/srv/irclogs.ubuntu.com/2013/09/20/#juju-dev.txt

thumperhi axw01:29
axwhey thumper01:29
thumperaxw: what does ReadFull do?01:30
axwReadFull reads exactly len(buf) bytes from r into buf. It returns the01:30
axw    number of bytes copied and an error if fewer bytes were read.01:30
thumperoh cool01:30
thumperyeah, that's better than the loop01:30
thumperaxw: I only noticed it because I occasionally habitually run "make check" when I switch to an updated trunk01:36
thumperand it errored the first time and not the second01:36
thumperactually, it didn't fail again01:36
axwthat's race conditions for you :)01:37
axwglad it didn't cause too much strife01:37
wallyworldthumper: last one for the 1.15 release i hope https://codereview.appspot.com/1380004301:42
thumperaxw: no, not strife. I've spent lots of time in the past dealing with race conditions01:42
* thumper objects horribly01:50
axwthumper: re jc.Set?01:52
thumperyes01:52
thumpernow on juju-cev01:52
thumperdev01:52
thumperhow about testing/basesuite01:53
thumperor basesuties01:53
thumpersuites01:53
axwtesting/testbase?01:53
thumperI'd be happy with that01:53
thumpershall I do it now?01:53
axwdoesn't have to be about suites01:53
thumpersure01:53
thumperworth having any test helper methods and functions that don't drag in extra dependencies01:54
thumperkeep it loopless if possible01:54
thumpershould be possible01:54
axwI say go for it01:54
* thumper finishes wallyworld's review then goes for it01:55
wallyworld\o/01:55
wallyworlds/review/reviews01:55
wallyworldand we need to talk about logging01:55
thumperack01:56
* thumper dedicates 20 minutes to this task01:56
thumperif it takes longer than that, interrupt me01:56
thumpermake that 30 minutes01:56
axwthumper: it can wait till later, but am I likely to get a lgtm on sshstorage today?01:57
thumperaxw: yes, I'm at the disposal of you two today01:57
thumperafter this hacking01:57
axwbueno01:57
axwthanks01:57
* axw looks at authenticated httpstorage01:58
thumperbugger02:06
* thumper forgot to add the directory before moving things to it02:07
thumperbzr lost track02:07
* thumper found bzr mv --after02:08
mrammhey thumper got a couple of min?02:08
mrammI think robbie (kvm robbie not cdo robbie) is out all next week02:09
thumpersure02:09
mrammso if you want to round trip with him on kvm issues it is going to have to be today :/02:09
thumperyeah...02:09
thumperhe scheduled a meeting 1am nz saturday02:09
mrammouch!!!!02:09
mrammthat is nutto02:09
thumperyeah, at least I'm down as optional :)02:11
thumpermramm: did you want to hangout?02:11
mrammI just wanted to tell you that news02:11
mrammand to let you know that Dan is *very* concerned about kvm stuff02:12
mrammhe says it is pretty well tested on his end, and wanted me to give him a detailed view of what problems you've had02:12
thumperI'll forward you the email02:12
mrammthanks02:13
davecheneymramm: is there background on the kvm issue ?02:14
davecheneyi have heard 512mb + kernel overcommit == 0 is the issue ?02:14
davecheneymramm: we believe this is fixed in golang trunk02:14
davecheneyfixed is an oversimplification of the problem02:16
davecheneyif this is the problem I think it is02:16
davecheneydan will have the same issue running jvm applications in those small kvm containers02:16
davecheneythumper: mramm ping02:17
thumperhi davecheney02:17
davecheneythumper: may i butt into this issue02:17
davecheneyi believe i have experience02:17
mrammdavecheney: the issue is with the kvm wrappers that they wrote02:18
mrammnot in any way with jvm02:18
mrammand it's just about us supporting kvm containers *inside juju*02:19
davecheneymramm: ok02:19
davecheneythanks02:19
davecheneymramm: what does kvm inside juju mean ?02:19
mrammlike lxc containers but kvm containers02:21
mrammEG deploy ceph inside kvm on a machine as part of an openstack install02:21
mrammok, gotta run02:23
mrammgetting late here.02:23
axwgnight02:25
davecheneymramm: how does this problem relate to the 1gb issue ?02:29
davecheneyare they unrelated02:29
davecheneyi'm only interested in the latter02:29
mrammwhich problem?02:29
mrammkvm?02:29
davecheneymramm: yes02:29
mrammthey are completely unrelated02:29
mrammyes02:29
davecheneyok02:29
davecheneyplease view my comments above with respect only to the 1gb issue02:30
mrammand the later is that we can't bootstrap on a 512 node anymore02:30
davecheneymramm: ahh, that is why i am confused02:30
davecheney512mb nodes are a kvm thing ?02:30
davecheneycorrect ?02:30
