/srv/irclogs.ubuntu.com/2014/10/28/#juju-dev.txt

cmarscontrib Casey Marshall00:13
cmars(that was for mup_)00:13
wallyworldthumper: i disagree that bug 1385289 should block landings. stuff isn't broken, it's slower00:18
mup_Bug #1385289: local storage migration is very slow <ci> <regression> <test-failure> <upgrade-juju> <juju-core:Triaged> <https://launchpad.net/bugs/1385289>00:18
wallyworldit should be critical for sure00:18
davecheney$ go version00:18
davecheneygo version devel +211c346906d0 Tue Oct 28 09:56:33 2014 +1100 linux/power64le00:18
thumperwallyworld: agreed00:18
davecheney^ real go, on power00:18
wallyworld\o/00:18
thumpernice00:19
davecheneythumper: fixed a few bugs there today00:19
thumper\o/00:19
davecheneyincluding some stuff that makes me wonder how the guys at google were testing this00:20
davecheneyhttps://code.google.com/p/go/source/detail?r=cf3c9803a3d2116a9621fb1e9b2226a8fa3d5e1000:20
* thumper -> physio00:20
davecheneyfor example00:20
waiganimenn0: you about?00:32
menn0waigani: yep00:33
waiganimenn0: i've hit a tricky one, can we talk it over?00:33
waiganimenn0: standup channel?00:33
menn0yep and yep00:33
perrito666sorry did you say that the critical bug was fixed?00:45
waiganimenn0: should we add a check for the EnvUUID as a generic test for all the upgrade steps?01:25
menn0waigani: it already does :)01:26
menn0see checkAddEnvUUIDToCollection01:26
waiganimenn0: ? how did the tests pass then?01:27
menn0that checks that document has the env-uuid field01:27
menn0but it doesn't check that the struct has it added01:27
menn0there's no generic way to do that01:27
menn0well... I maybe there is but it won't be pretty01:27
waiganireflection and assuming the name is EnvUUID?01:28
waiganiyeah, it would be ugly01:28
waiganiI'm happy to let it go01:28
menn0I think the best bet is to check in the test function after checkAddEnvUUIDToCollection has done its thing01:28
menn0so for instanceData, check EnvUUID where MachineId and Status are checked01:29
rick_h_wallyworld_: I'm going to check with frankban on adding this as a new bug but fyi on https://bugs.launchpad.net/juju-core/+bug/1336843 as it's a fix in alpha2 that has a follow up effect.01:31
mup_Bug #1336843: bootstrap without a jenv destroys an existing environment  <bootstrap> <juju-core:Fix Released by cox-katherine-e> <https://launchpad.net/bugs/1336843>01:31
wallyworld_rick_h_: no worries, file a bug if needed and katco can look into it01:32
rick_h_wallyworld_: cool01:32
wallyworld_rick_h_: if there's an issue, we'll fix asap01:33
wallyworld_thanks for letting us know01:33
davecheneythumper: $ juju version01:37
wallyworld_axw: long story wrt SyncBuiltTools, maybe my explanation was crap. i took out the stream param and wrote to a "custom" directory/stream when building the tools but it failed and i traced through the code to figure out what was happening. but the logic that syncs expects the tools in the source datasource to be in the same stream as the destination, since it needs to read from the public cloud that way. hard to explain,01:37
wallyworld_happy to discuss if needed01:37
davecheney1.21-alpha3-trusty-power64le01:37
thumperdavecheney: awesome to see01:44
waiganimenn0: http://reviews.vapour.ws/r/269/01:45
* menn0 looks01:45
davecheneyok, now back to what I was supposed to be doing01:46
axwwallyworld_: SyncBuiltTools needs a stream param; I was just saying --upload-tools should set an obviously arbitrary value01:47
wallyworld_axw: trouble is that an arbitrary value breaks the processing, but i'll take another look to be sure01:48
menn0waigani: done01:48
waiganisweet, thanks menno01:48
wallyworld_since upload-tools calls a method which copies from source to dest, and that copy needs to have same stream for source and dest, not an arbitrary one for source01:48
axwwallyworld_: upload-tools does not do that anymore.01:49
axwlook at uploadTools in cmd/juju/upgradejuju.go; it opens the file and passes it to UploadTools01:49
wallyworld_some infrastructure did in the tests01:49
wallyworld_i'll take another look to be sure,but a whole bunch of sync tools tests failed01:50
wallyworld_if the source stream != dest stream01:50
wallyworld_and sync tools calls that same shared function as upload tools ses from emory01:51
waiganimenn0: testing for EnvUUID field, we currently are not doing it for any of the collections. Shouldn't be for all or none?01:51
menn0waigani: probably01:51
menn0waigani: your call01:52
menn0waigani: the most important thing I guess is that the field is there in the DB, which we are testing01:52
menn0waigani: it's trivial to add the field to the struct later if we've forgotten01:52
