/srv/irclogs.ubuntu.com/2013/08/22/#juju-dev.txt

davecheneyunsubtle reminder, https://docs.google.com/a/canonical.com/document/d/1eeHzbtyt_4dlKQMof-vRfplMWMrClBx32k6BFI-77MI/edit#heading=h.reab82nqtett00:00
arosalesbigjools, thanks again for the help on the azure bits00:08
bigjoolsarosales: no prob00:09
bigjoolswallyworld gets the baton next week and he told me he's really looking forward to it00:09
wallyworldno he's not and no he isn't00:10
bigjoolsdavecheney: I want to move the gwacl code on to the juju landing bot. Where do I start?00:57
davecheneybigjools: i'm sorry I don't know how to do that00:57
davecheneyjohn owns the bot00:57
bigjoolsok HAL00:57
bigjoolsjam, when you get in.... ta :)00:58
smoserbigjools, so saucy + azure + juju core "works for me"01:32
bigjoolssmoser: yeah I just got it working too01:32
bigjoolsNFI why it didn't before01:32
smoserhttp://paste.ubuntu.com/6012428/01:33
smoserbigjools, i think you just were'nt being patient.01:33
bigjoolssmoser: 30 minutes ought to be enough patience.01:34
bigjoolsand actually I left it several hours01:34
bigjoolsalthough I have a sneaking suspiscion that it was booting an image without the new cloud-init01:35
smosernot on saucy.01:36
smoserCloud-init v. 0.7.3 finished at Wed, 21 Aug 2013 21:33:21 +0000. Datasource DataSourceAzureNet [seed=/dev/sr0].  Up 534.78 seconds01:36
smoserthats from bootstrap node.01:36
smoseri certainly dont think you could have been running saucy without cloud-init.  there was a fix in a few days ago01:41
smoserCloud-init v. 0.7.3 finished at Wed, 21 Aug 2013 21:48:55 +0000. Datasource DataSourceAzureNet [seed=/dev/sr0].  Up 520.95 seconds01:41
smoseroops01:41
smoserhttps://bugs.launchpad.net/ubuntu/+source/cloud-init/+bug/121454101:41
_mup_Bug #1214541: hostname setting is erroring out <amd64> <apport-bug> <cloud-images> <saucy> <cloud-init (Ubuntu):Fix Released> <https://launchpad.net/bugs/1214541>01:41
smoserbut that didn't really stop boot.01:41
smoserbigjools, one thing to note, it sure appears to me that you're sending in the linuxprovisioningxml01:43
smoser<HostName>default</HostName>01:43
bigjoolssmoser: not sure we even set that01:44
smoseroh. maybe not.01:44
smoserit probably makes sense to send the hostname as the hostname :)01:44
thumperaxw: I don't suppose you can work your test speed ups on the ec2 tests?02:15
axwthumper: heh :) I was planning to go across the codebase02:15
thumper\o/02:15
axwjust responding to fwereade's comments on manual provisioning now02:15
thumpercool02:22
* thumper is trying to finish off the refactoring branch of doom02:22
axwthumper: which is that? agent config changes?02:23
thumperaxw: yeah, replacing agent.Conf with an interface02:23
thumpergetting kinda epic02:24
axw:)02:24
thumpereven for the minimum case02:24
axwthumper: https://codereview.appspot.com/12831043/patch/8002/9024  -- second last comment. fwereade suggested I ask you about whether the upstart job's name ought to go into agent config02:31
axwwhat do you reckon?02:31
thumperhmm...02:32
* thumper looks and considers02:32
thumperaxw: so my answer is yes, but not now02:33
axwokey dokey02:33
thumperaxw: I want to add an environment type thing to agent config02:33
thumperbut that is coming after the mega branch02:33
axwenvironment type thing?02:33
thumperso the agent config can have key/value string pairs02:33
axwah right02:33
thumperfor arbitrary extra info02:33
axw*nods*02:34
axwI will add a TODO here then02:34
thumperkk02:34
wallyworldthumper: axw: you guys run the ec2 tests on trunk recently? i have a failure in TestAddresses02:45
thumperwallyworld: have you updated goamz?02:45
wallyworldthumper: axw: also, reading back scroll, rodger is working on ec2 test speedups02:45
wallyworldmaybe not02:45
thumpermy trunk from the other day works02:46
wallyworldok, i was out of date, thanks. will try again in a bit02:46
axwwallyworld: I think gz changed something. and yeah I saw the other day he sped them up significantly - not sure where he's at with it now. I'll ping him later02:47
wallyworldit may have been rodger not martin02:48
axwsorry, I conflated two comments02:48
axwyes, rog02:48
thumperOOPS: 1 passed, 3 FAILED, 1 PANICKED, 4 MISSED02:53
