | bzzt | 05:47 |
davecheney | Apr 24 05:46:53 juju-goscale2-machine-0 mongod.37017[9222]: Wed Apr 24 05:46:53 [initandlisten] connection refused because too many open connections: 819 | 05:47 |
davecheney | Apr 24 05:46:53 juju-goscale2-machine-0 mongod.37017[9222]: Wed Apr 24 05:46:53 [initandlisten] connection accepted from 15.185.176.240:44746 #55756 (820 connections now open) | 05:47 |
_mup_ | Bug #55756: installer partman crash selecting swap partition <ubiquity (Ubuntu):Invalid> <https://launchpad.net/bugs/55756> | 05:47 |
davecheney | Apr 24 05:46:53 juju-goscale2-machine-0 mongod.37017[9222]: Wed Apr 24 05:46:53 [initandlisten] connection refused because too many open connections: 819 | 05:47 |
davecheney | Apr 24 05:46:53 juju-goscale2-machine-0 mongod.37017[9222]: Wed Apr 24 05:46:53 [initandlisten] connection accepted from 15.185.177.196:51276 #55757 (820 connections now open) | 05:47 |
_mup_ | Bug #55757: "Send a mail" to a contact list adds only the first contact to "To:" <deskbar-applet:Fix Released> <deskbar-applet (Ubuntu):Fix Released by desktop-bugs> <https://launchpad.net/bugs/55757> | 05:47 |
fwereade | morning everyone | 05:59 |
fwereade | how screwed up is everything today? | 05:59 |
=== ChanServ changed the topic of #juju-dev to: ttps://juju.ubuntu.com | On-call reviewer: fwereade | Bugs: 2 Critical, 64 High - https://bugs.launchpad.net/juju-core/ | ||
davecheney | fwereade: we're looking at a hard limit of ~300 machines or 800 agents (the smaller of the two) | 07:17 |
fwereade | davecheney, thanks, I just read your email -- definitely mongodb connections? | 07:17 |
davecheney | fwereade: absolutelu | 07:17 |
davecheney | it says so in the mongo log | 07:17 |
davecheney | fwereade: also, https://codereview.appspot.com/8931044/ | 07:18 |
davecheney | not critical | 07:18 |
fwereade | davecheney, that is ludicrous, isn't it? surely that's not as far as it can go? | 07:19 |
davecheney | the default limit is 80% of nfiles, which is 1024 on ubuntu | 07:19 |
davecheney | these defaults can be changed | 07:19 |
davecheney | during service deployment we average 3x the number connections as machines | 07:20 |
davecheney | the number falls back to 2x after deployment | 07:20 |
davecheney | (although that may not be successful deployment as I restarted monogo | 07:20 |
davecheney | 2x sounds logical, 1 machine agent, 1 unit agent per service unit | 07:20 |
fwereade | davecheney, yep, agreed | 07:20 |
davecheney | the conn limit for mongo can be increased | 07:20 |
davecheney | fwereade: the other problem is i'm seeing continual cpu spikes as 600 agents wake up | 07:21 |
davecheney | and probe each mongodb server | 07:21 |
davecheney | this is a very unhelpful feature | 07:21 |
fwereade | davecheney, heh, I had been less worried about that in particular but it goes to show you never can tell:/ | 07:22 |
davecheney | at 2000 agents I estimate the mongodb server will be under constant polling pressure | 07:23 |
fwereade | davecheney, *if* we had an internal API that would be a non-issue but it doesn't help us today | 07:23 |
davecheney | which will probably fuck us with sockets in TIME_WAIT | 07:23 |
davecheney | fwereade: yes, that is the logical solution | 07:23 |
davecheney | federated mongodb might help | 07:23 |
davecheney | but the polling logic will still be a massive burdon | 07:23 |
fwereade | davecheney, yeah,doesn't feel like it'd really address the issue | 07:24 |
rogpeppe | mornin' all | 08:26 |
dimitern | rogpeppe: morning | 08:28 |
rogpeppe | dimitern: hiya | 08:28 |
rogpeppe | fwereade: hope you've recovered a bit! | 08:29 |
fwereade | rogpeppe, yeah, more or less, although I remain baffled at the actual state of play -- I am pondering dave's scale tests for now and hoping that mramm will wake up in an hour or two and bring us the gift of clarity | 08:30 |
rogpeppe | fwereade: unfortunately my network connection is still borked, so i missed the kanban meeting yesterday, so i don't really know | 08:31 |
rogpeppe | fwereade: i don't think i've seen anything about dave's scale tests other than the conversation just above | 08:34 |
dimitern | rogpeppe: I think nobody knows for sure, except probably mgz | 08:34 |
fwereade | rogpeppe, I had a brief chat with him this morning, it seems we are hitting mongo connection limits at around 300 machines | 08:35 |
rogpeppe | fwereade: interesting. i'm not entirely surprised actually. | 08:35 |
rogpeppe | fwereade: need to get that internal API done. and HA on that. | 08:36 |
fwereade | rogpeppe, yep | 08:36 |
fwereade | rogpeppe, considering possible mitigations in the meantime | 08:37 |
rogpeppe | fwereade: anything particular in mind? | 08:37 |
fwereade | rogpeppe, still going through internal sanity-vetting ;p | 08:37 |
dimitern | hmm.. cmd/juju tests running time seems to have improved slightly | 08:40 |