waiganimenn0: yep, I'll leave it as is01:52
axwwallyworld_: http://paste.ubuntu.com/8712745/    <- all tests pass in cmd/juju with this, and upgrade-juju works as expected. haven't tested bootstrap, but should be the same.01:54
wallyworld_axw: trouble is, when i made that change, most tests in environs.sync fail01:55
axw?? how could a change in cmd/juju affect environs/sync?01:55
wallyworld_oh, i see what your saying01:55
wallyworld_you're01:55
wallyworld_i had experimented with not having the streams parameter at all01:56
wallyworld_and just using "custom" all the time01:56
axwright, the param is needed for *other* cases, just no in this and the bootstrap case01:56
axwnot*01:56
wallyworld_right ok, fair point. sorry i misunderstood you01:56
axwnps01:56
wallyworld_i'll change that when i do the index file fix01:57
axwit's a fairly minor thing, just wanted to make sure we both were on the same page01:57
wallyworld_yeah, np. thanks for pushing the issue01:57
waiganimenn0: the instanceData upgrade step. To add a post-env UUID, based on the machine _id have to either: assume machine _id is pre-env and pass it through st.docID() (in which case, the upgrade step has to come before machine upgrade), or assume the machine _id is post-env (in which case, the upgrade step has to come after the machine upgrade)02:22
waiganimenn0: i.e. I don't see how we can make the step not care about order02:22
menn0waigani: or you could detect if the machine is pre or post env UUID upgrade02:23
menn0waigani: by checking the env-uuid field02:23
waiganimenn0: or that02:24
waiganimenn0: okay, will do02:24
menn0waigani: if that seems too messy, then pick and ordering and add a comment in the list of steps explaining that the ordering is important and why02:24
waiganimenn0: I didn't suggest it as it seemed a bit messy, I'll see how bad it looks and fall back to a comment02:26
waiganimenn0: I also forgot about just checking the EnvUUID - thought we'd have to regex to _id02:26
menn0waigani: cool. yeah, just see how it looks.02:27
axwwallyworld_: what's the simplest way to create tools+metadata from source?02:32
axwI only want the built tools in the metadata I create02:32
wallyworld_axw: there's no easy way as such. if you have a tarball, you can easily generate metadata, but as far as i know, there's no easy way to get a tarball from source02:35
axwwallyworld_: that's what I though - thanks02:35
wallyworld_axw: what are you trying to do?02:35
axwjust want to live test the upgrade bug02:36
axwI've got a unit test02:36
axwI'll bootstrap local and pull the tarball out of there02:36
wallyworld_yep, that's what i do02:36
wallyworld_maybe we need a maketarball plugin02:36
waiganimenn0: so we search for instanceData docs with the machineID - pre or post env depending on whether or not the instanceData upgrade step has been run02:39
waiganimenn0: I'm thinking order just matters: migrate 1. instanceid, 2. machines envUUID 3. instanceData envUUID02:40
menn0menn0: that's probably less messy isn't it?02:41
menn0menn0: just write a comment explaining it and be done with it02:41
waiganimenn0: sgtm!02:42
waiganimenn0: http://reviews.vapour.ws/r/177/02:53
waiganimenn0: http://reviews.vapour.ws/r/269/03:06
* menn0 is looking03:12
menn0waigani: done03:16
waiganimenn0: cheers03:17
waiganithumper: http://reviews.vapour.ws/r/269/03:17
waiganithumper: http://reviews.vapour.ws/r/177/03:17
thumperack03:18
wallyworld_thumper: remind me, we're not meant to run apt-get update for local provider are we? i thought we'd check for dependencies like cpu-checker and complain if they weren't there, but not just install shit onto a user's host http://pastebin.ubuntu.com/8714033/03:58
davecheneythumper: menn0 https://github.com/juju/juju/pull/97604:03
davecheneyi haven't created an rbt review for this yet04:03
davecheneyi want to get a sense for how gross you think this is04:03
davecheney(I think it's 50% gross)04:03
=== mup_ is now known as mup
thumperdavecheney: rbt gets magically created for you now04:03
thumperwallyworld_: yes we are supposed to run update by default for local04:04
thumperwallyworld_: we should be able to configure it off for those who know what they are doing04:04
wallyworld_thumper: ok, ta04:04
thumperand have done an update recently themselves04:04
wallyworld_thumper: that was on my hst04:04
wallyworld_host04:04
wallyworld_not the lxc container04:04
thumperoh...04:04
thumperno, we install shit04:05
wallyworld_that log from when i bootstrapped04:05
thumperwhich is just another reason to fix the local provider04:05
wallyworld_yeah :-(04:05
