2015/08/24/#juju-dev.txt

[01:22] <menn0> waigani: would you mind looking at: http://reviews.vapour.ws/r/2456/
[01:22] <waigani> menn0: sure
[01:22] <menn0> waigani: it's a fix for intermittent allenvwatcher test failures that i've noticed
[01:22] <menn0> waigani: there's another problem to fix as well but this one issue is happening more often
[01:25] <waigani> menn0: shipit
[01:36] <menn0> waigani: thanks
[01:39] <davecheney> i love the fact that all our upgrade and provisioner tests use "ppc64le" as the "i cannot possibly happen" operating system
[01:41] <anastasiamac> davecheney: isn't it easier to support impossible operating systems? :D
[02:06] <davecheney> we're basically using quantal-ppc64le as our "nope, cannot happen" sentinal
[02:37] <mwhudson> davecheney: good think the ubuntu architecture is called ppc64el then
[02:37] <mwhudson> *THING
[02:37] <mwhudson> doh can't brain today
[02:49] <davecheney> axw_: godeps: cannot get information on "/home/dfc/src/launchpad.net/gwacl": bzr revision-info has unexpected result "244.1.1 [email protected]\n"
[02:50] <davecheney> mwhudson: oh god
[02:50] <davecheney> don't mention the war
[02:50] <davecheney> or the worst joke in the world that continues to keep on giving
[03:09] <davecheney> state tests now take 450+ seconds to pass
[03:09] <davecheney> % go test ./state/...
[03:09] <davecheney> ok      github.com/juju/juju/state      438.504s
[03:09] <davecheney> golf clap
[03:09] <davecheney> FAIL    github.com/juju/juju/state/leadership   600.006s
[03:09] <davecheney> fucking fantastic
[03:12] <mup> Bug #1487947 opened: provider/ec2: localServerSuite.TestNetworkInterfaces  <juju-core:New> <https://launchpad.net/bugs/1487947>
[06:46] <mup> Bug #1487947 changed: provider/ec2: localServerSuite.TestNetworkInterfaces  <juju-core:New> <https://launchpad.net/bugs/1487947>
[07:16] <dimitern> rogpeppe, hey, have you seen https://code.launchpad.net/~dimitern/godeps/fix-bzr-revision-info-parsing/+merge/268223 ?
[07:16] <rogpeppe> dimitern: nope :)
[07:16] <rogpeppe> dimitern: only just back from 2 weeks' holiday
[07:17] <dimitern> rogpeppe, have a look when you have a moment :) it's just 2 characters change
[07:17] <rogpeppe> dimitern: looking
[07:17] <dimitern> dooferlad, TheMue, hey guys, it seems net-cli is blessed for merging in master
[07:20] <rogpeppe> dimitern: interesting. out of interest, do you actually want to use non mainline revision numbers as dependencies?
[07:21] <dimitern> rogpeppe, I suppose not, but at least when there is such a case now godeps refuses to work anymore
[07:21] <rogpeppe> dimitern: yeah, i agree it shouldn't fail in that case
[07:21] <dimitern> rogpeppe, maybe print a warning when there are dots in the bzr revision and go on?
[07:22] <rogpeppe> dimitern: i think i'll just go with your fix actually. if you really want a non mainline revision, you can have one
[07:22] <rogpeppe> dimitern: it should be obvious in the .tsv file anyway
[07:22] <dimitern> rogpeppe, sounds good
[07:26] <rogpeppe> dimitern: reviewed
[07:31] <dimitern> rogpeppe, thanks!
[07:32] <dimitern> rogpeppe, I think you should manually merge the MP, as there's no bot to pick it up
[07:38] <rogpeppe> dimitern: merged (ooo, my bzr-fu is getting rusty!)
[07:40] <dimitern> rogpeppe, :) mine too
[07:57] <dooferlad> dimitern: great! Lets do it and go drink some beer!
[07:57]  * dooferlad knows that TheMue would rather drink whisky
[07:57] <TheMue> dooferlad: good idea, got time next week monday. wonna meet in London?
[07:57] <dimitern> \o/
[07:58] <TheMue> dooferlad: ah, you know me right
[08:00] <dooferlad> TheMue: I think you meen tuesday. Well, maybe Monday evening...
[08:01] <TheMue> dooferlad: yeah, Monday evening
[09:01] <dimitern> dooferlad, voidspace, standup?