davecheneymramm: is there an issue for the 1gb / 512mb problem ?02:37
thumpermramm: sorry those emails hadn't arrived, I thought I had sent them but the email client was stuck asking me to sign things02:43
mrammdave I don't know where the issue is or even if there is a bug filed, sorry02:45
davecheneymramm: i will follow up with rog to see if he raised an issue02:47
davecheneymramm: i've seen the other side of the conversation on the golang-dev ML02:47
davecheneythe gui has just crapped out on me as well02:49
davecheneylooking at the unit02:49
davecheneythe gui process has died ...02:49
davecheneyoooooooooooooh, interesting02:51
davecheneythis is a chrome issue02:51
davecheneyone chrome tab was showing the SSL warning02:51
davecheneywhich was blocking any other connectino to that site ....02:51
thumpernice huge mechanical branch coming02:54
thumperwallyworld, axw: rack up the reviews you need in the order you need them02:58
thumperI'll drill through them after I have kids from school02:58
* thumper goes to put on shoes02:58
axwthumper: I see you've lgtm'd sshstorage (thanks); just this one now: https://code.launchpad.net/~axwalk/juju-core/sanity-check-constraints/+merge/18501502:58
wallyworldhttps://codereview.appspot.com/13763044/02:58
wallyworldhttps://codereview.appspot.com/13800043/02:59
thumperaxw: you can take a look at my testbase branch02:59
axwthumper: yep, will do02:59
thumperhttps://codereview.appspot.com/1369404602:59
thumperall it does is move a few things02:59
thumperaxw: are you landing the sshstorage?03:09
thumperif so, I'll wait for my as they'll clash03:09
axwthumper: I can wait for you to land that one if you like03:09
thumperok, I'll land it03:10
axwthanks03:10
thumperaxw: I'd prefer to avoid importing checkers into testbase, but we could, as long as we make sure that it is just that03:10
axwthumper: nps, just leave it till it's an issue03:10
thumperapproved it, so lets see if it lands :)03:11
axwthumper: did you land the sshstorage test one?03:11
axwyes03:11
thumperyes03:12
axwthumper: does mocking out ssh/sudo really make the test whitebox? the fact that it's in the same package is really just an unfortunate requirement03:14
thumperhmm...03:14
thumperperhaps not03:14
thumperI'm flexible03:14
axwI'll leave it. I think we may move the SSH stuff out to another package later anyway, for centralising common flags03:15
thumperwallyworld: /environs/tools/tools_test.go is still using filepath, do you know where?03:23
wallyworldin the mp?03:23
thumpernm03:24
thumperI' look03:24
thumperand it is ok03:24
wallyworldsigh, another one https://codereview.appspot.com/1369404703:24
=== jtv1 is now known as jtv
thumperkid says "take me to hockey now, I have a game at 4"03:32
* thumper afk for a few minutes.03:32
thumperaxw, wallyworld: where are we up to on reviews needed?03:57
wallyworldthumper: i've got 303:57
axwthumper: I'm all good now03:57
wallyworldlet me check them to see what's approved03:57
axwthumper: I'm just going to make sure my null provider stuff still works with all the changes to sshstorage and co, then I'll land manual bootstrap and propose null provider03:58
wallyworldthumper: default hp cloud set up is not needed anymore because there's nothing hp cloud specific anymore03:58
thumperwallyworld: how do the users point openstack at hpcloud?03:59
wallyworldwhat do you mean?03:59
thumperaxw: ok03:59
thumperwallyworld: well, I want to go bootstrap03:59
wallyworldthey set their credentials same as now03:59
thumperhow does it know where to look?03:59
wallyworldit uses the credentials03:59
thumperyou suggest os env vars03:59
thumper?04:00
wallyworldno, they set in yaml or env vars04:00
wallyworldtheir choice04:00
thumperok, fair enough04:00
thumperdo the users know that hpcloud is openstack?04:00
wallyworldbefore, hp cloud was different because the public bucket url was needed04:00
wallyworldthey don't need to know that04:00
thumperI'm just wondering if it is nice to keep hpcloud as an option since it is a certified cloud04:01
wallyworldthe only hp cloud difference was the public bucket url, else it is just standard openstack04:01
thumperwallyworld: but you have removed hpcloud as an option04:01
wallyworldno need04:01
thumperin the config04:01
wallyworldyes, because there is no hp cloud specific config04:01
thumperI know there is no *need*, I just think it is useful04:01
wallyworldit's done via credentials04:01
wallyworlduseful how?04:01