thumperso that machine zero is a container too04:05
wallyworld_i thought we'd check but not install04:05
thumperI don't remember that04:05
wallyworld_maybe i'm making it up04:05
wallyworld_jam: i have an lxc question - you suggested in a bug comment setting a http/apt/whatever proxy to http://10.0.3.1:8000 instead of http://localhost:8000 where 10.0.3.1 is the lxcbr0 bridging address. i have set up a little http server on the host and can wget to it fine from the host, but inside the lxc container, wget can't get out due to connection refused. could be a firewall issue between container and host? do you04:20
wallyworld_have any suggestion?04:20
jamwallyworld_: so you put an http server on your host machine, and then are unable to access 10.0.3.1:PORT from lxc, is that correct?04:21
jamwallyworld_: did you start your HTTP server before starting LXC?04:21
jamIs it binding to "*" (0.0.0.0) or is it binding to 127.* or to an explicit IP address?04:22
wallyworld_jam: yes, but can access localhost:PORT from host04:22
jamwallyworld_: from the *host* can you access 10.0.3.1:PORT04:22
wallyworld_yes04:22
wallyworld_from my PC04:22
wallyworld_oh04:22
wallyworld_sec, let me check04:22
jamwallyworld_: my first thought is that your service isn't bound to all networks04:23
wallyworld_jam: yeah, can't connect to 10.0.3.1 from host. i was just trying to validate the replacing localhost with the bridge address actualy worked04:23
wallyworld_maybe there's no need to do that04:24
jamwallyworld_: so are you just doing something like python SimpleHTTPServer ?04:24
wallyworld_the http proxy, apt proxy etc are all set to <bridge addr> as required04:24
jamAre you handing it an address to bind to ?04:24
jamRather than "0.0.0.0" ?04:24
jamwallyworld_: it might certainly be a real life issue, as having your HTTP proxy set up may not bind it to the LXC bridge address if it comes up first.04:25
wallyworld_jam: ah, i see the issue, the little go prog is binding to localhost04:25
wallyworld_http.ListenAndServe("localhost:8000", nil)04:25
jamwallyworld_: right, I think you can change that to ":8000' or maybe "0.0.0.0:8000"04:25
wallyworld_jam: i assumed that would cover all connections into the physical host04:26
jamwallyworld_: nope, specifying localhost restricts it to 127.* as it is how you avoid making something publi04:26
wallyworld_oh, i'll try that, ty :-)04:26
wallyworld_jam: you are awesome, that worked, ty04:27
jamwallyworld_: np04:28
menn0davecheney: looking now04:38
menn0davecheney: done04:58
stokachuanyone notice when doing a juju bootstrap with a maas provider that when it runs its apt-get update/upgrade it returns a non zero status?05:13
stokachueven though its chugging right along05:13
axwstokachu: apt-get update/upgrade fails, but bootstrap continues on? what version of juju is that?05:24
stokachu1.20.1005:24
stokachuthe upgrade doesn't fail though05:24
axwcan't say I've seen that. we do "set -e" ... can't see any obvious reason why that'd happen05:27
stokachuit returns the same 124 status code05:27
davecheneystokachu: exit 124 is timeout05:29
davecheneythat is coming from deep the bowels of apt05:29
stokachuok05:30
davecheneystokachu: did you paste the bootstrap log ?05:30
stokachunah im doing another deploy05:30
stokachuill paste once it fails again05:31
davecheneykk05:37
davecheneyThe following packages have unmet dependencies: gdb : Depends: libpython3.4 (>= 3.4~b1) but it is not going to be installed06:14
davecheneykill me06:14
tasdomasmorning06:48
tasdomasashipika was trying to land his PR and got this response from the buildbot: http://juju-ci.vapour.ws:8080/job/github-merge-juju/1029/console06:49
tasdomasany ideas as to what may be causing this?06:49
ashipikaand this one is interesting, as well http://juju-ci.vapour.ws:8080/job/github-merge-juju/1028/console06:51
davecheneymy tasdomas mongodb shat itself on startup06:59
davecheneythat is why it cannot complete the TLS handshake07:00
tasdomasdavecheney, should we retry the merge?07:01
davecheneyyes07:01
davecheneyit'll pass eventually07:01
tasdomasdavecheney, ok - ashipika's first attempt failed when godeps failed to retrieve launchpad.net/tomb07:02
davecheneyyup07:02
davecheneyin that case07:02
davecheneylaunchpad shat itself07:02
ashipika:D07:02
ashipikashit is definitely transitive..07:02
davecheneyashipika: we checkout all the build deps on each build07:02
davecheneymakes the build suceptable to other systems that are weakly available07:03
ashipikadavecheney: no problem, should i just keep retrying them until the build succeeds?07:03
davecheneyashipika: yup07:03