thumperbut at least they now freaking compile02:53
thumperthat is the agent tests02:53
* thumper digs02:53
thumper2 passed02:55
thumper3 passed02:57
thumper4 passed02:58
thumperkid run02:59
* thumper stabs useless tests in the face03:19
thumpertesting something that never happens and doesn't make sense03:19
thumperomg might be there...03:39
thumperrunning make check03:39
axwwoo :)03:40
thumperheh03:41
thumperI got more strict in what I expect03:41
thumpernow I have a few more failures to fix03:41
thumperalthough should be trivial03:41
thumpernot as trivial as I had hoped...04:03
thumper22 files changed, 781 insertions(+), 866 deletions(-)04:05
thumpertotalling around 2.5k changes04:05
thumperat least william will do it :)04:05
thumperfinally, all tests pass04:17
thumperfor the interested: https://codereview.appspot.com/13164043 (and fwereade)04:24
* thumper goes to take kid to sport04:24
thumperback for the meeting04:25
=== thumper is now known as thumper-afk
axwadios04:25
jambigjools: I'm around, where do you need gwacl moved to?05:24
jamalso, wallyworld, dimitern and some other people should know how to do it in case I'm not available05:24
bigjoolsjam: I currently have a landing bot on my home server, and I want it moved so the juju bot lands it please05:26
jambigjools: ah, just moved to the bot. Sure. Is it just tarmac?05:26
bigjoolsthe project is part of juju05:26
bigjoolsyes, thanks05:26
bigjoolsI can give you the tarmac conf later if you need it - my home server decided to crash and burn so I am rebuilding it05:26
jambigjools: so the "standard" config I have is this: http://paste.ubuntu.com/6012846/05:28
jamI think we need a little bit of change for your code, because we need to put it in a different GOPATH so you can land stuff that would otherwise break juju-core until you have the patch that matches juju-core again.05:29
jambigjools: but otherwise, whatever verify_command, etc you want , just let me know (make check, etc)05:29
jambigjools: maybe it was only with older tools but "-gocheck.v" wasn't compatible with "./..." when I've tested it in the past.05:30
jamI guess it doesn't matter terribly, but I'm guessing your bot wasn't testing "dedent" and "logging" on each run.05:31
jam"make check | grep dedent" doesn't show anything for me05:32
bigjoolsjam: make check is all you need05:32
jambigjools: but it doesn't run the tests in the subdirs05:32
jam(as I mentioned just now)05:32
jamit is a bug in "go test" but if you pass special flags like "-gocheck.v" it doesn't notice "./..."05:32
bigjoolsmake check works fine locally, what's up with the bot?05:33
jambigjools: test it yourself "make check | grep dedent"05:33
jamempty output for me05:33
jamon my personal machine05:33
jamgo 1.1.105:33
bigjoolswhy are you grepping for dedent?05:34
jambigjools: because that is part of your test suite "dedent/dedent_test.go"05:34
bigjoolsoh////05:34
bigjoolsmeh05:34
bigjoolssorry I am slow todya05:34
jambigjools: so your "make check" doesn't run tests in subdirs05:34
bigjoolsjam: ok so the regular go test ./...05:36
bigjoolsand05:36
bigjoolsgo build example/*/run.go05:36
bigjoolswe can add something for formatting later05:36
bigjoolsalternatively we can add a new make target for the bot05:36
davecheneyafternoon all, see in yo in a few hours for the meeting05:57
=== wallyworld__ is now known as wallyworld
=== jam1 is now known as jam
=== tasdomas_afk is now known as tasdomas
=== thumper-afk is now known as thumper
natefinchjam: is the hangout link different from the morning standup link?08:03
axwnatefinch: https://plus.google.com/hangouts/_/bf3f4cfe715017bf60521d59b0628e5873f2a1d3?authuser=108:04
natefinchaxw: thanks08:04
jamnatefinch: you should be on the calendar entry, which is the g+ hangout we use. Just checking that it is working for you.08:08
mrammhey, for some reason I can't get back in the hangout08:16
nate-finchjam: Yep, I have the calendar entry - I just missed the link in the event... I'm too used to seeing a long ugly URL, not a small "Join Video Call" link :)08:19
wallyworldfwereade: if you want a (slightly) smaller branch ( 2392 lines (+425/-580) 36 files modified) as an entree to thumper's: https://codereview.appspot.com/13081044/   plus I updated https://codereview.appspot.com/13047043/09:11
