[00:09] fwereade__: hey, thanks for the reviews now go to bed! [00:11] fwereade__: are you up again? [00:31] wallyworld_: want to do an old fashioned review? https://code.launchpad.net/~thumper/golxc/mockable/+merge/169317 [00:31] sure [00:31] diff still updating [00:31] i really don't like roetveld [00:31] for latest additions [00:32] you're all good now [00:33] wallyworld_: also, can I have a hangout with out to teddy-bear some design ideas? [00:35] sure [00:41] hmm... [00:42] * thumper needs a recursive find and replace [01:07] wallyworld_: thanks for the golxc review, will get to those [01:07] just doing a massive move branch [01:07] np [01:08] i wish we used interfaces more in juju [01:08] wallyworld_: fwereade__ asked to have environs.Instance moved to instance.Instance [01:08] so container.Container doesn't need environs.Instance [01:08] ok, sounds good [01:08] * thumper goes to make lunch [01:33] wallyworld_: chat? [01:34] sure [01:34] https://plus.google.com/hangouts/_/3fc078d1d8dacd2c85f1f72fb71d3ed67d97b9c1?hl=en [02:08] wallyworld_: you have two reviews [02:12] wallyworld_: trivial review for you Rietveld: https://codereview.appspot.com/10253053 [02:15] * wallyworld_ looks [02:18] wallyworld_: happy for a trivial on that? [02:18] yep [02:18] * wallyworld_ looks for the other review [02:18] * thumper woners [02:18] woners [02:18] try again [02:19] wonders? [02:19] wonders perhaps why the D key is broken? [02:19] I typed woners again first tiem [02:19] tiem? [02:19] our tarmac lander checks for two LGTMs [02:19] :-P [02:19] time [02:19] pierce off fuck knuckle [02:20] now, now. public forum :-P [02:20] i forgot about the tarmac thing [02:20] maybe we need to patch it to allow "LGTM trivial" through [02:21] wallyworld_: yeah, or perhaps we should just get people to review in lp :) [02:21] to get the two reviews from the merge propose [02:21] PLEASE YES [02:21] proposal [02:21] that way we can check for a trivial tag [02:25] thumper: i'm sure i didn't merge your branch - only trunk. i last merge late yesterday i think. i normally merge one last time anyway prior to submitting [02:25] wallyworld_: that change wasn't in trunk [02:25] hmmmm. [02:26] * wallyworld_ is confused [02:26] however, I have done things like this with this workflow [02:26] switch to someone elses branch to look [02:26] create a new brack with "switch -b new-work" for something new [02:26] nah, didn't do that [02:26] however since I wasn't on trunk, it starts with the last revision of the other branch [02:26] anyways, i'll sort it out [02:26] after lunch :-) [02:28] kk === tasdomas_afk is now known as tasdomas [04:41] * thumper off early due to late work last night [04:41] ciao everyone [07:24] morning [07:34] fwereade__: thx ;) the resumer is indeed not very complex. [07:34] TheMue, haha, yeah :) [07:38] fwereade__: hmm, somehow rogers and my branch merge proposals of yesterday don't go into trunk but they have status Approved [07:38] fwereade__: any idea as an old launchpad user? [07:38] TheMue: https://code.launchpad.net/~themue/juju-core/025-cleaner-worker/+merge/169228 ? [07:38] As I warned, it doesn't have a Commit Message [07:39] you can copy the description into "Set Commit Message" [07:39] jam, btw, thank you ever so much for doing all that work [07:39] (its probably the biggest case where things fail for people with Tarmac, I tried to be upfront that it is a failure mode) [07:39] I've had that problem as well [07:40] there is a patch somewhere that has tarmac set it back to Needs Review with a note that it needs a commit message, which at least avoids the black-hole of 'why isn't this working' [07:40] but I haven't tracked it down yet. [07:40] jam: Hmm, I thought I have something in the commit message. *wonder* [07:40] TheMue: https://code.launchpad.net/~themue/juju-core/025-cleaner-worker/+merge/169228 you have a Description set, but not a Commit Message [07:40] separate fields [07:41] (Launchpad models it as one is the conversation you want to have, vs the thing you want saved for posterity) [07:41] jam: *aaargh* have to learn this first, thanks [07:42] TheMue: as an added bonus, I believe LP's resubmit actually removes a commit message if it was set in the original proposal. And *that* is pretty stupid. :) [07:42] *drumroll* will see now [07:43] TheMue: no proposals for your prerequisite branch [07:43] TheMue: I'm cheating by looking at the tarmac log [07:44] your 024-cleanup-watcher is marked as 'this must land before I can land 025' but you don't have '024' proposed for merging (and approved) [07:44] jam: that is already merged. do i have to resubmit it too? [07:44] TheMue: I think it is just tarmac