[00:04] anastasiamac: ^^^ that is the first part of my fix [00:04] thumper: looking ;) [00:20] thumper: reviewed :D [00:35] wallyworld: did the KVM/LXC image cache expire old images? [00:36] I'm guessing not, since that would require some smarts in the server - and it was just a proxy IIRC [00:36] axw: no, was quite a simple implementation [01:31] axw: babbageclunk : standup? [02:03] wallyworld: here's that bug: https://bugs.launchpad.net/juju/+bug/1679948 [02:03] Bug #1679948: juju bootstrap is failing in the MAAS CI lab [02:03] TA [02:08] wallyworld hml: got a call from oracle about my trial... did I miss anything? [02:09] oh you're still there [02:09] axw: no, just explaining how build agent etc works [02:09] anastasiamac: thanks [02:09] anastasiamac: next one coming up shortly [02:10] thumper: nps :D m looking forward to the sequel [02:16] veebers: sorry, caught up in meeting, will miss the QA catchup i think [02:16] wallyworld: nw [02:53] anastasiamac: https://github.com/juju/juju/pull/7233, +251 −730 (as long as you ignore the 217 files bit) [02:54] thumper: wow... anything in particular to pay attention too? coz otherwise m going to hit the button with my eyes closed :D [02:54] * thumper looks for the files of interest [02:55] cmd/juju/commands/main.go [02:55] thumper: ack [02:56] cmd/jujud/main.go [02:56] cmd/plugins/juju-metadata/metadata.go [02:56] that's it really [02:57] everything else is dealing with using cmdtesting from the juju/cmd package rather than juju/juju/cmd [02:57] and the rename of juju/juju/cmd/cmdtesting to juju/juju/cmd/cmdtest [02:57] thumper: ack. looking -tyvm!! [02:57] The only thing left in the juju/juju/cmd/cmdtest now is some weird arse function about running a command with the dummy provider [02:58] can't fix everything at once [02:59] thumper: agreed. lgtm'ed [02:59] ta [03:12] wallyworld: Can you take a look at https://github.com/juju/juju/pull/7234 [03:12] sure, i just need a quick bit to eat [03:13] wallyworld: I need to pop out but after that can we chat about the changes needed for short-circuiting? [03:13] wallyworld: cool cool [03:16] babbageclunk: sure, can do [04:16] babbageclunk: left some comments, happy to chat whenever [04:16] awse, thanks [04:16] looking now [04:37] menn0: if you happen to find urself with nothing to do \o/ PTAL at my update of https://github.com/juju/juju/wiki/MgoPurgeTool with everything u've mentioned [04:41] Bug #1492237 changed: juju state server mongod uses too much disk space

[04:43] anastasiamac: "runaway transactions on apihostports or machines" could be lumped into cleaning up txn references from all documents [04:43] at one point we had specific problems on a couple collections, but now we cleanup all of them [04:43] (the newest one is model documents) [04:43] I don't know if people care as much about the specific issues vs the category of issue, though [04:44] oh jam, that the wording I did not touch :) [04:44] Prunning I think is spelled Pruning [04:44] ah, that one would be mine - anything misspelt cannot b menn0 :) [04:44] I'm pretty sure pruning has to run on the primary [04:44] compact needs to be run on secondaries [04:45] we've also found that it isn't 100% compatible with Juju 1.18, though it is close [04:45] jam: yeah, that is why is said juju 1.25.x :) [04:45] mgopurge drops the db entirely for presence, and we can't do that, but you can drop the individual tables [04:47] Bug #1492237 opened: juju state server mongod uses too much disk space

[04:47] anastasiamac: unfortunately I am famous for typos [04:48] jam: pruning can run anywhere I think (all writes go to the primary), but it's most efficient to run on the primary [04:48] menn0: jam: fixed everything except does pruning need to/must run on primaries? [04:48] k. i'll add that as a note too :) [04:49] menn0: well, if it redirects all writes to the primary, are you running on a secondary? I'm not sure what guarantee we are opening the session in [04:49] (strongly consistent auto redirects all requests to the master) [04:49] which would mean you might be running the script on another machine, its still running on the primary [04:50] Bug #1492237 changed: juju state server mongod uses too much disk space