fwereade | mramm, heyhey | 08:47 |
mramm | heyhey | 08:49 |
mramm | I am trying to get caught up on the packaging issues | 08:50 |
fwereade | mramm, great | 08:50 |
mramm | I thought things were moving smoothly until late yesterday when antonio informed me that the server guys did not think it would go in | 08:50 |
mramm | I still don't have a clear picture from them of what needs to happen, just some hand waving about "issues" | 08:51 |
fwereade | mramm, my understanding had been that monday was the actual razor-sharp cutoff in any case | 08:52 |
mramm | well, that was my understanding too | 08:52 |
mramm | but I thought we'd given them something by then, and I hear from them yesterday that the razer sharp cutoff is actually today | 08:53 |
fwereade | mramm, but, regardless, if there is anything I can do to help I would be happy to; and if you do manage to glean some measure of clarity I, and others, would be most grateful for it | 08:53 |
mramm | looks like jamespage is on the case | 08:54 |
mramm | if you can join #server on canonical IRC that would be helpful | 08:55 |
mramm | so anybody that can help jamespage in some way today gets an extra gold star from me ;) | 09:13 |
davecheney | good evening | 09:25 |
davecheney | could I draw your attention to | 09:25 |
davecheney | https://bugs.launchpad.net/juju-core/+bug/1172122 | 09:25 |
_mup_ | Bug #1172122: state/presence: FAIL: presence_test.go:253: PresenceSuite.TestWatchPeriod <juju-core:New> <https://launchpad.net/bugs/1172122> | 09:25 |
davecheney | and | 09:25 |
mramm | so, on subject of dave's e-mail | 09:25 |
davecheney | https://bugs.launchpad.net/juju-core/+bug/1172121 | 09:25 |
_mup_ | Bug #1172121: environs/maas: multiple test failures <juju-core:Confirmed> <https://launchpad.net/bugs/1172121> | 09:25 |
mramm | haha | 09:25 |
davecheney | mramm: i can leave again if you want to talk about me | 09:26 |
davecheney | i don't mind, i like the idea of being popular | 09:26 |
davecheney | also, https://docs.google.com/a/canonical.com/document/d/1p_OzWxqxaXalHBI3ohkUsB9_iQBPSGWwM-qODm5FSbI/edit# | 09:26 |
mramm | haha | 09:27 |
mramm | so, I think we should talk about the internal API stuff sooner rather than later | 09:27 |
mramm | given that we are hitting scalability limits now | 09:27 |
fwereade | davecheney, did you update gomaasapi? | 09:29 |
davecheney | mramm: SGTM | 09:29 |
davecheney | fwereade: probably not | 09:29 |
fwereade | davecheney, I think that resolves it | 09:29 |
davecheney | right, i'll check that tomorrow | 09:29 |
fwereade | davecheney, I'll close it then, reopen if I'm wrong please | 09:30 |
davecheney | SGTM | 09:32 |
davecheney | what about state/presence ? | 09:32 |
fwereade | davecheney, I think I have no option but to do a deep dive there and try to figure out WTF is going on | 09:33 |
fwereade | davecheney, occasional failures from there have always been a thing, but not one that ever quite rose high enough to be looked at properly | 09:33 |
davecheney | has anyone else seen that problem ? | 09:33 |
fwereade | davecheney, it doesn't look entirely unfamiliar, but I thought we had a bug for it already; apparently not | 09:35 |
mramm | so, can somebody help jamespage out with updating our "release" package to build from local source rather than latest trunk, so the source gets included in the package | 10:10 |
mramm | we can pull the local source from trunk and then build | 10:10 |
jamespage | mramm, forget the package | 10:10 |
jamespage | I just want a release tarball of juju-core 1.10.0 that contains everything that is juju-core 1.10.0 aside from the packaging | 10:10 |
mramm | ahh | 10:11 |
mramm | cool | 10:11 |
mramm | so rogpeppe, TheMue, fwereade, jam, dimitern: can one of you get that for jamespage now? | 10:12 |
rogpeppe | mramm: ok, i'll do that | 10:12 |
mramm | rogpeppe: thanks! | 10:13 |
rogpeppe | jamespage: presumably not including binaries, right? | 10:13 |
jamespage | rogpeppe, you got it | 10:14 |
TheMue | ah, just wanted to ask too | 10:14 |
davecheney | if you fancy a break to look at it | 10:18 |
mramm | morning mgz! | 10:21 |
mgz | hey! seen your email. | 10:23 |
jamespage | mgz, hey! | 10:24 |
mgz | so, there are a couple of issues for me: | 10:26 |
mgz | the update-alternatives bits didn't get merged into dave's packaging branch, but he did then add a manpage, so that needed updating | 10:27 |
mgz | I don't have rights to upload to the public bucket on ec2, nor have I done what dave normally does for releases | 10:28 |
mgz | and there's a change in 1.10 working around the updates-during-raring to upstart bug that I don't think we want to release with | 10:29 |
jamespage | mgz, I've merged all the various bits of packaging - lp:~james-page/ubuntu/raring/juju-core/1.10 | 10:30 |