confused because of the resubmit [07:44] * TheMue is sometimes really astonished [07:44] the old proposal was against a different branch, so it isn't seeing it. [07:45] jam: yes, it has been before the switch [07:45] Because you used 'resubmit' things got copied across that aren't relevant (It is no longer a prerequisit) [07:45] TheMue: fortunately most of this irons itself out naturally with time, I'll unstick it in a sec [07:46] jam: thankfully my next branch is one of the post-switch-era [07:46] jam: great, thanks. otherwise i would have resubmitted and approved it (with commit message) [07:47] TheMue: fortunately for prerequisite requirements, Tarmac *does* post to the request: https://code.launchpad.net/~themue/juju-core/025-cleaner-worker/+merge/169228/comments/376486 [07:48] TheMue: what other branch is affected? [07:48] jam: nice guy ;) [07:49] TheMue: pre-commit hook is now running. It failed to merge, though. [07:49] jam: aha? [07:49] TheMue: my fault [07:50] I made a change which broke the tarmac config [07:50] will fix now [07:51] jam: and i'll change my "Getting started" document with our usage of the tools and the procedures [07:52] TheMue: precommit now firing again [07:52] * jam crosses fingers [07:54] TheMue: it is currently on cmd/juju tests, so looks pretty good. We'll find out in about 15min. [07:55] jam: great, fantastic work [07:55] TheMue: sorry it didn't work first time right away, but thanks for being patient [07:55] jam: no problem, typical migration time experience [07:56] jam: i'm more happy about your effort for tarmac [07:56] hmm, sound somehow wrong [07:56] sounds [07:57] jam: happy about the work you did :) [07:57] thanks [07:58] jam: so to get sure: if it fails the status will be set to "Needs review" again, but currently no mail notification? [07:58] TheMue: it should set the status, and post the failure to the Launchpad proposal, which should generate an email for you [07:58] the only thing it doesn't send the email for (but should) is missing a commit message. [07:59] jam: ok, thanks [08:00] TheMue: I certainly see emails from earlier, check if you got them as well [08:00] * TheMue now also has an eye on the video stream from Toulouse, where the A350XWB starts for its first flight [08:01] Ah, just took of, great! [08:10] TheMue: and it has landed [08:12] jam: yep, just seen, fantastic. great job! === danilos__ is now known as danilos [10:26] * TheMue is at lunch [12:15] danilo, laaaging, you take over for a sec :) [12:15] you can also install a different editor if you like :P [12:46] mgz, I just tried '-not -name' on find and it works (though -exec seems harder since it doesn't return the subprocess return code, so I am leaving that as xargs :)) [13:43] *pinDrop* [13:44] ping drop? :) [13:45] mgz: no, just a pin drop, it's so quiet here [13:46] mgz: but ping drop sound better on irc, indeed [13:46] sounds === wedgwood_away is now known as wedgwood [15:05] fwereade__, chance I could get another look at https://codereview.appspot.com/10237043/ SetCharm soon? [15:06] Makyo, sorry, ofc [15:06] fwereade__, thanks, much appreciated :) [15:21] Makyo, LGTM, someone else review it too please :) [15:21] fwereade__, thanks. [15:29] Makyo: 2nd LGTM ;) [15:29] TheMue, Thanks \o/ [15:29] Makyo: yw [15:46] TheMue: you have a review [15:47] rogpeppe1: yust seen the mail, thanks === gmb`` is now known as gmb [15:50] rogpeppe1: the non-fatal error is simply due to the fact that it may happen during a short network problem but shall be automatically retried later (currently each minute). [15:51] TheMue: how is that different from any other error we encounter in the other workers when talking to the state? [15:51] rogpeppe1: if the resume fails what would we win if we let the worker crash? [15:51] rogpeppe1: the cleaner also only logs and doesn't crash [15:51] TheMue: if the state is borked because its connection is broken, won't we need to reconnect to it? [15:52] * rogpeppe1 wishes he had a better understanding of mgo failure-modes [15:52] rogpeppe1: me too [15:53] rogpeppe1: imo both, cleaner and resumer, don't have an own state that me be confused by an error beside their connection to mongo [15:54] rogpeppe1: so assuming that the connection is able to reconnect they can repeat their work [15:54] rogpeppe1: but indeed, i'm right now not sure with this connection [15:55] TheMue: yeah, i dunno [15:55] niemeyer: ping [15:57] rogpeppe1: Hea [15:57] heya [15:58] niemeyer: yo! [15:58] niemeyer: how's tricks? [15:59] rogpeppe1: Going :) === tasdomas is now known as tasdomas_afk [15:59] niemeyer: a couple of questions [16:00] rogpeppe1: Shoot.. I may not be able to answer them right