[04:53] anastasiamac: why did you mark https://bugs.launchpad.net/juju/+bug/1492237 as Won't Fix vs Fix Committed for 2.2 ? [04:53] Bug #1492237: juju state server mongod uses too much disk space

[04:53] though I suppose the other aspects are "configurable log" and "statuseshistory" would be the other 2 commits for 2.2 [04:53] and then what to do about the charm blob store [04:54] esp. wrt JAAS cache where the admin of the db isn't the one whose data we are working with. [05:01] anastasiamac: coming to https://hangouts.google.com/hangouts/_/canonical.com/discuss-roadmap?authuser=1 ? [05:04] anastasiamac: r u joining? [05:05] sorry sorry got distracted by imprtant ppl [05:12] wallyworld: take another look? https://github.com/juju/juju/pull/7234 [05:12] babbageclunk: will do, just in yet another meeting, *sigh* [05:12] wallyworld: ok, no worries :) [05:14] wallyworld: running a test against GCE now - wasn't sure whether rebooting the machine was enough - wouldn't enter scope also happen in that case anyway? [05:14] babbageclunk: don't think so - scope entry is on the controller side [05:15] ah right [05:42] babbageclunk: i'm multi-tasking - how did the GCE testing go? [06:19] axw: quick review if you have time? https://github.com/juju/juju/pull/7235 [06:19] wallyworld: sure [06:19] ty [06:20] bbiab after coffee run [06:43] anastasiamac: I see, so we're not "imprtant ppl"... :,( [06:45] jam: u r... but there is also a higher category of "very imprtant" [06:46] :D [06:50] anastasiamac: ah, I see. so priority inversion, unable to interrupt the 'imprtant' people to meet with the 'very imprtant' ones [06:53] :) === frankban|afk is now known as frankban [07:54] wallyworld: sorry, was afk. [07:54] no wuckers [07:55] wallyworld: it looks like it didn't work because the bounced machine hasn't picked up its new public address - I guess a problem with the GCE provider? [07:55] wallyworld: :( [07:57] babbageclunk: there's an instance address poller than runs but then backs off once the address is known. the agent is also suposed to report addresses from memory. i can't recall the specifics [07:58] but that's not for this PR [07:58] wallyworld: hmm - I guess the agent wouldn't know that the machine's public address changed, right? [07:58] wallyworld: so it would have to be the instance poller. [07:59] wallyworld: Ok, I'll merge my PR anyway [07:59] babbageclunk: there's supposed to be an address poller in the agent i thought [07:59] but i could be wrong, will have to check [07:59] wallyworld: but it would need to talk to the provider to find out the public address. [08:00] why? [08:00] the agent would poll the nics [08:00] on the machine [08:00] i guess it depends on how things are configured [08:00] maybe you're right [08:01] maybe the agent start up should trigger the instance poller [08:01] maybe that's already supposed to happen [08:01] so many questions [08:06] axw: i've pushed changes, too bad ParseDuration() doesn't support "days" [08:07] wallyworld: yeah. we could always add another function that extends the syntax, but at least this way it's more flexible [08:10] wallyworld: LGTM [08:12] axw: ty [08:13] will test again befor elanding [08:13] babbageclunk: i've added a card to ensure we remember to check for why rebooting an instance causing an address change is not being picked up by juju [08:25] wallyworld: oh, it eventually did! Haven't had a chance to investigate - talking to the maas guy. [08:26] babbageclunk: that matches my understanding then, whew. the instance poller will wake up and do it. we just need to be smarted about triggering it after a reboot [08:30] axw: turns out that the TMPDIR-clearing antics are the dynamic loader's fault [08:30] (because there is a setuid executable in the chain of invocations when you run a snap) [08:33] mwhudson: ah, right [08:34] jam: order does matter if you want to have something to test against :) [08:40] wpk: I'm not worried about SortedValues as much as I am whether we need to preserve the *users* order [08:41] wpk: and I believe it doesn't matter, so shoving it