* thumper slaps wallyworld for not listening04:01
* wallyworld slaps thumper for the same reason04:02
* thumper slaps wallyworld for slapping him04:02
wallyworldhangout perhaps?04:02
thumperwallyworld: yes, please04:02
thumperhowever04:02
thumperI need to go get kid 3 in 2 minutes04:02
wallyworldok04:02
thumperonly takes a few minutes04:02
wallyworldwith the tests, i deliberately used text04:02
thumperI'll create a hangout and ping you when I'm back04:02
wallyworldnot a const04:02
wallyworldok04:02
thumperwallyworld: https://plus.google.com/hangouts/_/6f17da147ed6be734a284c49eb9c67086eef59b0?hl=en04:19
wallyworlddavecheney: hi still working on getting the code for the release ready. i'll update the release notes tonight. i assume that will be ok?04:57
davecheneywallyworld: sure04:59
davecheneywallyworld: it's more important to get it right, than to get it out there tonight05:00
davecheneyif you need more time, just say05:00
wallyworldyep. pushing shit uphill, but getting there05:00
davecheneywallyworld: the friday date is fairly arbitary05:01
wallyworldi've 3 branches to land then test05:01
=== tasdomas_afk is now known as tasdomas
wallyworlddavecheney: do you have credentials to write to the s3 public bucket?05:26
davecheneywallyworld: will /msg05:26
wallyworldok05:26
thumperick05:29
thumperaxw: got a minute?05:29
axwthumper: yes?05:29
thumperaxw: hangout?05:30
axwsure05:30
thumperaxw: https://plus.google.com/hangouts/_/43172cdd0670d5faaae767a84e0c0fe72923a9ea?hl=en05:30
davecheneyAAAAAAAAAAAAARHRHG05:37
davecheneylucky(~/charms/precise/galaxy/hooks) % juju add-machine -n1005:37
davecheneyerror: flag provided but not defined: -n05:37
davecheneyWHY05:37
wallyworlddavecheney: are you thinking of add-unit?05:38
wallyworldadd-machine doesn't take a -n parameter05:39
davecheneywallyworld: it would be good if it did05:46
wallyworldyeah. no one has asked for it that i know of05:46
davecheneyhttps://bugs.launchpad.net/juju-core/+bug/121420905:46
_mup_Bug #1214209: cmd/juju: add-machine should take a -n param <papercut> <juju-core:Triaged> <https://launchpad.net/bugs/1214209>05:46
wallyworldfair enough05:47
axw:(05:58
axwthumper: you know how I changed sshstorage to create/remove the tempdir on Put?05:59
axwit's not going to work05:59
thumperwhy05:59
axwstorage goes to /var/lib/juju/storage05:59
axwtempdir to /var/lib/juju/storage.tmp05:59
thumperand...06:00
axw/var/lib/juju/storage gets created under sudo, and chown'd to $SUDO_UID06:00
axwthen.,..06:00
thumperhmm06:00
thumperhow about /var/lib/juju/storate/.tmp ?06:00
axwanother, non-sudo shell is executed to do the transfers06:00
axwthat just means we can't have .tmp in storage, which I suppose is reasonable06:00
rogpeppemornin' all06:01
axwmorning rogpeppe06:01
rogpeppeaxw: hiya06:01
davecheneyave06:01
axwthumper: by which I mean, the user can't Put anything starting with ".tmp"06:02
thumperaxw: yeah...06:03
thumperaxw: however your tmp dir is now removed after right?06:03
thumperaxw: doesn't the function to create a tmpdir take a prefix?06:03
thumperand won't clash with names?06:03
axwthumper: yes, but if you die before removing it06:04
axwthen you have something in storage06:04
thumperyes, that is a risk06:04
thumperbut managable compared to the other options I think06:05
rogpeppethumper: testbase looks good (not keen on "base" in names in general, but can't think of a better one here)06:10
thumperrogpeppe: axw and I brainstormed some names06:10
thumperand it was the best we could come up with06:10
thumperI'm particularly fond of the test that makes sure it doesn't grow dependencies06:10
rogpeppethumper: one thing though: how about Patch rather than PatchValue? the lines that use it tend to be longish anyway, so the extra horizontal space is useful.06:11
thumperPatch doesn't say what it is patching06:11
thumperit isn't a descriptive name06:11
rogpeppethumper: it's patching its argument, which is (by definition) a value06:12
thumperno, sorry I disagree06:12
rogpeppethumper: "Value" doesn't really add any, erm, value there06:12
rogpeppethumper: i think it works well for the basic patching mechanism06:13
rogpeppethumper: and it's a reasonable compromise between PatchValue and Set :-)06:13
thumpermy suggesting is this: if you still think strongly that we should change it by the time we all get together in SFO, then we can get a team concensus06:14
thumperthen we can run sed -i on the codebase06:14
thumperbut give it a little time to settle06:14