jam | mgz: I have some questions for you as well if/when you have some time. | 10:30 |
mgz | jam: sure | 10:30 |
jamespage | mgz, and raised a FFe - 1172215 | 10:30 |
jamespage | bug 1172215 | 10:30 |
_mup_ | Bug #1172215: [FFe] Please include juju-core 1.10.0 in Ubuntu 13.04 <Ubuntu:New> <https://launchpad.net/bugs/1172215> | 10:30 |
jam | mgz: 1) it is known that 1.10.0 is in the ppa but not uploaded to ec2, so you can't actually bootstrap (no tools found) | 10:31 |
mgz | those are the only things I'm aware of. | 10:31 |
mramm | davecheney: can you help out with the tools upload stuff if you are still around? | 10:32 |
jam | 2) I still can't ssh to the original goose instance, as near as I can tell chinstrap isn't letting ssh get to lcy01 (it is working to lcy02, but I get No route to host trying to get to 10.55.60.94) | 10:32 |
mgz | any of the core guys should have rights, but no one responded when I poked the other day | 10:32 |
mgz | so maybe they don't? | 10:33 |
davecheney | mramm: sure, mgz do you want the creds ? | 10:33 |
jam | 3) I tried starting a new tarmac bot with similar config and juju-core doing the bootstrap to lcy02, but it seems charms:precise/tarmac uses puppet, and that just doesn't work now. | 10:33 |
davecheney | or I can do the push if you cna point me to the deb in the archive | 10:33 |
mramm | giving mgz creds makes sense to me | 10:33 |
mramm | lp:~james-page/ubuntu/raring/juju-core/1.10 looks like the proposed release | 10:34 |
davecheney | sure, the only reason I hesitate is they belong to gustavo | 10:34 |
mramm | yea | 10:34 |
davecheney | so, be careful with 'em | 10:34 |
jam | davecheney: load testing time! | 10:34 |
jam | how many CC4.xxlarge can we run? :) | 10:34 |
mramm | we ultamately need to take that over and make it something owned by a team, not a person :/ | 10:34 |
jam | mramm: is it possible to change ownership of an s3 bucket? | 10:35 |
mramm | not easily | 10:36 |
davecheney | mgz: ceheck your make | 10:36 |
davecheney | mramm: jam much easier to change the source in the code | 10:37 |
davecheney | mgz: check your mail | 10:37 |
mramm | davecheney: right | 10:37 |
mramm | you can delete it, and then try to pick the name up under a new account | 10:37 |
mramm | but 1) anybody can pick it up when it becomes available | 10:38 |
mramm | and 2) it can take up to 24 hours to go back into the available names pool | 10:38 |
mgz | jam: one funny thing I noticed was there are a couple of goosebot instances, one of which is shutoff, on lcy01. it's possible the routing is broken just for that host. | 10:38 |
mgz | jam: the tarmac puppet charm not working with juju-core is a bug here I guess, unless it does something very bogus | 10:39 |
jam | mgz: the shutoff one is the python-juju bootstrap node | 10:39 |
jam | I wasn't able to start and get to that one either, but didn't really need to. | 10:40 |
mgz | there are two shutoff ones... | 10:40 |
TheMue | mramm: thats why I suggest a dns name like tools.juju.ubuntu.com in our code to point to ANY bucket we want (or other server) | 10:42 |
rogpeppe | jamespage: i'm pushing a source-only branch containing all the latest source and its dependencies. | 10:43 |
mramm | TheMue: file a kanban card for it and add it to tomorrow's agenda ;) | 10:43 |
davecheney | TheMue: https://docs.google.com/a/canonical.com/document/d/1p_OzWxqxaXalHBI3ohkUsB9_iQBPSGWwM-qODm5FSbI/edit# | 10:44 |
TheMue | mramm: yep, will do | 10:44 |
rogpeppe | jamespage: unfortunately my network upload speed is outrageously slow at the moment (about 50Kbits/s) so it will take a while | 10:44 |
TheMue | davecheney: agenda is already edited, for a different topic ;) | 10:44 |
jamespage | rogpeppe, ack | 10:44 |
mgz | jam: seems 10.55.60.94 is indeed unreachable, worth raising with canonistack support | 10:45 |
jam | mgz: is that ask web-ops and then get it escalated ? | 10:47 |
jam | mgz: do you know if you have to have the puppet agent running to have the script run ? (default is that puppet agent in /etc/defaults/puppet is to not run) | 10:47 |
mgz | for the record, the packaging branch is at lp:~james-page/ubuntu/raring/juju-core/1.10.0 with the trailing .0 | 10:47 |
rogpeppe | jamespage: lp:~rogpeppe/+junk/juju-1.10.0-source-only | 10:51 |
mgz | jamespage: your packaging branch looks good to me | 10:51 |
jamespage | rogpeppe, that really need to be somewhere official | 10:51 |
* TheMue is at lunch | 10:51 | |
jamespage | like lp:juju-core/1.10.0 | 10:51 |
rogpeppe | jamespage: ah | 10:51 |
rogpeppe | i'm worried i might start stepping on someone's toes here - i'm not usually involved with this stuff | 10:52 |
mgz | needs a different name if you're putting it under the juju-core project | 10:52 |
mgz | as it's packaging | 10:52 |
mgz | I can push it somewhere though | 10:52 |