fwereadewallyworld, aww, thanks ;p09:11
wallyworldi'm here to please :-P09:11
wallyworldit's a net deletion of code which i a good thing, a lot of duplicated stuff gone :-)09:12
fwereadewallyworld, cool, thanks09:15
axwjam: just looking at the upgrade bug again...09:19
axwwhat happened was, I did an upgrade-juju --upload-tools in local09:19
axwthe bootstrap machine agent got the tools it needed to upgrade, and restarted09:19
axwthen upgrade-juju proceeded to attempt to upload additional tools, and failed09:20
axw*additionally* the machine agent just restarts in a loop09:20
fwereadeaxw, huh, don't we upload all the tools before we set the agent version?09:28
* axw has a look09:28
axwI have no idea :)09:28
fwereadeaxw, (and, sorry, which machine agent? 0?)09:28
axwyes, 009:28
fwereadeaxw, don't suppose the log said anything useful about why?09:29
axwthere are some errors about connecting to a websocket? but I haven't had time to dig09:29
* fwereade sighs09:30
fwereadethanks axw09:30
fwereadeI think it's probably clear how to fix it09:30
fwereadewe just need to wait to make connections rather than bouncing every time things go wrong09:31
axwfwereade: yes it does appear that the version is set last. Perhaps I already had machine-0 in a bad state when I attempted to upgrade again09:36
axwbut the simple upgrade-juju is still a problem09:37
axwi.e. the steps in #1214676 are repeatable09:37
_mup_Bug #1214676: upgrade-juju in local environment causes bootstrap machine agent to restart continuously <juju-core:Triaged> <https://launchpad.net/bugs/1214676>09:37
fwereadeaxw, sure, I'm not trying to claim otherwise -- thanks :)09:39
* fwereade breakfast09:39
jam axw: it sounds like we should be uploading all the tools before we ask things to upgrade themselves.09:40
axwjam: possibly we are. I'm no longer sure about the interaction of this bug and the one I link to the description09:41
axwI'll have to try and reproduce the original one again09:41
axwanywho, gotta get dinner on - I'll be back for a bit later09:41
jamnp09:42
jamhave a good evening09:42
fwereadewallyworld, remind me: are we now requiring that we bootstrap with matching version rather than picking latest? or is that really really imminent?10:23
wallyworldfwereade: soon. i justed wanted to get the core refactoring done first. there will be a lot of tests etc to fix when we change the boot strap tools search10:24
wallyworldand the branch was already huge10:25
fwereadewallyworld, sure, understood :)10:36
fwereadewallyworld, btw, why the change from tools-metadata-url to tools-url? the first feels more correct to me10:40
fwereademgz, you have quite a lot of branches up that I think are already LGTMed, are there any I should be looking at?10:41
mgzfwereade: nope, though last check I think a couple of the trivial ones weren't lgmted, I'll look again10:43
yolandahi, i'm using juju-core with canonistack. I destroyed a service but now it keeps on status "dying" for all the time. Tried to terminate the machine, removing the unit, without success, and now i cannot redeploy that again although i destroy all the environment. Is there some way where i can manually destroy that buggy service?10:43
mgznope, all ready to land, I'll queue 'em up10:43
mgzyolanda: you may be able to resolve the unit, then destroy it, depending on exactly what state things got confused in. otherwise, it's destroy-environment to wipe everything10:45
yolandamgz, machine was stuck with out of memory errors, but charm state looked as started10:46
yolandai tried everything without success, now i manually removed the machine with nova delete, but no progress10:46
mgzyolanda: what's the current state of the unit? can you pastebin status?10:46
mgzah, probably not easily cleanupable any more then10:46
mgzdestroy-environment for the wipe :)10:47
yolandahttp://paste.ubuntu.com/6013589/10:47
yolandabut that's after all the things i've done10:47
yolandai was trying to setup a jenkins unit, but complained about memory10:47
yolandaok, i'll destroy everything and try again10:47
mgzyou probably need a constraint on mem=somethingbig for the jenkins service10:48