[09:01] <voidspace> dimitern: trying, it's not letting me in :-/
[09:01] <voidspace> restarting the browser
[09:01] <dimitern> voidspace, oh boy :/ monday hangout issues
[09:02] <voidspace> dimitern: not having used this machine for six weeks issues too...
[09:02] <voidspace> 350mb of updates
[09:02] <voidspace> including a new firefox - that maybe the issue
[09:02] <voidspace> that looks better
[10:03] <voidspace> dimitern: ready
[10:03] <dimitern> voidspace, omw
[11:33] <dimitern> fwereade, hey, are you around today?
[12:49] <TheMue> dimitern: btw, the provider I meant has been the vsphere one. i has a partial implementation of the networking interface
[12:53] <dimitern> TheMue, ok, so for vsphere then SupportsSpaces will be false, but the other method will remain
[12:53] <TheMue> dimitern: yep, did it that way
[12:54] <dimitern> TheMue, +1
[13:11] <redelmann> hi, im having some problem with juju-gui
[13:11] <redelmann> machine-0 logs says:  error stopping *state.Multiwatcher resource: unit not found
[13:12] <redelmann> everytime i load juju-gui on browser
[13:12] <redelmann> and no icons appears
[13:12] <redelmann> this is a juju bug or gui?
[13:18] <redelmann> i open this issue: https://bugs.launchpad.net/juju-gui/+bug/1485249
[13:18] <mup> Bug #1485249: Juju gui is not loading. <juju-gui:New> <https://launchpad.net/bugs/1485249>
[13:56] <rogpeppe> anyone know about the openstack provider?
[13:57] <urulama> dimitern: ^
[13:59] <mup> Bug #1486106 changed: TestNetworkInterfaces fails on vivid/wily <ci> <ec2-provider> <intermittent-failure> <test-failure> <juju-core:Fix Released by dooferlad> <https://launchpad.net/bugs/1486106>
[13:59] <rogpeppe> in particular at this moment I'm interesting in understanding the different roles of the tenant-name (taken from env var OS_REGION_NAME) and region (no env var) attributes
[14:01] <mgz> rogpeppe: you have that mixed up, tenant-name is taken from OS_TENANT_NAME
[14:02] <mgz> it's a user-specific grouping of resources
[14:02] <rogpeppe> mgz: so they are - not sure how i read that wrong
[14:02] <mgz> whereas regions are a cloud-level seperation
[14:03] <rogpeppe> mgz: for some reason i'd copied/pasted the wrong env vars and didn't check back to the source
[14:03] <rogpeppe> mgz: i did think it was potentially very confusing :)
[14:04] <rogpeppe> mgz: one weird thing: why does openstack fall back to using the ec2 passwords?
[14:07] <mgz> rogpeppe: there's a semi-compatible ec2 api, so many of the rcs set the same vars for ec2
[14:07] <rogpeppe> mgz: it seems like a potential security leak to me
[14:08] <rogpeppe> mgz: and i don't see any other values that do the same thing
[14:08] <wwitzel3> fwereade: ping
[14:08] <fwereade> wwitzel3, pong
[14:09] <wwitzel3> fwereade: katco isn't feeling well, are you able to move up our discussion to now? so she can take off and get some rest
[14:09] <fwereade> wwitzel3, sure, let's
[14:09] <rogpeppe> mgz: and surely if it was using AWS_SECRET_ACCESS_KEY it ought to use AWS_ACCESS_KEY_ID too?
[14:09] <wwitzel3> fwereade: we are in moonstone, https://plus.google.com/hangouts/_/canonical.com/moonstone?authuser=1
[14:12] <mgz> rogpeppe: the names used are just based on what was looked at in practice by the clients
[14:15] <mgz> and there's not really a security issue, the access ley/secret key are not passed along unless you say you're using that authmode
[14:15] <rogpeppe> mgz: do you think it might be reasonable to drop support for using AWS_SECRET_ACCESS_KEY and EC2_SECRET_KEYS for the openstack password?
[14:15] <mgz> so you need to be pretty deliberate about wanting it
[14:16] <rogpeppe> mgz: isn't userpass the default auth mode?
[14:17] <rogpeppe> mgz: so ISTM that it'll send that password in the usual case
[14:17] <mgz> and it's basically always going over https to somewhere that won't be loging your auth attempts
[14:18] <mgz> you'd need to have and endpoint url manually set to something hostile, or dns poisoned
[14:18] <rogpeppe> mgz: from the p.o.v. of the user that's not necessarily true - they don't know what's at the other end
[14:18] <mgz> and also have accidentally sourced your ec2 creds
[14:18] <mgz> I don't think the ec2 provider is robust against that
[14:19] <mgz> we don't check that the ec2 endpoint is really amazon
[14:19] <rogpeppe> mgz: i think we do, don't we
[14:19] <rogpeppe> mgz: 'cos we use https
[14:19] <rogpeppe> mgz: without using InsecureSkipVerify
[14:19] <rogpeppe> ?