jamespage | mgz, no - thats the point | 10:53 |
jamespage | I want an upstream release of juju-core | 10:53 |
jamespage | forget the packaging | 10:53 |
mgz | it's not juju-core 1.10.0 | 10:53 |
mgz | it's that plus all the deps | 10:53 |
jamespage | yes | 10:53 |
mgz | okay. | 10:53 |
jamespage | juju-core plus the deps that the juju-core dev team say are good for 1.10.0 | 10:53 |
rogpeppe | basically i fetched all the deps from scratch and removed the .bzr and .hg directories | 10:53 |
jamespage | mgz, I want to switch the packaging away from 3.0 (native) - its not required | 10:53 |
mgz | to quilt? | 10:54 |
jamespage | mgz, yes | 10:54 |
jamespage | native rarely makes sense | 10:54 |
mgz | I can live with that | 10:55 |
jam | mgz: I wonder if you could grab them and use 'bzr-upload' to create light dirs so that it is easy to update them to newer tools. | 10:55 |
mgz | so, issue #1 is resolved by james' branch, issue #2 is okay now I have the ec2 creds, last question is if we carry the cloud-init hack or not... | 10:56 |
mgz | jam: we could certainly do something more elegant, for now I'm happy with just dumping the code and adding a bunch of fresh unrelated stuff to the repo on launchpad | 10:58 |
mramm | be back in a few -- getting breakfast and etc. Ping me if needed. | 10:59 |
mgz | jamespage: probably a question for you as much as any of the juju guys, see <https://codereview.appspot.com/8648047/> | 11:00 |
jam | mgz: did you still want the hp-cloud instance runnng? | 11:01 |
mgz | jam: nope, I fixed my script so I can restart it myself as needed | 11:02 |
jam | mgz:except you couldn't reach http, right? | 11:03 |
jam | non-http because of the non-standard port | 11:03 |
jam | for keyauth | 11:03 |
mgz | well, "myself", provided I remember to do it in advance, otherwise without manual intervention (lesson: use sed -r when being fancy) | 11:03 |
rogpeppe | my pesky phone line seems like it might be out for another whole week | 11:18 |
* rogpeppe didn't know about sed -r | 11:20 | |
rogpeppe | i always get bitten by the fact that standard sed doesn't do "proper" regexps, 'cos i'm used to plan 9's sed which does them by default | 11:20 |
jamespage | mgz, rogpeppe: which of you two are working on preparing the juju-core 1.10.0 snapshot release and sticking it something official? | 11:21 |
jamespage | not clear from my backscroll | 11:21 |
rogpeppe | mgz: given my (lack of) current bandwidth, you might be best doing that | 11:22 |
jam | rogpeppe: still no home internet? ouch | 11:22 |
rogpeppe | jam: yeah. i just talked with the phone company, and they have no idea when it'll be resolved. it's been out since last tues | 11:23 |
jam | rogpeppe: for sed, you mean '\d' vs [[:digit:]] ? | 11:23 |
rogpeppe | jam: well, there is *some* internet, but the upload speed is stupidly bad, and i've been getting 3-5s ping response times | 11:23 |
rogpeppe | jam: no, i mean (foo|bar) | 11:24 |
mgz | jamespage: I can push the source roger put up to somewhere... and I guess we can just leave the other questions for now | 11:28 |
mgz | lp:~juju-core/ubuntu/raring/juju-core/1.10.0-source okay? | 11:30 |
jam | mgz, dimitern: standup? | 11:31 |
jamespage | mgz, not sure why we need the ubuntu/raring prefix | 11:31 |
jamespage | its not ubuntu or raring - its juju-core 1.10.0 | 11:31 |
jam | jamespage: because ~juju-core/juju-core is completely unrelated code. | 11:31 |
jam | well, "unrelated" | 11:31 |
jam | in that it is just the base tree, not all deps | 11:31 |
jamespage | well like I said - I'm good with a tarball published on launchpad.net | 11:32 |
jamespage | like we do for juju | 11:32 |
mgz | hm, that would be ideal really | 11:33 |
jamespage | i'd rather have it that was - otherwise I have to cut my own tarball still | 11:33 |
jamespage | which is not always deterministic | 11:33 |
jam | mgz: mumble? | 11:33 |
mgz | I'm there. | 11:34 |
mgz | so, we don't have a 1.10.0 milestone yet? ...ah, the fun is it's half the 2.0 milestone | 11:41 |
wallyworld_ | fwereade: want to join us on mumble? | 11:52 |
mgz | jamespage: https://launchpad.net/juju-core/1.10/1.10.0 has tarball based on the rogpeppe branch | 12:02 |
fwereade | wallyworld_, actually, could we have a quick hangout for 15 mins? | 12:02 |
mgz | I'll tidy up the rest of the release sutff for now | 12:02 |
wallyworld_ | sure | 12:02 |
fwereade | wallyworld_, I forgot our differing interpretations of wednesday | 12:02 |
mgz | blast, no hangout for me | 12:02 |
jamespage | mgz, ta muchly | 12:02 |
wallyworld_ | fwereade: if mgz can't do hangout, can you do mumble? | 12:03 |
fwereade | wallyworld_, er maybe, how can I set it up in 2 minutes or less? | 12:03 |
rogpeppe | mgz: you might want to include this info somewhere around the place: http://paste.ubuntu.com/5598150/ | 12:03 |
