[00:51] I assume we already have some tooling for storing time values as json on disk. Can someone point me at an example? [00:53] going for a run, hope it doesn't all go wrong while I'm out! [00:56] thumper: ping [01:03] voidspace: hey [01:03] thumper: hey, hi [01:04] thumper: I'm working on bug 1632362 [01:04] Bug #1632362: error during juju-wait ERROR cannot load cookies: file locked for too long; giving up: cannot acquire lock: resource temporarily unavailable [01:04] thumper: which *appears* to be a problem with the juju/persistent-cookiejar library around the way it does file locking [01:04] thumper: the suggestion in the bug is to switch to a many-readers-one-writer style lock [01:05] ugh, otp right now [01:05] thumper: I'll email it [01:05] ack [01:05] thumper: I hear you have "opinions" on file locking... [01:05] thumper: :-) [01:05] :) [01:05] thumper: although it seems like really we want a database... [01:06] thumper: which may only be as complex as rewriting file locking from scratch [01:06] or less... [01:06] thumper: assuming that's the real bug... [01:22] OK. I'm calling it. Time to go watch the incoming results. [01:22] good luck America [01:22] I'm also calling it time [01:22] g'night all [02:19] thumper, ping [02:20] alexisb: pong [02:20] heya do you have a second to chat? [02:20] yep [02:20] hangout or bluejeans? [02:20] k meet you in bluejeans [02:20] ack [02:20] let me plug in my laptop [04:23] Bug # changed: 1543362, 1544158, 1565943, 1607786 [04:51] someone shoot me [05:24] * anastasiamac does not want to shoot redir [07:04] alexisb: ping, can you grab wallyworld for the session? [07:08] rick_h: where r u? [07:08] anastasiamac: in the bolero room for the cross cloud controller [07:10] rick_h: we r in opereta, talking with uros [07:10] anastasiamac: wallyworld is there as well? I've got some guests in here it'd be good to get backup here [07:11] rick_h: come over here - the same conversation is taking place [07:11] rick_h: yes, wallyworld is here [07:11] anastasiamac: k, the calendar says this room so we've got 5 folks in here to talk about it [07:12] rick_h: there is more here :D [07:59] good morning juju-land. can I get a review of https://github.com/juju/charm/pull/225 and https://github.com/juju/juju/pull/6549 to fix https://bugs.launchpad.net/juju-core/+bug/1639855 ? [07:59] Bug #1639855: 2.0 metrics breaks 1.25 [08:30] no seriously. [08:30] Shoot me. [08:32] * redir goes to bed hoping not to have a 4 year hangover [08:44] could someone review these patches? https://github.com/juju/juju/pull/6523 https://github.com/juju/utils/pull/249 [12:45] dooferlad: ping [13:57] rick_h: we doing 1:1 today? [14:00] voidspace: how can I help? [14:00] rick_h: I'll be a bit late anyway, nothing *burning* that I need to talk about that can't be done by email [14:01] rick_h: so happy to skip, but also very happy to chat if you're around [14:01] dooferlad: hey [14:01] dooferlad: so I'm fixing the file locking to use juju/mutex instead of filelock [14:01] dooferlad: which uses named pipes (etc) to do the locking instead of the filesystem [14:01] dooferlad: but it has a 40 char max, and cookiejar uses full file paths which can exceed 40 chars [14:02] dooferlad: so I'm thinking that a hash of the full path is the thing to do [14:02] dooferlad: and am wondering which hashing algorithm to use - SHA1 looks promising in terms of a small output hash sized [14:02] voidspace: can you give me a path to the file that has the problem? [14:02] dooferlad: but I thought you might have "opinions" that would be helpful [14:02] voidspace: :-) [14:02] dooferlad: I'll have to dig to get that [14:03] dooferlad: and my wife just got back, let me come back to you on it [14:03] voidspace: sounds good [14:05] rick_h: I just got the MAAS storage email [14:05] rick_h: I'm very close to a complete PR for the issue I'm on currently, so I'll complete that then switch to MAAS storage [14:06] rick_h: I've not looked at storage before, but there are plenty of people to ask if I need to (most of them on sprint with you though...) :-) [14:08] Voidspace thank you. Understand it's not an immediate turnaround. [14:21] voidspace: on the subject of locking I would say that looking at the persistent-cookiejar code we could name a mutex 'juju-cookies' as long as we only have one location for the cookie file (next to the Juju settings) and that we need to import any existing cookies from the current location (JUJU_COOKIEFILE, GOCOOKIES or $HOME) if [14:21] $HOME/.local/share/juju/cookies doesn't exist. [15:35] jam: ping [15:46] dooferlad: I think it's ok to use 29 character hash (of SHA256) : last ten chars of file path [15:46] dooferlad: and as we're the only user for juju cookies I don't think the cross-version incompatibility will be a problem [15:47] voidspace: if we are the only users, why not just ask for a jar name on creation and use "juju/cookies/ sorry, juju/cookies as that name [15:48] when calling from our code [15:48] dooferlad: because the underlying library needs to generate locks for arbitrary paths [15:49] voidspace: good point, well made :-) [15:54] this feels a lot like we forked the cookie code instead of implementing a library to wrap file access inside a juju/mutex [16:00] Bug #1640521 opened: Unable to deploy Windows 2012 R2 on AWS [16:18] Bug #1640521 changed: Unable to deploy Windows 2012 R2 on AWS [16:21] Bug #1640521 opened: Unable to deploy Windows 2012 R2 on AWS [19:08] rick_h - correct me if i'm wrong, but to bootstrap a controller with proxy settings we just drop http-proxy|https-proxy flags in a config.yaml and bootstrap with --config=myfile.yaml?