[14:19] <mgz> that just verifies it has been signed by something in your chain
[14:20] <mgz> it doesn't verify what it is.
[14:20] <rogpeppe> mgz: well, same security as any website :)
[14:20] <rogpeppe> mgz: and, yes, i'm well aware that https security is a farce
[14:21] <mgz> the ec2 signing mechanism is better though, in not passing along the secret
[14:21] <rogpeppe> mgz: but we all blithely go on assuming that verisign et al are somewhat reliable
[14:22] <rogpeppe> mgz: i mean, fair enough i guess if people actually use AWS_SECRET_ACCESS_KEY for their openstack password, we don't wanna break that
[14:22] <rogpeppe> mgz: it just triggers my Something Wrong Here sense.
[14:23] <mgz> I think it's likely not really used any more, all the current rc things will at least include both, and generally only the OS_ versions
[14:23] <mgz> just whether it's worth futzing with
[14:24] <rogpeppe> mgz: i'm moving the env vars into the openstack provider to do the environschema stuff, and it made me wonder...
[14:25] <rogpeppe> mgz: i mean, maybe goose.v1/identity should export the env var names
[14:29] <mup> Bug #1465404 changed: worker/provisioner: fail lxcBrokerSuite.TestStartInstanceLoopMountsDisallowed <intermittent-failure> <unit-tests> <juju-core:Fix Released by dooferlad> <https://launchpad.net/bugs/1465404>
[14:29] <mup> Bug #1473209 changed: github.com/juju/juju/service/windows undefined: newConn <ci> <intermittent-failure> <test-failure> <unit-tests> <windows> <juju-core:Fix Released by bteleaga> <juju-core trunk:Fix Released by bteleaga> <https://launchpad.net/bugs/1473209>
[14:35] <mup> Bug #1465404 opened: worker/provisioner: fail lxcBrokerSuite.TestStartInstanceLoopMountsDisallowed <intermittent-failure> <unit-tests> <juju-core:Fix Released by dooferlad> <https://launchpad.net/bugs/1465404>
[14:35] <mup> Bug #1473209 opened: github.com/juju/juju/service/windows undefined: newConn <ci> <intermittent-failure> <test-failure> <unit-tests> <windows> <juju-core:Fix Released by bteleaga> <juju-core trunk:Fix Released by bteleaga> <https://launchpad.net/bugs/1473209>
=== redelmann is now known as popi_
=== popi_ is now known as redelmann
=== redelmann is now known as s0plete
=== s0plete is now known as redelmann
[14:39] <dimitern> fwereade, hey
[14:40] <dimitern> fwereade, does this look in line with what we discussed? http://reviews.vapour.ws/r/2463/
[14:41] <rogpeppe> mgz: this is what i've got: http://paste.ubuntu.com/12184056/
[14:41] <mup> Bug #1465404 changed: worker/provisioner: fail lxcBrokerSuite.TestStartInstanceLoopMountsDisallowed <intermittent-failure> <unit-tests> <juju-core:Fix Released by dooferlad> <https://launchpad.net/bugs/1465404>
[14:41] <mup> Bug #1473209 changed: github.com/juju/juju/service/windows undefined: newConn <ci> <intermittent-failure> <test-failure> <unit-tests> <windows> <juju-core:Fix Released by bteleaga> <juju-core trunk:Fix Released by bteleaga> <https://launchpad.net/bugs/1473209>
[14:41] <rogpeppe> mgz: does that look reasonable to you?