* TheMue => lunchtime10:50
yolandamgz, will try with 2G now10:52
jamfwereade: https://plus.google.com/hangouts/_/6c44984654c782bf948ffabe55b69414a3a38aea ?11:05
arosalesjam, thanks for the ideas on azure11:15
nate-finchI was up at 4am, and I'm in the standup....11:32
jammgz: wallyworld: https://plus.google.com/hangouts/_/f497381ca4d154890227b3b35a85a985b894b471 standup ?11:34
nate-finchmgz: standup11:34
mgzta11:34
jammgz: poke if you could ask jamespage whether they will be getting go-1.1.x into precise's backports12:00
yolandamgz, problems with jenkins again, now nova show machine is on error state, and juju shows it as pending, is there any way to solve it without destroying?12:07
yolandai got a nasty error:12:07
yolandahttp://paste.ubuntu.com/6013816/12:08
jamespagejam: you could ask me yourself :-)12:11
jamjamespage: I thought given it was Thurs he might be sitting right next to you, and could have the high-bandwidth conversation12:11
jamthe question also originated on our phone call, but he lagged out for a bit.12:12
jambut yes, jamespage, what is the current plan for getting juju into precise-backports and its associated build dependencies12:12
jam?12:12
fwereadeTheMue, ping12:30
fwereadeTheMue, I'm in the standup hangout when you're ready12:32
TheMuefwereade: I wanted to to it here, but hangout is fine too.12:41
TheMuefwereade: still there?12:41
jamjamespage: that's why I don't ask you, you don't answer :)13:15
jamespagejam: urgh - sorry13:15
jamespageI have my head in a nasty openvswitch datapath compatibility issue13:15
jamespageand was mid cherry pick13:15
jamnp13:16
jamit isn't urgent13:16
jamespagejam: right now backporting is on hold13:16
jamespagejam: two reasons - 1) we now have a stable PPA and 2) 12.04 cloud archive is going to get a cloud-tools pocket which would have golang + other stuff in13:16
=== tasdomas is now known as tasdomas_afk
* fwereade needs to be off a bit early today, see you all soon16:43
mgz_later fwereade!16:44
TheRealMuecu16:54
=== tasdomas_afk is now known as tasdomas
=== tasdomas is now known as tasdomas_afk
thumpermorning21:07
* thumper stretches21:10
weblifemorning thumper21:15
thumperhi weblife21:17
thumperfwereade: hey21:32
thumperfwereade: you really around?21:32
* thumper sighs21:44
thumperlots of test failures after merging trunk21:44
* thumper does some manual testing22:05
thumperbah humbug22:21
* thumper has broken the local provider... locally at least22:21
* thumper enfixorates22:21
fwereadethumper, heyhey22:23
thumperfwereade: hey22:23
thumperfwereade: I was going to ask about the breaking bootstrap change22:23
thumperand now my local provider is unauthorized in bootstrap22:24
* thumper needs to figure out how...22:24
fwereadethumper, it's just that juju currently bootstraps with the latest available tools, regardless of client version -- but is it possible it's finding *older* installed tools and trying to bootstrap with those?22:25
fwereadethumper, wallyworld is going to be changing it to demand matching tools on bootstrap, which I think will help22:25
thumperI remember change some behaviour around the initial password22:25
thumperfwereade: but I didn't change that, right?22:26
fwereadethumper, ehhh, I felt a tiny twinge around that but read it quite carefully and *thought* it was solid22:26
thumperyeah, I thought it was solid too22:26
thumperbut not enough to commit it to trunk without trying it out :)22:26
thumperec2 is working22:27
thumperhmm... "bzr ssh 0 " doesn't do what I mean22:27
thumperI mean "juju ssh 0"22:27
thumpermy computer needs more mind reading22:27
fwereadethumper, well, I am inclined to prejudge it as all being about the upload/sync confusion in the local provider, but that says more about me than about juju I think22:28
fwereadethumper, ha, yeah, I confuse the two quite a lot22:28
thumpercan't wait until we can change the certs to be stored in base6422:28
thumperkinda crazy looking at the config files now22:28
thumperfwereade: no, I don't think it is that...22:29
thumperfwereade: more likely to be assumptions made during the initial config writing and bootstrapping22:29