thumperwe should schedule a session for pure bikeshedding :)06:15
* thumper is at EOW06:15
* rogpeppe sobs that it was his shed originally :-)06:15
rogpeppethumper: have fun06:15
* thumper throws pink paint at rogpeppe's shed and runs06:15
wallyworldrogpeppe: are you an ec2 expert? we are getting access denied trying to access http://juju-dist.s3.amazonaws.com/ and i have no idea why06:16
wallyworldhence juju is currently broken for ec206:16
rogpeppewallyworld: oh shit06:16
rogpeppewallyworld: i am by no means an ec2 expert06:17
wallyworldi tried sync tools and aborted it after it had uploaded a few tarballs. not sure if that caused it. i can't see how06:17
wallyworldi also aborted uploads to hp cloud etc and no problem06:17
rogpeppethat *shouldn't* have caused a problem06:17
rogpeppewallyworld: so did it stop working immediately after your abort?06:18
wallyworldi think so06:18
wallyworldmay fwereade knows? or mgz?06:22
rogpeppewallyworld: i don't have any s3 access at all06:23
rogpeppewallyworld: i wonder if this might be an amazon issue06:23
wallyworldi sure hope so06:23
rogpeppe% s3 put /tmp/x s3://pefkesnjesesvds06:23
rogpeppeERROR: S3 error: 403 (InvalidAccessKeyId): The AWS Access Key Id you provided does not exist in our records.06:23
wallyworldso i guess we wait a bit06:24
wallyworldand see if they fix anything06:24
rogpeppewallyworld: hmm, status.aws.amazon.com doesn't indicate anything amiss06:24
rogpeppewallyworld: could you install s3cmd and try making a bucket, please?06:25
rogpeppewallyworld: just to see if you get the same issue06:25
wallyworldok06:25
wallyworldi can create one06:26
davecheneywallyworld: do i need to declare an emergency06:28
hazmatwallyworld, i'm an expert..06:28
davecheneyit's 3:30pm in japan06:28
wallyworldhazmat: any idea why we are getting access denied all of a sudden trying to read the juju-dist bucket?06:28
hazmatdavecheney, its sounds like its just an upload issue not a public issue06:28
hazmatwallyworld, are you using the correct account..06:29
hazmatwallyworld, are you sure you don't have some openstack/canonistack stuff in your env06:29
wallyworldhazmat: i was testing sync-tools and aborted a tarball upload part way through with Ctrl^C. but surely that could not have caused it?06:29
rogpeppehazmat: it should be globally readable06:29
hazmatie. your actually using an aws account06:29
wallyworldhazmat: yes06:29
hazmatrogpeppe, we're talking about upload06:29
hazmatwallyworld, i've seen that error when i accidentally try to use an openstack account against aws06:29
hazmatalternatively when an iam account i'm using is deleted06:30
rogpeppehazmat: i can't access that url, which i think i should be able to06:30
hazmatrogpeppe, who is the owner of the account / bucket?06:30
wallyworldhazmat: i'm almost 100% sure i used the correct account details06:30
davecheneyhazmat: gustavo06:30
rogpeppehazmat: niemeyer06:30
hazmathmm.. too early for him..06:31
hazmatone moment.. checking the bucket with creds06:31
rogpeppewallyworld, hazmat: FWIW i see the same issue, though i can in fact upload to my own buckets (i'm not sure why s3cmd was failing to do that)06:35
hazmatwallyworld, davecheney, rogpeppe ... okay verified the account works06:35
hazmati can upload files to the bucket, and remove them06:36
rogpeppehazmat: to juju-dist?06:36
hazmatrogpeppe, yes06:36
hazmatthe problem is that the permissions on the account are restricted06:36
wallyworldso why do we get access denied reading it?06:36
hazmati can't for example list all the buckets on the account06:36
hazmatbut i can list th contents of that bucket, and r/w to that bucket06:36
hazmatread access to juju-dist would be a separate issue06:37
hazmater... public read access06:37
rogpeppeoh, ffs, s3cmd doesn't look at the usual env vars06:37
wallyworldhazmat: so you can fix the permissions?06:38
wallyworldi wonder how they changed06:38
hazmatwallyworld, well first i want to verify if its an actual issue06:38
wallyworldjuju doesn't work06:38
hazmatwallyworld, it works for me..06:38
wallyworldnot for me or rogpeppe :-( or davecheney06:38
hazmati'm in the middle of a class, and we've been deploying units and services in different envs this whole time06:39
hazmatthose envs were already bootstrapped though.. but the tool logic is pretty constant on pinging the pub bucket06:39
wallyworldonce bootstrap happens i don't think it uses the pub bucket anymore06:39
hazmatwallyworld, hmm.. ic what you mean06:40
hazmatwallyworld, checking06:40