[14:42] <rogpeppe> mgz: actually, i should avoid copy/pasting those env vars
[14:45] <mgz> rogpeppe: you have EC2_SECRET_KEYS ... with an S
[14:46] <rogpeppe> mgz: that's what goose uses
[14:46] <mgz> also, not really to do here, but we should add tenant-id already
[14:47] <mgz> heh, in which case you probably can just delete it, as it's been wrong for a while then
[14:47] <rogpeppe> mgz: :)
[14:47] <rogpeppe> mgz: i suspect the same of AWS_SECRET_ACCESS_KEY too
[14:47] <mup> Bug # changed: 1458717, 1461393, 1463641, 1467372, 1474788, 1478660, 1484617
[14:47] <rogpeppe> mgz: but i'm actually thinking i don't want to risk breaking anything at all
[14:48] <mgz> rogpeppe: no, that one is right, the old naming there was just dumb
[14:48] <rogpeppe> mgz: i know it's right - i just think that using an AWS name for openstack creds is silly
[14:49] <rogpeppe> mgz: particularly when it doesn't use AWS_ACCESS_KEY_ID for one of the user names
[14:49] <mgz> it was the only way to use non-password auth on the old hp cloud
[14:49] <mgz> as there aren't OS names for the different values
[14:51] <rogpeppe> mgz: ah, makes sense
[14:59] <rogpeppe> mgz: ok, here's a review for you: https://github.com/go-goose/goose/pull/14
[14:59] <mup> Bug #1459064 changed: worker/leadership: data races in test and code <intermittent-failure> <tech-debt> <unit-tests> <juju-core:Fix Released by dave-cheney> <https://launchpad.net/bugs/1459064>
[14:59] <dimitern> voidspace, dooferlad, TheMue, please have a look at these last 2 PRs we need to unblock merging net-cli into master: http://reviews.vapour.ws/r/2463/ and http://reviews.vapour.ws/r/2464/
[14:59] <mup> Bug #1460893 changed: many unhandled assigned values <tech-debt> <juju-core:Fix Released by dave-cheney> <https://launchpad.net/bugs/1460893>
[15:01]  * dimitern steps out, bbl
[15:02] <frobware> hello all - new starter on the Juju Sapphire team today.
[15:02] <mgz> rogpeppe: looks fine. as I hinted earlier, there are some bigger bits that could do with fixing in the os auth stuff, but nothing that blocks this change.
[15:02] <mgz> frobware: hey there!
[15:02] <rogpeppe> mgz: thanks
[15:03] <mgz> frobware: another Andrew?
[15:03] <frobware> mgz, tis the way...
[15:06] <ericsnow> fwereade: we didn't mean to chase you away! :)
[15:11] <mup> Bug #1459064 opened: worker/leadership: data races in test and code <intermittent-failure> <tech-debt> <unit-tests> <juju-core:Fix Released by dave-cheney> <https://launchpad.net/bugs/1459064>
[15:11] <mup> Bug #1460893 opened: many unhandled assigned values <tech-debt> <juju-core:Fix Released by dave-cheney> <https://launchpad.net/bugs/1460893>
[15:14] <rogpeppe> mgz: could you land my goose branch for me please? i'm not a member of goose collaborators currently
[15:17] <mup> Bug #1459064 changed: worker/leadership: data races in test and code <intermittent-failure> <tech-debt> <unit-tests> <juju-core:Fix Released by dave-cheney> <https://launchpad.net/bugs/1459064>
[15:17] <mup> Bug #1460893 changed: many unhandled assigned values <tech-debt> <juju-core:Fix Released by dave-cheney> <https://launchpad.net/bugs/1460893>
[15:18] <redelmann> anyone know about juju-gui?
[15:18] <redelmann> im trying to debug an issue
[15:18] <natefinch> redelmann: probably better to ask in #juju
[15:19] <redelmann> thank natefinch
[15:19] <voidspace> dooferlad: I see you reviewed the first one, are you looking at the second too?
[15:19] <mgz> rogpeppe: erm... can't remember exactly but the bot may actually just still check if you're a juju memeber
[15:19] <dooferlad> voidspace: yes
[15:19] <mgz> rogpeppe: added the comment anywa
[15:19] <rogpeppe> mgz: thanks
[15:21] <voidspace> dooferlad: ok, I'll leave it to you :-)
[15:21] <voidspace> thanks
[15:21] <dooferlad> voidspace: I haven't graduated yet. I suggest you look at the first one first - it seems to have been committed with the second as well.
[15:21] <dooferlad> voidspace: then you know which bits to ignore :-)
[15:22] <voidspace> ah
[15:22] <voidspace> I did think the second had some extraneous stuff in it
[15:23] <voidspace> dooferlad: dimitern: first looks good and it tells me which bits of the second I can ignore too
[15:24] <voidspace> in which case the second looks trivial
[15:24] <voidspace> dooferlad: although I missed the concern you raised, need to read the code agian!