wallyworld_ | fwereade: not in 2 mins i don't think :-( | 12:03 |
rogpeppe | mgz: it's the full list of the revision numbers | 12:04 |
wallyworld_ | mgz: can you reduce your bandwidth for a hangout? | 12:04 |
mgz | fwereade: `sudo apt-get install mumble` then <https://wiki.canonical.com/StayingInTouch/Voice/Mumble> but hangout safer, go for that | 12:04 |
fwereade | mgz, cheers, I'll set that up after this then | 12:04 |
mgz | wallyworld_: issue is google don't provide arm debs for their binary blobs | 12:04 |
wallyworld_ | mumble can be fiddly | 12:04 |
rogpeppe | mgz: hmm, except the revision of the go tree itself. that should probably be included too, i suppose, although we're not including it in the tarball | 12:04 |
rogpeppe | mgz: maybe we should | 12:05 |
rogpeppe | mgz: although i suppose we're building against a known go revision | 12:05 |
mgz | I can, if needed, cheat and re-up the tarball with fixes | 12:05 |
fwereade | wallyworld_, mgz: anyone else I should invite? jam? | 12:05 |
mgz | I'm not certain the current dir layout makes sense for instance | 12:05 |
wallyworld_ | fwereade: he is afk for a little bit, will join when he gets back | 12:05 |
mgz | fwereade: jam was interested, but has to go off now | 12:05 |
wallyworld_ | so invite him | 12:06 |
rogpeppe | mgz: i just did it so the root of the tree could be used as $GOPATH | 12:06 |
fwereade | mgz, I need to in 10 mins bit I think we can cover some stuff usefully | 12:06 |
mgz | yeah, go go, you and ian | 12:06 |
* dimitern bbin1h | 12:06 | |
jamespage | mramm, mgz: I noticed the debian/copyright file was not complete in the packaging - working on that now | 12:36 |
jamespage | which means I have to document the copyright and license for all of the bundled projects as well | 12:36 |
* jamespage sighs | 12:36 | |
mgz | ...sorry about that | 12:40 |
jamespage | mramm, mgz: OK - I have two problems | 12:41 |
jamespage | goose has no explicit Copyright holder | 12:41 |
jamespage | and lpad has neither a Copyright holder or license | 12:41 |
mgz | fwereade: sorry for the incoming launchpad email surge, you had a bunch of fixed bug targetted at 2.0 that you actually fixed long ago | 12:45 |
mgz | ...is lpad actually a dependency? | 12:45 |
mgz | or should we remove it from the tarball? | 12:45 |
mgz | goose is fixable | 12:46 |
mgz | we've had no external contributions, the copyright holder is just canonical | 12:46 |
mgz | jamespage: can you try building what you have as well, to see if tests etc are all fine? I'll upload a -1 tarball with any fixes needed | 12:47 |
jamespage | mgz, builds just fine - this is just a distro copyright/license thing | 12:48 |
jamespage | mgz, I don't see any tests executing fwiw | 12:48 |
mgz | jamespage: where should we put the copyright holder if we're not doing per-file licencing? | 12:49 |
mgz | jamespage: I doubt they're run as part of the packaging | 12:49 |
mgz | we should probably add that, but not now | 12:50 |
jamespage | mgz, good question - most of niemeyer's projects have it in LICENSE | 12:50 |
niemeyer | mgz, jamespage: I tend to do it per file as well | 12:51 |
niemeyer | mgz, jamespage: and certainly on a LICENSE file or similar | 12:51 |
jamespage | niemeyer, I spotted | 12:51 |
jamespage | :-) | 12:51 |
mgz | in juju-core that seems to just be a copy of the agpl | 12:51 |
jamespage | mramm, mgz: LICENSE/LICENCE or suchlike is sufficient - every file is best practice | 12:51 |
niemeyer | jamespage: +1 | 12:52 |
niemeyer | Even because some projects (e.g. goyaml) do have mixed licensing | 12:52 |
* jamespage looks at goyaml again | 13:06 | |
mgz | hm, and now canonistack is refusing to talk to me | 13:07 |
mgz | okay, I can't build or upload anything till canonistack is back, so having lunch | 13:20 |
=== wedgwood_away is now known as wedgwood | ||
mgz | jamespage: okay, have written a hacky script for rolling up tarball | 14:51 |
jamespage | mgz, yeah for hacky scripts ;-) | 14:51 |
mgz | I just need to know, exactly, what change you want me to make for the goose license thing | 14:51 |
jamespage | it was missing copyright - hrm | 14:51 |
mgz | I currently have COPYING and COPYING.LESSER in the branch | 14:51 |
mgz | I don't really want to modify those, as they're just the text from gnu | 14:52 |
mgz | not that I know why we have both... | 14:52 |
mgz | modifying every darn file in the tree is also not sane, though it's what the gpl generally wants (header on each source file) | 14:53 |
jamespage | mgz, indeed - I think just adding a LICENSE file detailing which of those two licenses its licensed under and details of the C holder would be OK | 14:53 |
jamespage | for now anyway | 14:53 |
jamespage | infact I'm happy to leave what you have in tarball as-is - so long as there is a commit in the bzr branch with the details on it - I can refer to that with a comment | 14:53 |