davecheneyboth of those failures are transiant07:03
ashipikadavecheney: rogerroger07:03
davecheneysorry for the inconveneince07:03
ashipikadavecheney: no, no, i completely  understand.. thanks for the info07:04
davecheneyashipika: np07:05
dimiternmorning all07:26
jammorning dimitern07:30
dimiternhey, jam!07:31
=== ashipika is now known as ashipika_afk
mattywmorning all07:53
mattywdimitern, morning07:53
TheMuemorning08:02
voidspacemorning all08:26
jamdimitern: I'm not sure if you saw the calendar update, but we moved the standup time by 45 minutes to see if it would work better for everyone (you may not be back yet, I guess)09:58
jamTheMue: voidspace: just a poke that it is soon since I know the calendar is changed09:59
* fwereade out for a bit09:59
jamhey fwereade09:59
jamhave fun :)09:59
voidspacejam: omw09:59
voidspacejam: hmmm... firefox is in a redirect loop10:00
voidspacerestarting browser time...10:00
jamvoidspace: sounds exciting10:00
jamwhere will you end up next ? :)10:00
voidspaceoh it is10:00
voidspacehah10:00
voidspacealways an adventure10:00
dimiternvoidspace, jam, TheMue, hey guys, I've just came back - omw for the standup10:05
voidspacedimitern: hey10:05
perrito666morning10:06
voidspaceperrito666: morning10:06
mattywdimitern, after the standup could you take a look at this? http://reviews.vapour.ws/r/265/10:09
dimiternmattyw, sure10:11
mattywdimitern, thanks very much - it's the fix for the add-unit bug you found last week10:12
dimiternmattyw, sweet!10:15
perrito666is anyone using exuberant-ctags?10:40
mattywperrito666, I am10:46
perrito666mattyw: and that with vi?10:47
mattywperrito666, of course (are there other editors ;) )10:47
perrito666mattyw: I am getting an empty tagbar with go10:47
perrito666yet, if I ctags -x on a go file I get results10:47
mattywperrito666, you use tagbar? wow - not used that in a loooong time.10:48
mattywperrito666, which ctags are you using for go?10:48
perrito666mattyw: care to share what are you using?10:48
perrito6665.9-svn10:48
mattywperrito666, https://github.com/jstemmer/gotags10:49
perrito666mattyw: I am a bit wary of using go get as a package manager, since I tend to nuke my /bin in go10:51
mattywperrito666, I don't update it enough to worry about package management - just build it then move it somewhere you don't nuke10:52
mattywperrito666, do you use godef?10:52
mattywperrito666, and by extension vim-godef?10:52
perrito666mattyw: I dont, what is it?10:52
mattywperrito666, https://github.com/dgryski/vim-godef10:54
perrito666sweeeet10:54
wwitzel3fwereade: so I decided to go the explicit route and add a --no-remote-unit option13:09
fwereadewwitzel3, ok, that sounds sane13:10
fwereadewwitzel3, what do we do if nothing's specified and no units are in the relation? require --no-remote-unit, or just execute without one?13:11
perrito666ericsnow: ping13:14
wwitzel3fwereade: my thought was require --no-remote-unit .. easiest is to just ask what they'd prefer as the desired behavior and make sure it is documented accordingly13:14
perrito666rogpeppe: your timing is not the best :p13:18
rogpeppeperrito666: i'm sorry - i only just saw it13:19
rogpeppeperrito666: it shouldn't be hard to change though, should it?13:20
perrito666nope, its just that I have to revert it :p it would be nice to have a $$revert$$13:20
rogpeppeperrito666: ha, yes13:20
rogpeppeperrito666: git makes it very easy to do that though13:21
perrito666rogpeppe: It does since the bot treats it as a merge13:24
perrito666rogpeppe: Ill make sure that you did not miss anything before making the change but you might be right13:25
rogpeppeperrito666: thanks13:25
rogpeppeperrito666: i don't like seeing nice small APIs made bigger because of testing issues :)13:26
perrito666rogpeppe: I like it no more than having smart dummy objects :p It was a fast trade off, anyway going for your way now BUT, you will have to review it :p13:28
rogpeppeperrito666: np13:28
fwereadewwitzel3, I'm inclined to think that not having a remote unit is actually perfectly valid13:37
fwereadewwitzel3, and I'm also starting to feel sure that specifying a remote unit that appears not to exist *may* be valid13:37
fwereadewwitzel3, so it shouldn't be an automatic error13:38
fwereadewwitzel3, but it should probably be some sort of warning..?13:38
fwereadesorry bbiab13:39
wwitzel3fwereade: ok, that sounds fine to me13:39
voidspacerebooting13:46
voidspacedimitern: ping14:03
ericsnowdimitern: could you take a look at the patches I have up?14:18
ericsnowdimitern: the most important are http://reviews.vapour.ws/r/67/ and http://reviews.vapour.ws/r/135/)14:18