hazmatwallyworld, you can still download tools from the bucket.. http://juju-dist.s3.amazonaws.com/tools/juju-1.13.2-saucy-i386.tgz06:40
davecheneyahh, but we cannot list the bucket ...06:40
hazmatfor example06:40
wallyworldhazmat: yes06:41
davecheneyi woner if we've passed some limit06:41
hazmatwe should be using an explicit index anyways..06:41
davecheneyie, the number of rows to return ?06:41
hazmatie. a json/yaml file we create..06:41
davecheneymaybe it isn't 20006:41
hazmatrelying on aws default output isn't sane06:41
rogpeppehazmat: you can't rely on list-bucket?06:42
* hazmat looks into the bucket perms06:42
davecheneyhazmat: i guessed at the limit06:42
hazmatrogpeppe, juju does not do list-bucket06:42
hazmatrogpeppe, we don't require every use to have an aws account06:42
hazmatevery user.06:42
wallyworldjuju does list buvket06:43
rogpeppehazmat: list-bucket is just a GET of the bucket, no?06:43
wallyworldwhen it looks for the tools06:43
hazmatnot quite06:43
hazmatrogpeppe, yes.. but its a bit more involved for pagination, etc06:43
* wallyworld has to run away for a bit. back a bit later06:43
rogpeppehazmat: agreed. but we do that.06:43
rogpeppehazmat: is there a problem with doing that?06:43
rogpeppehazmat: there are definite advantages to relying on ec2 itself to provide data about what files are uploaded06:44
hazmatrogpeppe, wallyworld, davecheney fixed06:45
rogpeppehazmat: in particular it means that you can concurrently upload tools without worrying about maintaining a shared "directory" file (something i'm not is possible in s3)06:45
hazmathttp://juju-dist.s3.amazonaws.com06:45
rogpeppehazmat: yup, that's better06:45
rogpeppehazmat: what was the problem?06:45
hazmatrogpeppe, we control the upload process, we should define our format06:45
hazmatrogpeppe, the bucket perms had changed to not be public06:45
rogpeppehazmat: how tf did that happen?06:45
hazmatrogpeppe, and updating an index file allows us the flexibility to host tools elsewhere06:46
rogpeppehazmat: we're moving towards using "simple" streams for all of this06:46
hazmatrelying on aws .. means they change something and we break.. and we have multiple impls if we have to go somewhere else06:46
hazmatrogpeppe, good06:46
hazmatrogpeppe, wallyworld, davecheney if you want the gory xml details.. here's previous and current of the bucket acl.. http://pastebin.ubuntu.com/6131371/06:48
hazmatsomething reset the bucket perms..06:49
hazmatit seems like06:49
hazmatto only allow owner access.. anyways.. its fixed afaics, heading back to class06:50
rogpeppehazmat: yeah06:50
rogpeppehazmat: thanks06:50
hazmatnp06:50
rogpeppefwereade, dimitern, mgz, axw: small review anyone? https://codereview.appspot.com/13806043/08:55
axwrogpeppe: looking08:56
rogpeppeaxw: thanks!08:56
=== allenap` is now known as allenap
wallyworldi have a branch for review for the 1.15 release. tim started it but got stuck when ec2 failed earlier. i took his work and added a test and here it is https://codereview.appspot.com/1364204510:04
mgzwallyworld: looking10:06
wallyworldthanks :-)10:06
natefinchIs it just me, or do the menus on www.ubuntu.com look totlaly messed up?10:29
rogpeppenatefinch: me too10:39
rogpeppeanother trivial review: https://codereview.appspot.com/13780044/10:40
natefinchrogpeppe: I'll take a look10:40
rogpeppenatefinch: thanks10:40
natefinchrogpeppe: done10:42
rogpeppenatefinch: ta!10:42
natefinchrogpeppe: who should I poke about www.ubuntu.com?10:43
rogpeppenatefinch: i have no idea, sorry10:43
natefinchheh no problem10:43
dimiternnatefinch, fwereade, rogpeppe, wallyworld: standup?10:46
dimiternrogpeppe, fwereade : a quick review after standup? https://codereview.appspot.com/13636044/10:54
fwereaderogpeppe, wallyworld, mgz, dimitern, natefinch: sorry, I'm trying to get back in11:06
fwereadegents, my audio is appallingly bad, I think we stick with "don't do that then" and hope that everybody can remember that until we have simplestreams online11:19
* fwereade pops out for coffee12:16
* rogpeppe just got bitten by this bug. it really confused me! https://bugs.launchpad.net/gocheck/+bug/122810512:18
_mup_Bug #1228105: panic in TearDownTest obscures original test panic <Gocheck:New> <https://launchpad.net/bugs/1228105>12:18
=== natefinch is now known as natefinch-afk
sinzuinatefinch-afk, does https://launchpad.net/juju-core/+milestone/1.14.1 need a bug fix to address the missing azure documentation?13:26
rogpeppetrivial review anyone? https://codereview.appspot.com/1377504613:35