jamespage | time is of the essence and all that | 14:54 |
mgz | okay, done, uploading | 15:06 |
mgz | this really does need the tests run on it though, as it's pristine from export, rather than copied files from trees that have been tested together | 15:10 |
mgz | jamespage: (and everybody else) https://launchpad.net/juju-core/1.10/1.10.0/+download/juju-core_1.10.0-1.tar.gz | 15:11 |
jamespage | mgz, wanna check my branch again? just re-cut using that tarball | 15:22 |
rogpeppe1 | mgz: i ran the basic tests on it, but not live tests. | 15:29 |
rogpeppe1 | mgz: BTW, i'm pretty sure that davecheney builds from pristine for each release | 15:29 |
mgz | jamespage: looking, it seems good, one question for the others | 15:36 |
mgz | go.net has lost the html/ package, is that something we used in any way? | 15:36 |
m_3 | TheMue: I'm gonna turn that into a blog post though.. so give it a couple of days and I'll have a version that's a bit easier to read | 15:41 |
mgz | seem safe, I needed exp/html from go trunk long ago, but I think that was for rietveld not juju-core anyway | 15:42 |
TheMue | m_3: Great, that's what we need. Thx for your effort. | 15:48 |
mgz | jamespage: seems there were some changes to the cert code in your old cut of the source which I'm not clear on the origin of... | 15:49 |
mgz | what's there now is all that's been on trunk as far as I can see, and looks okay to me | 15:50 |
mgz | anyway, I shall build out of that branch, and upload to public bucket | 15:51 |
mgz | I guess we may also want to change the recipe to use this, and rebuild what's in the ppa? | 15:52 |
rogpeppe1 | simple git question for someone: what's the equivalent of bzr revision-info in git? | 15:57 |
mgz | hm, something is not happy | 15:59 |
TheMue | rogpeppe1: took a look at git show? | 16:06 |
rogpeppe1 | TheMue: i think that "git rev-parse HEAD" is what i need | 16:06 |
TheMue | rogpeppe1: oh, i'll take a look | 16:07 |
rogpeppe1 | TheMue: it seems there's no linear idea of commit history in git, unlike hg and bzr. is that right? i.e. no numeric log numbering. | 16:07 |
rogpeppe1 | TheMue: thanks. i've never used git in anger. | 16:07 |
TheMue | rogpeppe1: i've just started for private projects, used hg before | 16:09 |
TheMue | rogpeppe1: and yes, it seems to use large numbers like uuids and commit, tree, parent relations | 16:11 |
jamespage | mgz, OK- just got a ftbfs on arm | 16:11 |
rogpeppe1 | jamespage: interesting. what's the error? | 16:11 |
jamespage | apologies for the paste: | 16:12 |
jamespage | launchpad.net/goyaml | 16:12 |
jamespage | # launchpad.net/goyaml | 16:12 |
jamespage | src/launchpad.net/goyaml/goyaml.go:89: undefined: newDecoder | 16:12 |
jamespage | src/launchpad.net/goyaml/goyaml.go:90: undefined: newParser | 16:12 |
jamespage | src/launchpad.net/goyaml/goyaml.go:135: undefined: newEncoder | 16:12 |
rogpeppe1 | jamespage: ah, i think i know what the issue might be. i wonder if cgo is disabled/not working on arm | 16:12 |
jamespage | rogpeppe1, how do I check? | 16:13 |
rogpeppe1 | jamespage: are you getting the error at a command prompt? | 16:13 |
rogpeppe1 | jamespage: oh jeeze | 16:13 |
rogpeppe1 | jamespage: this is an old version of go we're using | 16:13 |
rogpeppe1 | jamespage: i'm not surprised actually | 16:14 |
jamespage | 1.0.2 as in raring | 16:14 |
rogpeppe1 | jamespage: yeah. hmm. | 16:14 |
rogpeppe1 | jamespage: davecheney's the man for knowing about go-on-arm stuff. | 16:14 |
rogpeppe1 | jamespage: if you type "go env" on the arm box, it should have an output line saying something like: CGOENABLED="1" | 16:15 |
rogpeppe1 | sorry, CGO_ENABLED | 16:16 |
jamespage | rogpeppe1, CGO_ENABLED="0" | 16:16 |
jamespage | yikes | 16:16 |
rogpeppe1 | jamespage: right, so that's the issue | 16:16 |
jamespage | rogpeppe1, OK - we'll drop arm for juju-core in raring | 16:16 |
rogpeppe1 | jamespage: +1 | 16:16 |
rogpeppe1 | jamespage: we've done no testing on arm | 16:16 |
jamespage | mramm, ^^ | 16:16 |
mramm | jamespage: no arm is fine | 16:17 |
mramm | jamespage: arm support was explicitly pushed back to post 13.04 anyway | 16:17 |
mgz | well, nearly very smooth: <http://paste.ubuntu.com/5598730/> | 16:18 |
jamespage | mramm, OK | 16:18 |
rogpeppe1 | jamespage: from http://code.google.com/p/go-wiki/wiki/GoArm : "currently the development version of Go includes better support for linux/arm, including full cgo support, than Go 1.0." | 16:18 |
mramm | rogpeppe1: correct, git has no version numbers, just commits (identified by the hash of the commit contents) | 16:19 |
rogpeppe1 | mramm: thanks | 16:19 |
mramm | commits define their parent commit, and you therefore get a graph of commits | 16:19 |
=== _mup__ is now known as _mup_ | ||
mramm | which git walks to show you a timeline | 16:20 |