[15:24] <voidspace> *again
[15:29] <mup> Bug #1488139 opened: juju should add nodes IPs to no-proxy list <juju-core:New> <https://launchpad.net/bugs/1488139>
[15:35] <mup> Bug #1488139 changed: juju should add nodes IPs to no-proxy list <juju-core:New> <https://launchpad.net/bugs/1488139>
[15:48] <mup> Bug #1488139 opened: juju should add nodes IPs to no-proxy list <juju-core:New> <https://launchpad.net/bugs/1488139>
[16:01] <bodie_> is the answer to why godeps is used rather than godep a simple one?
[16:01] <alexisb> welcome frobware !
[16:05] <mgz> bodie_: godeps is older than godep
[16:07] <bodie_> I see.  thanks mgz o/
[16:26] <rogpeppe> mgz and anyone else: fancy a smallish review of a change to the openstack provider code? http://reviews.vapour.ws/r/2465/
[16:27] <dimitern> katco & moonstone: fwereade texted me about an hour ago his power's out and wanted to convey this apologies
[16:27] <dimitern> voidspace, dooferlad, thanks for the reviews guys!
[16:30] <mgz> rogpeppe: looks fine in general, the descriptions are a bit all over the place though, do you want nitpicks?
[16:30] <rogpeppe> mgz: specific suggestions would be great
[16:30] <rogpeppe> mgz: i copied the descriptions from the boilerplate
[16:31] <rogpeppe> mgz: much appreciated
[16:31] <mgz> ah, so they're now it two places? that's a bit unfortunate.
[16:31] <rogpeppe> mgz: when i get some extra time, the plan is to generate the boilerplate from the schema
[16:31] <mgz> how much will you hit me over the head if I ask you about generating the boilerplate from the schema?
[16:32] <mgz> eheh, you win.
[16:32] <rogpeppe> :)
[16:33] <rogpeppe> mgz: i'm also not liking my decision to go with partial sentences for the descriptions. using go-like full sentences seems like it's a better plan in the end.
[16:34] <rogpeppe> mgz: but these things are fairly easy to change
[16:38] <mgz> rogpeppe: review'd
[16:38] <rogpeppe> mgz: ta v much
[16:39] <rogpeppe> mgz: not sure about mentioning that control-bucket will be auto-generated - that's kinda true of every parameter that's optional.
[16:41] <mgz> I'm a little antsy on that one as setting it can be actually harmful... but it can also just be required
[16:46] <rogpeppe> when would it be required?
[16:47] <rogpeppe> mgz: perhaps a comment in the description saying "(do not set this unless you know what you are doing)" might be appropriate
[16:47] <mgz> rogpeppe: yeah, but leave it for now, don't need to futz with here.
[16:48] <rogpeppe> mgz: k
[16:48] <mgz> we need it for ci testing on canonistack, I *think* it's related to being able to cleanup reliably
=== urulama is now known as urulama__
[16:57] <mup> Bug #1488166 opened: Service with just one unit left which doesn't think it's the leader <landscape> <leadership> <juju-core:New> <https://launchpad.net/bugs/1488166>
[17:21] <perrito666> bbl groceries shopping
[17:44] <natefinch> ericsnow: getting this error when running featuretests: charm not found in "/home/nate/src/github.com/juju/juju/testcharms/charm-repo": local:quantal/workload-actions
[17:46] <ericsnow> natefinch: is it actually there (/home/nate/src/github.com/juju/juju/testcharms/charm-repo/quantal/workload-actions)?
[17:47] <ericsnow> natefinch: perhaps still named "proc-actions"?
[17:47] <ericsnow> natefinch: (which should be fixed)
[17:49] <natefinch> ericsnow: doh, yep
[17:49] <natefinch> I blame wwitzel3
[17:49] <ericsnow> natefinch: :)
[17:51] <natefinch> ericsnow: I'll submit a PR to rename them
[17:51] <ericsnow> natefinch: sounds good
[18:19] <wwitzel3> :(
[19:28] <natefinch> I think I got whatever katco has.  I should remember to wear a mask to hangouts from now on.
[20:54] <mup> Bug #1488245 opened: Recurring lxc issue: failed to retrieve the template to clone  <canonical-bootstack> <juju-core:New> <https://launchpad.net/bugs/1488245>
=== natefinch is now known as natefinch-afk
[21:47] <davechen1y> ping http://reviews.vapour.ws/r/2458/
[22:22] <perrito666> davechen1y: ship it, I am happy that that is being worked :)
[22:22] <perrito666> there is only a typo in the description
[23:10] <davechen1y> perrito666: thanks