rogpeppetook a while to diagnose that!13:35
rogpeppefwereade, dimitern, mgz: ^13:36
fwereaderogpeppe, looking13:37
mgzmetoo13:38
rogpeppefwereade: oh it would be so good to sort out JujuConnSuite. i think we're heading in the right direction with the configstore interface actually13:38
mgzugh, this is all so fragile13:39
rogpeppemgz: yeah13:39
fwereaderogpeppe, LGTM13:40
rogpeppemgz: most of it's actually unnecessary in fact13:40
rogpeppefwereade: thanks13:40
dimiternstill waiting for a review on https://codereview.appspot.com/13636044/ please13:41
dimiternrogpeppe, fwereade ^^ ?13:41
rogpeppefwereade: i kind of agree about nestable fixtures, but i'm not really sure why this CL implies the need13:41
rogpeppedimitern: looking, sorry13:42
fwereaderogpeppe, it looked like we were half-assing a fixture with sub-Test-method setup/teardown effectively13:43
rogpeppefwereade: that actually works ok in general - it's not really nesting as such, more splitting.13:44
rogpeppefwereade: i wish fixtures looked less like test suites though13:44
rogpeppedimitern: can't a machine tell its container type just by looking at its id?13:47
dimiternrogpeppe, I prefer to have an explicit method13:51
rogpeppedimitern: why do we need another API round trip for this?13:51
dimiternrogpeppe, it's still the same - no more roundtrips13:52
rogpeppedimitern: oh, i guess that's true. but still, i'm not entirely convinced it's worth denormalising here.13:53
dimiternrogpeppe, denormalising what?13:54
rogpeppedimitern: storing the container type redundantly to the id which directly implies it13:55
dimiternrogpeppe, then why do we have a ContainerType method on machine, or even a doc field?13:56
rogpeppedimitern: hmm, good point13:56
rogpeppedimitern: about the doc field anyway13:56
dimiternrogpeppe, the fact the machine id implies the container type is a side-effect, I think, does not merit doing tricky magical stuff in the api, just because we know its format13:57
wallyworldfwereade: i have a small mp to fix an issue with hp cloud that i found while smoke testing the 1.15 release https://codereview.appspot.com/1378904513:58
rogpeppedimitern: you may be right there. tbh, i think that it's pretty dodgy that the agent needs to know its container type - it looks pretty hacky to me13:58
rogpeppedimitern: it would be more regular if lxc machines had an lxc job, or something like that13:58
dimiternrogpeppe, perhaps, but we're not there yet13:59
fwereadewallyworld, ha, Isee, good catch13:59
rogpeppedimitern: fair enough13:59
fwereaderogpeppe, agreed and understood hacky, but out of scope for now ;)13:59
rogpeppedimitern: reviewed14:00
dimiternrogpeppe, ta!14:01
fwereadewallyworld, LGTM14:09
wallyworldfwereade: great thanks, landing now14:09
fwereadewallyworld, there's some code commented out you might want to remove first14:11
wallyworldfwereade: yeah, already done :-)14:11
wallyworldit's late, i'm tired, stupid typo14:11
fwereadewallyworld, I know the feeling, get some rest :)14:14
wallyworldsoon, will just land the branch14:15
motterout of curiosity, does all of juju's logging code write to syslog?14:16
motter(i'm looking at https://bazaar.launchpad.net/~go-bot/juju-core/trunk/files/head:/log/)14:16
mgzmotter: we've been moving that way14:17
mgzfor the agents that run on remote machines at least14:18
motterthe loggo package does levels, is that configured to write to syslog too?14:18
mgzyes, there should be a correspondance14:19
=== tasdomas is now known as tasdomas_afk
mottermgz: thanks14:20
motterwe're also using syslog, but i'm curious as to what other projects are doing14:20
=== BradCrittenden is now known as bac
sinzuiI need help setting up a reliable test env/config. I always get failures running trunk and 1.14. I doubt the suite is really broken.14:58
rogpeppesinzui: what failures are you seeing?14:59
sinzuirogpeppe, http://pastebin.ubuntu.com/6133051/15:00
rogpeppesinzui: what version of mongod have you got installed?15:00
sinzuidb version v2.4.615:01
rogpeppesinzui: could you paste the output of mongod --help, please?15:01
sinzuirogpeppe, http://pastebin.ubuntu.com/6133061/15:02
rogpeppesinzui: hmm, that looks ok. although i'm actually using 2.2, so it's possible there's a problem with later versions.15:03
rogpeppesinzui: i think it's *probably* a mongod problem15:03
sinzuirogpeppe, I have mongod running on my system by default. I don't think I have messed with it (I build my apps in containers)15:04