fwereadethumper, ha, yes, tis true those bits don't quite match the non-local case either, right?22:30
thumperno, it is a bit hand wavey22:30
thumperand "this is not the config you are looking for"22:30
fwereadethumper, haha22:31
thumperthere is a very slight change in behaviour22:31
thumperprior, when writing out the new password22:31
thumperit would not keep the old one22:31
thumperbut replace it with ""22:31
thumpernow I keep the old one around22:31
thumperhowever I don't think that is the current problem22:32
thumperthere is an assumed two-phase commit with the passwords22:32
thumperwhich isn't quite there22:32
thumpersave the password, then change it22:32
thumperand hope all is god22:32
thumpergood22:32
thumperhowever since it is possible for the rename of the file to fail22:33
thumperthere is a possibility, however small, that we get stuffed up22:33
fwereadeheh, damn, I missed that; I'd prefer not to lose that just because it's a possible screwup22:34
fwereadeI doubt it's what you're seeing though22:34
thumperit is no different to before22:34
thumperjust I realised it is there22:34
thumperthe behaviour is the same22:34
thumperno, this isn't what I'm seeing22:35
thumperit is just an unauth on the initial create log db part in mongo22:35
fwereadeah, that's good then, I *thought* I had verified it still was, but was feeling ready to assume hallucinations on my part22:35
thumperah fuck22:44
thumperI think I know what it is22:44
thumperfuckity fuck fuck22:44
thumperit is the magic "localhost"22:44
thumperwhich I noticed while doing the local provier22:44
thumperprovider22:44
* thumper double checks22:45
thumperyep...22:48
thumperat least that fixed22:58
thumperit22:58
* thumper wonders if he'll conflict with wallyworld23:05
wallyworldhope not23:06
* thumper just checks23:06
thumperwallyworld: how are you doing?23:06
wallyworldok, landing a bunch of stuff23:08
wallyworldthumper: i was hoping to get in first with my stuff. now i will have lots of conflicts i'm sure :-(23:11
thumperwallyworld: well, mine is in the queue too :)23:11
bigjoolsgood day23:11
thumperthe freaking epic add an interface branch23:12
thumperhi bigjools23:12
wallyworldthumper: yours in landing now23:12
thumper\o/23:12
bigjoolswallyworld: where did you buy your microserver?23:12
wallyworldbigjools: some online place, i searched on staticice.com.au23:13
bigjoolsok ta23:13
bigjoolshp's site is useless23:13
wallyworldyep23:13
wallyworldthumper: only one conflict :-)23:24
* bigjools hugs hot coffee23:26
* wallyworld still has no coffee machine :-(23:26
thumperwallyworld: still?23:27
thumpergeez23:27
thumpergo get a rental23:27
* thumper needs another coffee23:27
thumperwallyworld: I have some errands in town23:32
thumperso will be back later23:32
bigjoolswallyworld: you can always work here23:32
thumperhappy to have my epic branch landed though23:32
wallyworldthumper: i'm about to go to the police station myself23:32
thumperwallyworld: is you tools branch the one that fixes bootstrap?23:32
thumperso don't try to bootstrap with old tools?23:32
thumperwallyworld: ack23:33
wallyworldthumper: not quite. next branch23:33
thumpercool23:33
thumperI'll be doing more reviews this afternoon23:33
wallyworldmine does a bunch of other pre -req stuff23:33
thumperaround axw's work on manual provisioning23:33
wallyworldyay23:33
* thumper nods23:33
thumperI want to add a new branch for adding string key/value pairs into agent config23:34
thumperwhich would lead to the logging branch working23:34
thumperand tweaking the maas stuff for containers23:34
thumperbigjools: I have a fix for the maas provider which is less smash of the network conf23:34
thumperbigjools: using "source" as an include mechanism23:34
thumperbigjools: so when maas gets better at network handling23:35
bigjoolsthumper: splendid.23:35
thumperit won't kill it23:35
bigjoolswhen you say better ...23:35
thumper:)23:35
bigjoolswe're adding more constraint types next week23:35
thumperwhat does that mean for me?23:35
bigjoolsso the core had better hurry up and support it :)23:35
=== thumper is now known as thumper-afk
bigjoolscustom tags as per the pyjuju23:36
bigjools--to in the new world IIRC23:36

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