ericsnowdimitern: however, I could really use eyes on http://reviews.vapour.ws/r/79/ and http://reviews.vapour.ws/r/126/ as well14:19
* perrito666 looks at the 47" tv and wonders how comfortable its as a work screen14:31
dimiternvoidspace, pong14:36
dimiternericsnow, sure, looking14:36
ericsnowdimitern: thanks :)14:36
dimiternI was on a call, sorry for responding slowly14:36
voidspacedimitern: so we have stub implementation of AllocateAddress for all providers14:37
voidspacedimitern: that needs to change14:37
voidspacedimitern: I've mostly done that14:37
dimiternvoidspace, great14:37
voidspacedimitern:  AllocateAddress(_ instance.Id, _ network.Id, _ network.Address) error14:37
voidspacedimitern: ?14:37
dimiternvoidspace, that sounds good to me14:38
voidspacedimitern: cool, there's an existing test (dummy provider) that needs to change too14:38
voidspacejust finishing that14:38
dimiternvoidspace, possibly with the addition of a concrete error type returned for "conflict" or "address already reserved" sort of error14:38
voidspacedimitern: I figured we could sort that out later...14:39
voidspacedimitern: as we also need to handle "request failed - couldn't reach server" errors too14:40
dimiternvoidspace, yeah, fair point -- just a reminder14:40
voidspacechanging error type to a concrete type is trivial14:40
dimiternvoidspace, "request failed", as in "i did 5 attempts so far and it failed all" ?14:41
voidspacedimitern: I guess14:41
voidspacewhatever strategy we pick14:41
voidspaceI did discuss that with John a bit yesterday but it morphed into a discussion about race conditions14:41
dimiternvoidspace, I think ec2 deserves some retrying logic, being occasionally flaky14:42
voidspacedimitern: yep, ideally we'd abstract the retry machinery so it can just be reused14:42
voidspacerather than ad-hoc retrying of all api calls14:43
dimiternvoidspace, there's *some* reusable bits already - around the AttemptStrategy type14:44
voidspaceright14:44
dimiternvoidspace, I even seem to recall a strategy that takes rate limiting into account for ec2, but I might be wrong14:44
voidspacedimitern: right, we did some rate limiting work14:45
voidspaceman, that was one of the first things I worked on14:45
dimiternvoidspace, yeah :) -- but wasn't that related to our apiserver only?14:46
* TheMue listens14:47
dimiternericsnow, hey14:48
voidspacedimitern: I don't recall the details...14:48
dimiternericsnow, so wrt http://reviews.vapour.ws/r/126/ - did you reach an agreement with davecheney about the approach ?14:48
voidspacedimitern: http://reviews.vapour.ws/r/274/14:49
dimiternvoidspace, cheers -- will have a look after I'm done with ericsnow's14:50
voidspacedimitern: cool, thanks14:50
TheMuevoidspace: AllocateAddress looks fine, implementation is nice short so far :D14:55
voidspaceTheMue: :-)14:55
voidspaceTheMue: yeah, done...14:56
TheMue+114:56
voidspaceTheMue: thanks14:56
ericsnowdimitern: wrt 126, while I addressed all his comments, I never heard back from Dave14:58
dimiternericsnow, right, I'd send him a mail and ask for a final look, considering his initial not lgtm, just to be on the nice side15:00
ericsnowdimitern: oh, that...we did sort that out.  I re-wrote it using direct HTTP requests rather than going through the websockets API15:02
dimiternericsnow, ah, ok -- it's fine then :)15:02
dimiternvoidspace, you've got a review15:38
dimiternericsnow, 1 review done, 3 to go15:38
ericsnowdimitern: rock on! :)15:38
voidspacedimitern: thanks15:41
voidspacedimitern: cool, I'll remove the TODO and merge15:41
dimiternvoidspace, +115:42
voidspacehow odd15:59
voidspaceinternet on my desktop just stopped working15:59
voidspacewas still working for everyone else in the house15:59
voidspaceanyway, a reboot fixed it15:59
perrito666 rogpeppe https://github.com/juju/juju/pull/97816:07
rogpeppeperrito666: one thing: please make sure that the call to MgoTestPackage is in peergrouper_test, not peergrouper16:09
rogpeppeperrito666: otherwise the tests in your workerJujuConnSuite will never be tested16:10
rogpeppeperrito666: (in Go 1.3 at any rate)16:11
rogpeppeperrito666: ah, i see that's already the case16:13
perrito666rogpeppe: it is there where already tests in peergrouper_test16:14
rogpeppeperrito666: BTW in the peergrouper version I've got, workerJujuConnSuite is already defined in peergrouper_test16:15
perrito666rogpeppe: ?16:15
rogpeppeperrito666: that was a week ago, commit 9301dcf0b882a1ab389297518a145b29668da51e16:16
rogpeppeperrito666: so i'd just revert the name to the same thing - worker_test.go16:16
rogpeppeperrito666: (there are other external tests too, such as initiate_test.go)16:17