rogpeppesinzui: what version of ubuntu are you running?15:04
sinzuisaucy15:05
rogpeppesinzui: hmm, i don't know what the status of juju on saucy is currently. mgz?15:08
rogpeppesinzui: the version of mongo i get with "sudo apt-get install mongodb-server" is v2.2.4, which is earlier than yours15:08
mgzshould e working...15:09
rogpeppemgz: do you know anything about what versions of mongo juju currently works with?15:09
mgzyoumean it's not all of them? :)15:09
rogpeppemgz: well, it looks like it fails with 2.4.615:09
sinzuiI have a precise container. I can try running the suite from there15:10
rogpeppesinzui: i'm running raring and it works ok there15:10
rogpeppesinzui: you could try grabbing the mongod binary from juju-dist.s3.amazon.com15:11
rogpeppesinzui: are you running the standard mongod installed with saucy? (i'm presuming so)15:11
sinzuiI am15:12
sinzuiand I have deployed many times with it. I see no signs of breakage15:12
rogpeppeanyone else here running saucy?15:13
rogpeppesinzui: i think i see a possible fix - looks like mongo have changed their error messages15:17
rogpeppesinzui: could you try editing testing/mgo.go for me please15:17
sinzuione moment, I am running the tests in a precise container15:18
rogpeppesinzui: and change the isUnauthorized function to this: http://paste.ubuntu.com/6133129/15:19
sinzuirogpeppe, I have the file open, I will make the change when the suite completes15:19
rogpeppesinzui: that might *possibly* fix your problem15:19
sinzuiThe test run is taking forever in this container. but it is also passing where it failed for saucy15:23
sinzuithank you rogpeppe, the patch is better, only one failure: http://pastebin.ubuntu.com/6133177/15:31
sinzuiThe tests passed in my precise container. I am unblocked for now, and can test more patches if you like15:31
rogpeppesinzui: ah, that looks like exactly the same issue in a different guise15:32
rogpeppesinzui: could you try one more thing for me, please15:35
rogpeppe?15:35
rogpeppesinzui: if you could try another version of isUnauthorized and paste me what the panic prints, that would be very useful: http://paste.ubuntu.com/6133195/15:35
rogpeppesinzui: then i'll be able to submit an appropriate patch to trunk15:35
* sinzui runs tests15:36
sinzuisorry rogpeppe , quite errors this time: http://paste.ubuntu.com/6133222/15:41
rogpeppesinzui: thanks; line 31 is all i need15:42
* rogpeppe wishes mongo could sort out its error codes15:42
mgzfiled bug 1228255 for my live test woes earlier16:31
_mup_Bug #1228255: Live bootstrap tests fail on canonistack <juju-core:Confirmed> <https://launchpad.net/bugs/1228255>16:31
rogpeppesinzui: ping16:34
sinzuihi rogpeppe16:35
rogpeppesinzui: could you pull this branch and try testing it, please. i hope it should fix your problem: lp~rogpeppe/juju-core/409-fix-for-mongo.2.416:35
* sinzui does16:35
sinzuirogpeppe, I have one failure. I don't think it is related to your work. I think I have the deps wrong when I switched back to trunk: http://paste.ubuntu.com/6133503/16:49
rogpeppesinzui: could you paste the whole of that failure log please?16:51
rogpeppesinzui: (no, i don't think it's related either)16:51
sinzuigoose and mgo had higher revnos. I just reverted to what the dependencies.tsv listed16:51
rogpeppesinzui: try this (in the juju-core root directory): go get launchpad.net/godeps; godeps -u *.tsv16:51
rogpeppesinzui: ah, or manually, sure16:52
sinzuirogpeppe, I did. it warned that the branches were on the wrong version. I assumed it would correct that16:52
sinzuithis is the full log: http://paste.ubuntu.com/6133515/16:53
* sinzui runs tests again16:53
rogpeppesinzui: you can run that test alone to get a quicker turnaround time: go test launchpad.net/juju-core/cmd/juju -gocheck.f TestAutoUploadAfterFailedSync17:04
rogpeppesinzui: i'd like to know if it fails consistently in saucy17:04
sinzuisorry rogpeppe. I don't know how to clean goose and v2/mgo. godeps says the are not clean when I run clean on them17:05
rogpeppesinzui: if in doubt, remove them and go get them again.17:05
rogpeppesinzui: e.g. rm -r $GOPATH/src/launchpad.net/goose17:05
rogpeppesinzui: rm -r $GOPATH/src/labix.org/v2/mgo17:06
rogpeppesinzui: go get launchpad.net/goose labix.org/v2/mgo17:06
rogpeppesinzui: if you want to know *why* they're considered unclean, go into the directory and run "bzr status"17:07
sinzuigoody. I will try that next time. I am getting the code again17:07
