[00:00] hml: hi, I just wanted to ping because I'm attempting to rain on a parade I guess. [00:00] rick_h: :-( [00:00] hml: and wanted to make sure my comment was coherent and make myself available if there were any questions/etc but please don't land https://github.com/juju/juju/pull/7449 [00:01] rick_h: okay - i’ll give it a look - we wanted to get something out there for feedback - expected input - you were on my list to ping. [00:01] hml: k, I think the intent is <3 but the way the command is put together is a bad precedent imo [00:05] rick_h: sounds like a chat is a good way to go. do you have some time now? or tomorrow? [00:08] hml: yes I can now [00:08] Let me head my headphones [00:09] hml: https://hangouts.google.com/hangouts/_/canonical.com/rick?authuser=1 when you want [01:39] wallyworld: do you know where I can find peer-xplod? jam's charm says unauthorized for me [01:39] wallyworld: ISTR you used it recently [01:39] axw: yeah, i copied the source locally [01:41] axw: i go tthe source from juju acceptance tests [01:41] there's a copy of peer-xplod in there that is up to date [01:42] wallyworld: cool, thanks [04:31] menn0: not sure how busy you are - it's ok if you won't get a chance to look at my recent cmr pr today - i'll bug someone else tomorrow. just need to know so that i can plan next steps [04:38] wallyworld: i'm in calls and usual household crazy period at the moment so am unlikely to be able to review today [04:38] wallyworld: I could do first thing tomorrow though [04:38] np, all good [04:38] i'll branch off that branch and continue work [04:38] wallyworld: email me with the link if you want me to take a look [04:38] will do ta [04:38] https://github.com/juju/juju/pull/7504 [04:39] i asked you because we discussed a bit yesterday [04:39] the new controller details collection [04:39] wallyworld: cool cool [04:39] jam: i'm likely to be a bit late to the standup. have to pick up a child. [04:39] menn0: understood [05:13] wallyworld jam: I've just added some data to https://github.com/juju/juju/pull/7502#issuecomment-308411294, going for a ride - I'll land when I get back if you're happy [05:14] ok [05:14] looking [07:10] Bug #1697175 changed: juju 2.1.2 trusty node put into recovery mode [07:13] Bug #1697175 opened: juju 2.1.2 trusty node put into recovery mode [07:19] Bug #1697175 changed: juju 2.1.2 trusty node put into recovery mode [14:03] in case anyone's around, here's a PR that provides a drop-in replacement for context.WithDeadline that uses mockable clocks: https://github.com/juju/utils/pull/282 [14:09] thumper: wallyworld: can either of you clarify what, if any, logging changes were made in 2.2? i noticed yesterday, using a 2.2 client, that i was not getting charm logging by default like i used to. is that expected? if it is, we really need to announce that and explain how to turn it back on, because w/o those logs it's much harder to debug charm problems [14:10] this was on a model that i created on an aws jaas controller === frankban is now known as frankban|afk [16:58] this PR makes api.Open time out more reliably: https://github.com/juju/juju/pull/7507 [21:27] axw: any objections? https://github.com/juju/juju/pull/7508 [23:02] wallyworld, axw: do you happen to know the prometheus collection interval? [23:02] typically [23:02] menn0: 15s i think by default [23:02] menn0: it depends on the yaml config [23:02] ok [23:03] prometheus will poll juju [23:03] wallyworld_: b/c the state metrics collector is written in a fairly inefficient way [23:03] every poll it calls ForModel for every model [23:03] that's going to hurt [23:04] yeah [23:04] menn0: we have a todo to fix that, maybe it's a mental todo [23:06] wallyworld_: ok cool [23:06] wallyworld_: I'm going through where ForModel is used, looking into problems axino is seeing [23:07] I've found one bad and unnecessary use in the allwatcher [23:07] easy to fix too so i'll do that today [23:08] menn0: yeah, there's no reason ti use for model anymore - state pool should be it. we just haven't updated evertything yet. we've been doing as a driveby [23:08] as we find things [23:10] wallyworld_: the reason i'm looking is that axino is seeing query storms on txns.log [23:11] wallyworld_: this is when the txn watcher resyncs [23:11] he's seeing 300 at a time [23:11] there shouldn't be that many watcher instances [23:12] possible cause is lots of new State creation for some reason [23:12] hence me looking at ForModel calls [23:13] wallyworld_: I also have an idea for getting rid of all the txn watchers so there's just one [23:13] wallyworld_: without major code churn (I think) [23:13] wallyworld_: I'd like to chat about that later [23:13] wallyworld_: does your PR still need review? [23:14] menn0: yeah it does :-) [23:14] menn0: axw is also looking at txn watchers, so let's chat for sure [23:40] anastasiamac: I fear that exposing that charm method will lead to misuse; it's just a convenience. maybe you can show how you would use it? [23:43] wallyworld_: can you please take a look at https://github.com/juju/juju/pull/7505 when you have a few minutes? [23:43] sure [23:44] menn0: I'd love to know how you plan to have just one txn watcher. I spent some time on it, it got messy... but I was going for a more drastic refactor too [23:46] menn0: standup? [23:59] menn0: btw I started looking at using StatePool for the statemetrics code yesterday, happy to continue if you have other things you'd rather do