rogpeppeperrito666: anyway, reviewed16:17
=== ChanServ changed the topic of #juju-dev to: https://juju.ubuntu.com | On-call reviewer: see calendar | Open critical bugs: 1385289 1386766
perrito666rogpeppe: I did not change the name, I have worker_external_test.go and worker_test.go is still there16:19
perrito666tx for the review btw16:19
rogpeppeperrito666: but worker_test.go is in peergrouper_test already, right?16:19
perrito666rogpeppe: no more, I also reverted that :)16:20
rogpeppeperrito666: ok16:20
perrito666my lunch was so diet that I actually took me less to cook it than a full test run16:22
perrito666rogpeppe: this patch was sponsored by my new diet :p you can thank my wife16:23
rogpeppeperrito666: i highly recommend the 5-2 diet BTW16:24
rogpeppeperrito666: means you only have to try hard on two days in a week16:25
perrito666rogpeppe: there is no spec on wp on how "normal" can be a normal day16:27
rogpeppeperrito666: whatever you like, pretty much16:27
perrito666rogpeppe: according to the dietitian my problem is not that I eat caloric things, its just that I eat a lot of it, for any definition of it16:28
rogpeppeperrito666: yeah, so the 5-2 diet is good for just that16:28
rogpeppeperrito666: you can still pig out :)16:28
rogpeppeperrito666: (all but those two days)16:28
perrito666:p16:28
* perrito666 gives it a shot16:28
perrito666uff, we are blocked.. again?16:29
rogpeppeperrito666: you should watch the original documentary16:29
rogpeppeperrito666: google for [horizon 5-2 diet documentary]16:30
sinzuinatefinch, can you arrange for a rollback to address bug 1386766? when it is removed, we can then test a fix the the upgrade regression that merged a few hours ago16:30
mupBug #1386766: generate-tools breaks scripted and documented behaviour <ci> <metadata> <regression> <juju-core:Triaged> <https://launchpad.net/bugs/1386766>16:30
TheMuedimitern: thx16:31
dimiternTheMue, np :)16:32
dimiternericsnow, all reviews done16:42
ericsnowdimitern: awesome! Thanks for doing that.16:42
dimiternericsnow, np :) thanks for doing all that work16:44
natefinchsinzui: ok17:00
voidspacedammit, dimitern has left17:02
voidspacehe always escapes before I can harrass him17:02
wwitzel3having this bag of pistachios by my desk was a mistake, I can't feel the tips of my fingers now17:13
voidspacewwitzel3: :-)17:13
voidspacewwitzel3: a bag of pistachios is never a mistake17:14
voidspacepretty much by definition17:14
perrito666wwitzel3: that sounds a lot like you are alergit to pistachios17:15
voidspaceTheMue: have you made progress?17:16
wwitzel3perrito666: haha, my fingers are alergic to cracking open 100 of them17:16
perrito666wwitzel3: oh, you shoult learn to do with your midle front tooth, its a life changer17:17
perrito666:p17:17
natefinchwwitzel3: you should buy them shelled :)17:17
natefinchwwitzel3: I actually did the math and at least at my grocery store, they actually come out to almost the exact same price per pound of edible bits17:18
perrito666oh, that is an option too17:18
perrito666natefinch: for that you actually had to peel all of them and weight them17:18
perrito666right?17:18
wwitzel3natefinch: yeah, but then, I eat too many and I've complaining about a stomach ache .. either way I'd be complaining :P17:18
natefinchhaha yep17:18
perrito666anyway I cannot eat pistachios while not having beer, so I eat peanuts which are easier to peel17:19
natefinchperrito666: yes, actually.  I bought a bag of the in-shell version, ate them, then weighed the shells17:19
perrito666natefinch: clever17:19
natefinchperrito666: I forget the percentage right now... maybe 30% of the weight was shell... something like that17:19
perrito666ericsnow: ping17:44
ericsnowperrito666: hey17:44
perrito666ericsnow: priv17:44
wwitzel3well that 14.10 upgrade did not go well .. had to boot a live CD and do a boot repair, and now things are a bit wonky18:17
wwitzel3all my dev stuff still works fine though, so that's good18:17
perrito666wwitzel3: ah I got an "interesting" behavior in network for a moment18:18
perrito666wwitzel3: and in the rest of my system too as you most likely saw in my attempt at a hangout yesterday18:19
natefinchsinzui: reverting that merge here: https://github.com/juju/juju/pull/97918:24
sinzuithank you very much natefinch .18:26
perrito666natefinch: you missed the fix tag18:26
natefinchdammit18:35
natefinchnow remind me, because I always forget... do I put that in the $$ $$ or do I have to put it in the previous commit message?  Also does it have to be 'fixes-1386766' or is it supposed to be 'fixes 1386766'  (space vs. hyphen)?18:37