into a Set is fine, and the SortedValues is appropriate === admcleod is now known as admcleod_afk [08:48] jam: technically it's a set [08:48] jam: also, that's what proxyupdater has been always doing [08:48] wpk: sgtm [09:06] jam: I'm chasing this bug 1679948 with a guy from the maas team [09:06] Bug #1679948: juju bootstrap is failing in the MAAS CI lab [09:07] jam: The controller looks like it's running fine, but the client can't connect to the api - we just see the Forbidden message in the log [09:15] jam: connectivity seems fine, we can ssh from the client to the controller [09:17] jam: and in the logs on the controller we can see the various workers connecting the the API as well, so I'm not sure why the client can't [09:18] babbageclunk: are you able to wget against the controller? [09:18] is the URL a valid uuid for a model? [09:19] jam: gah, didn't think of just wgetting - I was trying to find a websocket client I could use [09:20] it just tells you if you can open the socket [09:20] you can also try "telnet host 17070" or whatever the exact prot is [09:20] jam: yeah, will get in touch and try that [09:24] jam: I guess telnetting to it would tell something even though it's not going to do the https handshake [09:39] babbageclunk: you'll see the SSL headers come over [09:39] babbageclunk: there is also "openssl s_client" [09:39] which is 'telnet but for SSL' [09:41] jam: found a websocket client code example that might be a useful test bed - just seeing how far I can get to pass in some JSON to login to the API and eg list the models. === Spads_ is now known as Spads [11:18] jam: here's a PR that addresses one of the items we were discussing in today's call [11:18] https://github.com/juju/juju/pull/7237 [11:19] menn0: looking. what is cloud 'dev' ? [11:20] jam: just lxd with config for my local apt proxy and disabling auto apt upgrade [11:21] jam: ala https://github.com/juju/juju/wiki/Faster-LXD#suggested-juju-config-for-lxd-deployments [11:22] sure, I have the same, just put it on 'lxd' itself [11:22] menn0: https://github.com/juju/juju/pull/7237#pullrequestreview-32605432 lgtm, though in *my* opinion, if we're going to have '--no-switch' then we really should have '--switch' [11:22] and I'm hesitant to add a short option [11:22] we could live with just landing what you requested [11:23] *proposed [11:24] jam: I don't understand why we would you have --switch? if there's already --no-switch. [11:24] jam: I was borderline on the short option but it seemed like the kind of thing that some people would want to do all the time [11:24] menn0: so if you have "--dont-do-foo" it seems useful to have a "--do-foo" to invert it [11:24] it may be that "--do-foo" is currently the default [11:24] jam: but if that's the default.... [11:24] but that doesn't let us change it [11:25] *i* like the ability to be explicit, especially for things like scripts [11:25] if it's the default, most people won't use the option IMHO [11:26] and we can't really change the default (until 3.x anyway) for compatibility reasons [11:27] menn0: its personally a pet peeve when there is an option but you can't explicitly select a behavior, as then you have to know what the default behavior is, instead of just being able to say "this is the behavior I want" [11:27] that, however, is *my* opinion, so if you'd like a second opinion/consensus feel free to override/ask others [11:28] fair enough, i'll ask around next week [11:28] it's late and there's no particular rush to land this right now [11:28] menn0: for example, in 'mgopurge' we have -ssl=false as how you disable it, but you can specify -ssl=true [11:29] yep, I guess so [11:29] on the flip side we already have -no-gui and --no-browser-login [11:29] on bootstrap === hml_ is now known as hml === coreycb_vaca is now known as coreycb === rick_h_ is now known as rick_h === Dmitrii-Sh_ is now known as Dmitrii-Sh === salmankhan1 is now known as salmankhan === BradCrittenden is now known as bac === frankban is now known as frankban|afk === blackboxsw is now known as blackboxsw_bbl === blackboxsw_bbl is now known as blackboxsw === coreycb is now known as coreycb_vaca