now because lunch is waiting for me, but perhaps I can think over lunch. [16:00] niemeyer: just hoping to avoid delving into the source of mgo and wondering: if you get an error, is it necessary to reconnect to mongo, or will it happen automatically? [16:01] rogpeppe1: It will reconnect by itself, but! [16:01] rogpeppe1: The error on a session, after a connection is abruptly broken, won't go away until you either: 1) discard the session and create a new one; or 2) call Refresh on it [16:01] niemeyer: second: how easy would it be to add non-authorized s3 access to public-access buckets to goamz? and would you accept a patch that does that? [16:02] rogpeppe1: Can't parse the second question [16:02] rogpeppe1: I'd be very much against adding non-authorized access to public buckets.. (!? :-) [16:02] niemeyer: ok. even though they're open access? [16:03] rogpeppe1: Sorry, I jokingly meant the question makes no sense [16:03] niemeyer: open read access, to be more specific [16:03] niemeyer: currently you need amazon credentials to access them even though amazon doesn't require them, i think [16:03] niemeyer: when using goamz, that is [16:04] niemeyer: so... when does mgo reconnect by itself? [16:04] niemeyer: i'm not helping here by mixing the two questions up together :-) let's deal with the first one first, shall we? [16:04] rogpeppe1: I think it'd be okay to accept empty credentials for s3, yeah [16:05] rogpeppe1: I haven't thought about the implications of that in the code base, but I'd be happy to analyze with you a patch that does that [16:05] rogpeppe1: mgo reconnects whenever necessary [16:06] niemeyer: our existing plan was to just talk directly to s3 and parse the xml ourselves, but i thought that if it's easy to do in goamz we should probably do that [16:06] rogpeppe1: and once every few seconds either way, to keep the cluster state up-to-date [16:06] niemeyer: but "after a connection is abruptly broken" it's not necessary? [16:07] niemeyer: maybe i'm not getting the terminology - a "session" is a single connection? or a set of connections? [16:07] rogpeppe1: A session is an *mgo.Session [16:07] rogpeppe1: If you see a connection error on a session, that error doesn't go away until (1) or (2), as stated [16:08] rogpeppe1: You don't have to reconnect manually, you just need to ack the error via closing and recreating, or via the Refresh method [16:08] rogpeppe1: This is just because Bad Stuff could happen if a connection was recreated behind your bakc [16:08] back [16:10] niemeyer: i think i need to be refreshed on the terminology here. [16:10] niemeyer: a Session corresponds to... a single TCP connection? [16:11] rogpeppe1: http://godoc.org/labix.org/v2/mgo#Session.SetMode [16:11] I'll have lunch while you read this [16:11] biab [16:12] niemeyer: ah, perhaps you mean that once you've got a *Session it always automatically reconnects? [16:13] niemeyer: enjoy [16:14] rogpeppe1: btw, List() and URL() already work of the reader, have been trivial, like Get() will be too ;) [16:15] TheMue: ok, cool [16:16] rogpeppe1: http get, unmarshalling the two nested structs, iterate of contents, fetch key field, strings.HasPrefix(), simple [16:16] niemeyer: for when you come back: when can you see a connection error on a session? only at Dial time? or can you get a connection error at any later point? (in which case, how can you tell if it's a connection error so you know to call Refresh or discard the session?) [16:17] so, me has to leave for today, we'll continue on Monday morning [16:17] have a nice weekend [16:34] niemeyer: i've also got to go unfortunately. i'd like to continue the conversation though, as my understanding of this area is lamentably poor and it affects how we will approach some things. [16:34] niemeyer: have a great weekend [16:34] and good weekends to all! [16:40] rogpeppe1: You can get a socket error at any point [16:40] rogpeppe1: It's a normal network connection we're talking about [16:41] rogpeppe1: Supposedly, you won't want to continue doing whatever was being done on any unknown error [16:41] rogpeppe1: Re-establishing a connection may happen to a different primary, and a pretty different state from what was being accessed before [16:42] rogpeppe1: That's why we cannot simply transparently re-establish the connection [19:13] So - I think this might be one of the last few issues barring Rackspace OpenStack support for JUJU: https://bugs.launchpad.net/goose/+bug/1124561 [19:13] <_mup_> Bug #1124561: the Content-Length header is missing [19:14] I can't imagine why the requests wouldn't have the correct Content-Length value [19:14] Even adding Content-Length: 0 seems to be a fix ?