mgz | and yeah, I can't build juju-core on this box, we don't work on arm currently, but we were also not targetting it | 16:20 |
rogpeppe1 | mramm: that's true of all of those systems i think - but the linear history is quite nice for referring to a given trunk | 16:20 |
rogpeppe1 | mgz: you probably could if you used go tip | 16:20 |
mramm | rogpeppe1: well, git does not attempt to pretend that there is a linear history | 16:21 |
rogpeppe1 | mgz: i'd be interested to find out whether it works actually | 16:21 |
rogpeppe1 | mramm: yeah. | 16:21 |
mgz | I'll add that to my list of fun-time things to try :) | 16:21 |
rogpeppe1 | hmm, i wonder if i my san box upstairs would be up to the task... | 16:21 |
mramm | in a fully distributed system I have commits in a branch you don't have and vice versa, so linear history is impossible to get right | 16:21 |
rogpeppe1 | mramm: yeah, but hg and bzr both pretend quite well :-) | 16:22 |
mramm | linus was very opposed to pretending | 16:25 |
rogpeppe1 | mramm: another stupid git question: how do i update the current working tree to a given rev id? | 16:26 |
rogpeppe1 | mramm: git pull? | 16:27 |
fwereade | whoops, not been focusing on this: https://codereview.appspot.com/8939043 | 16:27 |
fwereade | rogpeppe1, quick look for form's sake please? | 16:27 |
mramm | git checkout <sha> | 16:27 |
rogpeppe1 | fwereade: LGTM trivial | 16:27 |
fwereade | rogpeppe1, cheers | 16:28 |
rogpeppe1 | mramm: thanks | 16:28 |
mramm | git pull will grab objects (commits and trees) from the remote repo an pull them down | 16:28 |
mramm | git checkout will switch the cwd | 16:28 |
fwereade | ok, I'm nearly up to date on my reviews, and everyone can start committing their approved bits and pieces to trunk now | 16:29 |
fwereade | will try to swing by again later -- if not, ttyall tomorrow | 16:29 |
fwereade | dimitern, if C+L sleep early I might ping you for a late beer/catchup re upgrade-charm | 16:30 |
fwereade | dimitern, otherwise, maybe 20 mins before the meeting tomorrow? | 16:30 |
dimitern | fwereade: sgtm | 16:32 |
dimitern | fwereade: when you can | 16:32 |
jamespage | mgz, rogpeppe1, fwereade: you guys happy with what we are proposing for release into raring? | 16:40 |
rogpeppe1 | jamespage: yup, sgtm | 16:40 |
rogpeppe1 | jamespage: assuming it's essentially still the sources i put together | 16:40 |
jamespage | rogpeppe1, its the re-cut sources mgz did | 16:40 |
rogpeppe1 | mgz: what did you change? | 16:41 |
jamespage | lpad got dropped | 16:41 |
jamespage | and some licensing clarification around goose | 16:41 |
dimitern | why is juju depending on lpad anyway? it's only used by lbox when interacting with LP | 16:42 |
* jamespage shrugs | 16:43 | |
jamespage | anyway I just uploaded to raring - Daviey and slangasek are lined up for review | 16:43 |
dimitern | jamespage: tyvm! | 16:44 |
mramm | jamespage: awesome work! | 16:47 |
jamespage | lets not get to excited - its just in the queue! | 16:48 |
mramm | thank you very, very much | 16:48 |
mramm | haha | 16:48 |
mramm | understood | 16:48 |
mramm | but it is progress | 16:48 |
jamespage | mramm, action prior to next release - get some copyright/license headers in all source across the board please! | 16:48 |
Makyo | fwereade, I got your review comments in the middle of submitting. Would you be alright with a separate branch with those implemented? | 16:48 |
mramm | jamespage: will do | 16:49 |
Makyo | Er, too late :/ | 16:49 |
rogpeppe1 | dimitern: it's actually just the store code that depends on lpad | 16:51 |
rogpeppe1 | dimitern: i included it just so i could do go test ./... without errors | 16:52 |
dimitern | rogpeppe1: it's about time to separate the store from juju-core now | 16:52 |
rogpeppe1 | dimitern: it's happened before and it will happen again :-) | 16:52 |
dimitern | rogpeppe1: i certainly hope it'll be soon :) | 16:53 |
dimitern | Makyo: it's never late for another branch ;) | 16:54 |
Makyo | dimitern, More branches, more branches! | 16:54 |
=== deryck is now known as deryck[afk] | ||
Makyo | Getting conflicting reviews on r1192 and it should be reverted. I've never been successful at that. Can someone help me out? | 17:28 |
fwereade | Makyo, sure, that's fine | 17:46 |
fwereade | Makyo, sorry about that | 17:47 |
fwereade | Makyo, no need to revert if it's already in, just note that you will update in the review please | 17:47 |
mattyw | what's the best way for a charm to work out if it's been deployed using py-juju or go-juju other than checking for location of the agent.conf files? | 17:48 |
Makyo | fwereade, even with rogpeppe1' | 17:49 |
Makyo | 's comments? | 17:49 |
fwereade | Makyo, ah sorry, just saw rog's | 17:49 |
fwereade | Makyo, 2 mins thinking time | 17:49 |
rogpeppe1 | mattyw: JUJU_CONTEXT_ID is a reasonable indication, i think. | 17:50 |