natefinchsinzui, perrito666 ^^18:37
sinzuinatefinch, fixes-138676618:45
sinzuiyou can optionally __fixes-1386766__ to make it bold and show the works you mean it18:46
natefinchsinzui: yes, but does that work inside the $$stuff$$?  I feel like I've tried that in the past and it didn't work18:48
natefinchsince I think it expects just alphanumeric inside the $$ $$18:48
sinzuinatefinch, "$$merge$$ fixes-1386766" in the same comment is fine18:48
natefinchok18:48
sinzuinatefinch, and if you think CI is blocking you from merging a critical fix, add __JFDI__ to make it be tested and merged on success. __JFDI__ is our safety to ensure an automated system doesn't stop os from doing the right thing18:49
natefinchok, I thought i remembered as JFDI, but couldn't remmeber the exact syntax.  Is there documentation of this stuff somewhere?18:50
sinzuinatefinch, in an email to canonical-juju.18:52
natefinchemail is not documentation :)18:52
sinzuinatefinch, I can add a page to reports.vapour.ws18:52
natefinchsinzui: that would be awesome, thank you :)18:52
sinzuinatefinch, I am not saying it is, but I write a lot of official gdocs, an no one reads them18:53
katcosinzui: i am familiar with that problem; one solution is to have much fewer docs, but more sections so that you have a single starting point to give people18:54
katcosinzui: and when these situations arise, you just point and go: "go forth and learn"18:55
natefinchsinzui: I understand the frustration.  hopefully we can help people remember that documentation exists18:55
natefinchyep18:55
katcoit doesn't help that docs is not very navigatable imo18:55
natefinchyes, sorry, I thought gdocs was a typo.   Google docs is a terrible place to keep documentation18:56
katconatefinch: well, it's not _horrible_, but i feel like there needs to be some sort of indexing mechanism18:57
natefinchdocs should be at a website with a url that means something18:57
rick_h_katco: howdy, I mentioned it to wallyworld last night. There was a small issue that came out of the work to not destroy envs when there's no jenv file that we hit. https://bugs.launchpad.net/juju-core/+bug/138670318:57
mupBug #1386703: New instance created when bootstrapping an existing env without the jenv file <juju-core:New> <https://launchpad.net/bugs/1386703>18:57
natefinchlike juju.ubuntu.com/docs18:57
katcorick_h_: thanks, i'll have a look. are you all blocked on this or what's going on?18:58
rick_h_katco: as an fyi, not sure where it'll fit on the todo list but wanted to bring it front/center as it's an issue for quickstart18:58
katcorick_h_: yeah... i'll bring it up to wallyworld_ tonight in our standup18:58
rick_h_katco: no blocked, but would be great if the release didn't close one bug and open another right next to it.18:58
rick_h_katco: ty18:58
katcorick_h_: gotcha. thanks for the ping.18:58
voidspaceg'night all18:59
natefinchI wish godeps didn't say "failed" when it's actually working as designed19:12
natefinchah dammit.... google's winsvc package isn't importable in linux, because it's all // +build !windows   ... and so godeps can't find it, so godeps -t doesn't include it in the output19:17
perrito666natefinch: fork it?19:22
natefinchahh, you know what works?  GOOS=windows godeps -t ./... > dependencies.tsv19:23
* natefinch feels pretty smart19:24
natefinchit works because windows dependencies are a strict superset of the linux dependencies19:24
natefinchperrito666: quick review? http://reviews.vapour.ws/r/277/diff/#19:30
perrito666natefinch: what is the 58 added at the end of service?19:31
natefinchperrito666: I forget.  some hg thing.... possibly an easy to remember incremental diff id.  It's optional, so people who insert dependencies in the file by hand often forget it.... but since I generate the file with godeps, it remembers and puts it in there19:33
perrito666good enough for me19:33
perrito666natefinch: lgtmd19:33
natefinchperrito666: thanks19:34
thumpermorning folks19:41
natefinchmorning thumper19:41
alexisbmorning thumper19:41
* thumper looks sadly at his very full inbox19:49
natefinchthumper or fwereade: got a second for a UX question?    charm level constraints include "container type blacklist"  what do we do if the user says to deploy a charm to a container that the charm says it doesn't support?  Warn the user?  prevent the deployment?  If we prevent it, is there a way to override?20:31
natefinch(or anyone else who has an opinion is free to speak up too :)20:32
thumperI think that we shouldn't stop them, but yes to warning20:32
thumperand say "the charm says this won't work"20:32
perrito666did the metadata fix land?20:32