sinzuirogpeppe, http://paste.ubuntu.com/6133582/ shows the failure again17:09
rogpeppesinzui: ok, that's useful thanks17:09
rogpeppemgz, fwereade, dimitern: saucy fix CL review? https://codereview.appspot.com/1380804317:11
mgzlooking17:13
rogpeppefwereade: ping17:24
mgzrogpeppe: reviewed17:26
rogpeppemgz: we're using standard base64 (as witness the StdEncoding below.17:27
mgzso, that change shouldn't be needed than?17:27
rogpeppemgz: but i'm actually inclined to use '[^']*'17:27
rogpeppemgz: why not? standard encoding includes + and /, no?17:28
rogpeppemgz: (neither of which were in the original regexp)17:28
mgzah, yes, I misread17:32
mgzthe change is correct17:32
natefinch-afksmoser: have a minute?17:37
smoserhere17:37
=== natefinch-afk is now known as natefinch
natefinchsmoser: Having difficulty ssh'ing into the maas virtual machines (it worked yesterday and then today somehow it's not... probably user error)17:37
natefinchsmoser: I recreated the machins with 1024M of RAM, since we'd seen some problems lately bootstrapping with 51217:38
smoserk17:38
natefinchsmoser: maybe I messed something up... what's weird is that juju bootstrap seems to run ok (at least from what I see locally), but then I try to ssh insto the machine, and it says it can't find a machine with that name17:39
smosernatefinch, its the hack that i told you about.17:39
smoseryou have to fix /etc/resolv.conf now. i'll do it.17:39
natefinchnatefinch: I figured there was something I'd forgotten17:39
smoserwhat happened that changed is that your dhclient lease came up17:39
smoserand it rewrote /etc/network/interfaces17:39
smoseri dont have a real solution for this at the moement.17:40
natefinchahh ok no problem17:40
natefinchsmoser: I'll leave myself a note so I don't forget next time17:41
smosernatefinch, i think its fixed now17:42
smoserbasically just re-wrote /etc/resolv.conf but to have '127.0.0.1' as the nameserver17:43
smosernow , 'host inst-001.master' works17:43
natefinchsmoser: yep, saw it.  That did it. Thanks.  Wrote myself a note to remind me to do that if this happens again.17:43
smoseri also just added 'search master'17:43
natefinchCool17:44
smoseri'd like to have a real fix for that, but ... time.17:44
natefinchyuuup17:44
smoseron canonistack instances the dhclient lease is 2 minutes.17:44
smoserso you get to re-apply every 2 minutes :)17:44
natefinchOMG17:44
natefinchcrazy17:45
rogpeppesinzui: i've managed to reproduce your test failure locally, and the fix is straightforward, i think. i reported the bug here: https://bugs.launchpad.net/juju-core/+bug/122831518:14
_mup_Bug #1228315: cmd/juju test failure under saucy <juju-core:New> <https://launchpad.net/bugs/1228315>18:14
sinzuithank you very much rogpeppe18:15
rogpeppesinzui: the mongo fix is now in trunk, BTW18:15
sinzuifab18:15
rogpeppenatefinch, fwereade, dimitern, wallyworld: a simple review? after this, saucy tests should run ok, theoretically. https://codereview.appspot.com/1381104318:51
natefinchrogpeppe: I can take it18:51
rogpeppenatefinch: thanks18:52
rogpeppesinzui: perhaps you could try out lp:~rogpeppe/juju-core/410-only-one-version and verify that it fixes the problem?18:53
* sinzui does18:53
rogpeppesinzui: ta18:54
natefinchrogpeppe: I like the change, but it doesn't look like it would change the behavior unless /etc/lsb-release changes while you're running.  What is it fixing, other than reparsing the same data over and over?18:56
sinzuirogpeppe, saucy loves your branch.19:02
rogpeppesinzui: good to hear!19:06
rogpeppenatefinch: it fixes tests that patch version.Current19:07
natefinchrogpeppe: aahhhh yep19:07
hatchwhen newbies follow the lxc guide and run `juju boostrap -e local` then follow up with a `juju status` the error message is that it can't connect and to check your credentials19:10
hatchthis will be very confusing19:10
hatchI am guessing there is no way to know that the bootstrap node is still starting up19:11
hatchbut maybe a 'your bootstrap node might still be starting' message could be added?19:11
natefinchhatch:  a friendlier error message might be all we really need19:28
hatchnatefinch: that's what I was thinking - any 'real' sollution is probably a little too fragile19:29
hatchshall I file a bug ?19:29
natefinchhatch: go for it19:33
jamespagesinzui, 1.14.1 all done - should build in the stable PPA shortly21:11
sinzuiwow. Thank you for your dedication jamespage21:12
jamespagesinzui, np21:12
hatchawesome!21:48

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