rogpeppe1 | Makyo: sorry for tardy review - i forgot to submit the comments earlier | 17:50 |
Makyo | rogpeppe1, That's okay, they're definitely necessary. I think reverting might be the best choice, though, to make sure things aren't half-right. | 17:51 |
fwereade_ | Makyo, yeah, it needs more thought -- sorry, Ineeded to look up what we did with deploy in that case | 17:53 |
fwereade_ | Makyo, I think we should keep charm-adding and charm-setting spearate, essentially as they are in deploy | 17:55 |
Makyo | fwereade_, alright, looking through deploy... | 17:58 |
fwereade_ | Makyo, trying to figure it out myself | 17:59 |
fwereade_ | Makyo, rogpeppe1: how do charms get into state from the GUI ie via the API? | 17:59 |
rogpeppe1 | fwereade_: currently they don't | 17:59 |
rogpeppe1 | fwereade_: we're restricted to charms in the charm store | 18:00 |
mgz | jamespage: what was the reason for juju-core rejection? ...I wish there was something recorded on launchpad | 18:00 |
Makyo | fwereade_, rogpeppe1, Correct, no support deploy/upgrade on local yet, though we can see them once they're deployed. | 18:00 |
fwereade_ | rogpeppe1, ah ok, so the client just has to pass a url known to the charms store? | 18:00 |
rogpeppe1 | fwereade_: yes | 18:00 |
rogpeppe1 | fwereade_: my vague plan is to have an optional extra call to upload a charm (probably a sequence of calls so we don't bundle up MB in one json message) | 18:01 |
rogpeppe1 | fwereade_: it may even end up being better as a PUT | 18:01 |
fwereade_ | Makyo, rogpeppe1: ok, then are we ok just putting the same restriction on upgrades for now? ie ServiceName, CharmURL, Force? | 18:01 |
Makyo | fwereade_, rogpeppe1, sounds good to me. | 18:02 |
rogpeppe1 | fwereade_: i think so | 18:02 |
fwereade_ | Makyo, rogpeppe1: I'm fine punting on local charms today at least, so long as we punt consistently | 18:02 |
rogpeppe1 | fwereade_: yup | 18:02 |
rogpeppe1 | i'm done for the day | 18:03 |
Makyo | Still on to revert 1192? | 18:03 |
rogpeppe1 | see y'all tomorrow | 18:03 |
Makyo | Later. | 18:03 |
gary_poster | Hi all. I'd like to announce the GUI's compatibility with juju core, but the Raring Juju from the devel PPA fails for me like this: http://pastebin.ubuntu.com/5598967/ | 18:04 |
gary_poster | Is that known? Is there some other, better way to suggest that people try out the GUI on juju core? I didn't figure installing juju from source was the right sales pitch :-) | 18:04 |
jamespage | mgz, release team said it was to late | 18:19 |
ahasenack | gary_poster: so now --upload-tools is failing too? | 18:29 |
gary_poster | ahasenack, as in the pastebin, bootstrap --upload-tools succeeds temporarily, but then I can't deploy anything. :-/ | 18:30 |
ahasenack | oh, ok, I missed the deploy command | 18:30 |
mramm | jamespage: mgz: we are still fighting the good fight. And if we don't get it in now, we will do everything possible to get it in via backports early next week | 18:47 |
mramm | so the end user visible difference will not be much | 18:47 |
mramm | either way after next week a user will be able to sudo apt-get install juju-core, and get our package | 18:52 |
mramm | and if it is in backports it will be even easier to get it updated with our monthly releases | 18:52 |
=== flaviami_ is now known as flaviamissi | ||
=== bac_ is now known as bac | ||
=== deryck[afk] is now known as deryck | ||
davecheney | fwereade_: what is the story with the trunk ? | 21:25 |
davecheney | can we land fixes ? | 21:26 |
mgz | davecheney: I thought it had been mentioned that it's okay to land again, but I now can't find a reference | 21:28 |
mgz | release things have all been branched though | 21:28 |
davecheney | mgz: mramm fwereade_ : please email juju-dev with the status of the trunk | 21:29 |
davecheney | mgz: i saw you were unassigning issues from 2.0 | 21:29 |
davecheney | thank you | 21:29 |
mramm | feel free to use trunk | 21:29 |
davecheney | mramm: email please | 21:29 |
mramm | will do | 21:29 |
davecheney | just copy and paste this discussion | 21:30 |
mramm | it's posted | 21:32 |
mramm | and I've updated the agenda so we talk about that | 21:32 |
mramm | and so that we talk about the backports stuff | 21:32 |
mramm | and general release status | 21:32 |
mramm | I will try to write up a release status update e-mail later this evening | 21:33 |
mramm | but right now there is a lot up in the air, and I need some time to get all the details organized and written down | 21:33 |
mramm | and I have some personal stuff to take care of in a few min, so I don't think it will make it before I leave for that... | 21:34 |
davecheney | thanks mark | 21:37 |
=== wedgwood is now known as wedgwood_away | ||
=== Makyo is now known as Makyo|out |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!