thumperbut if the user explicitly says "do it this way" we let them try20:33
natefinchthumper: sounds good20:33
natefinchthumper: do we still skip over newly created containers made with add-machine?20:33
natefinchthumper: actually... do we auto-deploy to an empty container?20:34
natefinchthinking about 'juju add-machine && juju deploy foo' will add a machine and then deploy foo to it... but not sure if that's true of 'juju add-machine lxc:4 && juju deploy foo'20:35
thumpermaybe...20:35
waiganimenn0_: was it relations or relationScopes that broke the world?20:35
natefinchthumper: easy enough to test.... still, I presume it would skip containers by default20:36
natefinchdammit20:39
natefinchstupid replicaset tests, who wrote that crap, anyway?20:39
natefinchsinzui: tests giving "No space left on device"20:41
sinzuihmm20:41
sinzuiwhich one natefinch20:41
natefinchsinzui: http://juju-ci.vapour.ws:8080/job/github-merge-juju/1042/console20:43
sinzuinatefinch, this might be tricky I see 29G free.20:45
* sinzui pokes arround20:45
sinzuiah, this is the machine instance20:45
=== menn0_ is now known as menn0
waiganimenn0: remember those selects from yesterday? They just selected _id. But then the unmarshalled charm doc is expected to have a URL - which *was* the id but is now it's own field. So I've added "url" to the select and all works again.20:50
sinzuinatefinch, this perplexing. I don't see evidence of the instance size being changed. We can see about 6G available on the root disk20:50
menn0waigani: that makes sense20:51
waiganimenn0: yep, cool20:51
* sinzui ponders changing the working dir to one of the mounted disks20:51
perrito666sinzui: I am having an interesting issue, I have my script, that creates a stream structure with "juju metadata generate-tools -d ${BUILD_FOLDER}/juju-dist/" creating, among other things, juju-dist/tools/streams/v1/index.json but then I have juju failing because it cannot find juju-dist/tools/streams/v1/index.sjson20:52
perrito666sinzui: happened to you?20:52
sinzuiperrito666, pull trunk, natefinch reverted wallyworld_'s commit that changed the default behaviour20:52
* perrito666 pulls again20:52
waiganimenn0: and with that state passes20:52
waigani\o/20:52
menn0waigani: awesome20:53
natefinchperrito666: it hasn't landed yet.... it keeps failing20:53
perrito666natefinch: ah20:53
perrito666a world of joy20:54
natefinchindeed20:54
natefinchsinzui: I'm happy to retry, but my EOD is in 5 minutes20:54
* perrito666 drinks a beverage at 90°C to protest against the summer heat20:54
sinzuiperrito666, maybe something else is broken or we aren't testing the reverted commit. CI still cannot deploy anything using streams20:54
sinzuinatefinch, retry your branch20:55
natefinchsinzui: it's going20:57
sinzuithank you for your time natefinch21:00
waiganimenn0/thumper: what do you thing about creating an interface for environment specific collection documents with DocID and EnvUUID fields? One use would be testing for those fields in the upgrade steps. Would there be other value to make it worthwhile?21:00
* thumper otp21:01
menn0waigani: what would the interface look like? remember that interfaces are only for methods, not for fields21:01
waiganimenn0: ah right, it would have to be an embedded struct - messy21:02
perrito666ok I am taking a break while natefinch chanes land, bbl21:13
perrito666and as I said that, the change landed21:13
natefinchhuzzah21:14
thumperperrito666: plz send some summer heat my way22:24
thumperbit chilly here still22:24
=== urulama is now known as urulama__
perrito666thumper: sure, Ill be glad to get rid of some summer heat, specially because its still freaking spring22:48
alexisbwallyworld_, you around?22:58
wallyworld_alexisb: yes, ia am, i have a few emails to respond to, did you want to chat? i'm in a meeting now but will be free soon22:59
alexisbyeah I do, if you are in a meeting now then I will ping you in about 20 minutes22:59
wallyworld_sure23:00
perrito666ericsnow: what implementation of tar are you using?23:11
ericsnowperrito666: pretty sure it's the one you put into the utils repo23:12
perrito666ericsnow: the symlinks implementation is broken and I recall fixing it23:13
perrito666ericsnow: 972 has a rb link?23:32
waiganiwhat should happen if you upgrade juju while a unit agent-state is pending?23:36
davecheneythumper: the uniter tests don't pass for me at trunk23:50
davecheneyhave you seen that ?23:50
anastasiamacperrito666: thnx 4 z review!23:58
perrito666np, don't take me as an authority anyway23:59
perrito666most where suggestions23:59

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