=== ming is now known as Guest47081 [02:55] How can I specify in the `requires` section of metadata.yaml that a database is required, but could be one of several choices (e.g., mongodb OR postgresql OR mysql)? [03:00] nwingfield: you would add an interface for both mysql and postgresql and mongodb but set them as `optional: true` [03:01] nwingfield: there may be a way to require that at least one is provided, but I'm not certain [03:01] yeah, i can’t find any examples of mutually exclusive interfaces [03:02] i’ll give that a shot though [03:06] nwingfield: hey there, yes have all three as optional [03:06] thanks thumper [03:06] but with new status work, the charm could be put into a status that says "awaiting database" [03:06] so anyone that goes "juju status" would see that it isn't yet running [03:07] this is the new 'workload status' work [03:10] that will be nice [03:11] thanks for the help blr, thumper [08:21] I've got a spinning jujud here - it's dumping tons of lines with "machine X has new addresses" over and over, cf. http://paste.ubuntu.com/11644054/ [08:21] Anyone seen something like this? [10:32] there are two manual environments running in rackspace. i have a task to move n machines from environment A to environment B. is something like that possible at all? [13:04] jamespage, if you have a moment: https://code.launchpad.net/~gnuoy/charms/trusty/percona-cluster/use-dc-stable/+merge/261372 [15:09] Is there a difference between 'juju-run ...' and 'sudo juju-run ...' on an instance? [15:11] Odd_Bloke: I wouldn't think so, as `juju-run` runs as root already [15:13] aisrael: I'm running it from a shell as not-root; does it escalate itself? [15:13] Odd_Bloke: there is one context in which juju run will run as the ubuntu user [15:13] and i beleive thats when you target --machine, vs using --service or --unit [15:14] Odd_Bloke: otherwise the context of a juju run session is executed as root, no elevation required. [15:44] Is there a way to tell Juju not to overwrite configs? I'm editing Glance config and when I reboot Glance the configs switch back to the Juju settings. [15:45] Bialogs: it's in the charm and what it does. The charm has to support non-charm related edits to the configs. [15:46] rick_h_: Thanks [16:32] hey marcoceppi, have you ever seen the juju unit commands just flat out hang? [16:32] eg: unit-get, juju-log, etc. [16:33] doesn't seem to matter which command i run, its just hanging around like its attempting to connect to the event socket and never timing out. [16:48] does it initiate an ssh connection? === kadams54 is now known as kadams54-away [17:26] I'm not sure about that Zetas, but i want to say no. [17:27] lazyPower: is there a gui running? canit connect to the wss? [17:28] rick_h_: i terminated the env and stood it back up. its worth noting that this also happened after dhx'ing into the unit, and that may have made an environmental change [17:28] oh hmm, ok [17:28] so i dont necessarily think this is core related, and more something wonky happened [17:30] lazyPower: have the url for charm testing handy? It's not linked in the bug and I can't find it in the docs under charm testing or anything. [17:31] lazyPower: is that not "live" atm? [17:31] rick_h_: thats a deceptively large topic, are you looking for the amulet docs? as in "how do i test my charm"? [17:31] lazyPower: the "charm tests failed on my charm, I think I fixed them, where's that test rnuner that would prove this to be true on a reporting page" testing [17:32] ah [17:32] http://reports.vapour.ws/latest-bundle-and-charm-results [17:32] that's not linked in the docs - the individual test runs should be embedded in the bug filed for the MP/Review request [17:33] lazyPower: right, just following up on https://bugs.launchpad.net/charms/+bug/1459345 and curious if the update ran/passed tests [17:33] Bug #1459345: Review/promulgation request for the Redis charm [17:33] rick_h_: that was probably run out of band by hand by a charmer. [17:34] i do know, that the current impl of the revq is linked to an older testing project in jenkins, and we have a newer setup that's using container isolation to kick off the tests that seems more reliable. [17:34] lazyPower: ah ok. Well not much we can do then. ty [17:34] we still have an open ticket to integrate the new method, and make it a tad more self-serving to end users, eg: open MP, get automated results like osci [17:35] thats upcoming this cycle by devx aiui, but no promises. I dont manage the workload on that team :D [17:35] lazyPower: we have more than an open ticket, there's a spec being drafted that makes CI way more robust [17:35] even. better. [17:35] hmm, juju-gui results going back to aug of last year? [17:41] rick_h_: i may have spoken too soon i was able to reproduce. Deploying the gui now [17:41] rick_h_: is there something specific to look for other than wss connectivity? [17:41] lazyPower: well if it can talk to the API but you can't from the cli it might be network issues or something [17:42] lazyPower: basically can you load up the gui, service details, unit details [17:42] ack, makes sense [17:43] yeah gui is working as expected [17:44] this is weird, if i detach and re-run the hook it works as expected. [17:44] yet when attached via debug-hooks, dhx, etc. it hangs on every unit-agent command [17:44] config-get, juju-log, unit-get [17:44] * lazyPower takes an env dump === kadams54-away is now known as kadams54 === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 [19:29] Hi, I was looking to deploy Openstack HA with juju and MAAS. I could not find any good step by step guide and architecture guide for it. Can someone help me point to it? I will be very thankful :) [19:52] beisner: ping :) ^ [19:52] o/ bilal [19:58] hi bilal - the openstack charms do generally support HA use cases, however only non-HA openstack bundles (solutions) have been published thus far in the juju charm store. === brandon is now known as web [20:05] Thanks beisner. So in order to make a charm e.g. keystone charm to work in HA usecase, what different steps do I need to do? Also do you know if maybe outside juju charm store, someone has posted HA reference architecture? [20:13] lazyPower: so I got the django 1.7 support tests all passing with bundletester over the weekend [20:13] GREAT SUCCESS \O/ [20:13] lazyPower: which included one with django 1.8, which also found a bug, which is also fixed in the branch [20:13] thats awesome news thumper [20:13] the autocommit attribute in the postgres config was removed in 1.8 [20:14] lazyPower: was using juju 1.24 to run the tests though, not 1.23 [20:14] lazyPower: I have celery worker support working locally too [20:15] thumper: ok, lets land the updates to 1.7 first [20:15] there is a wip merge proposal up so I can track changes [20:15] then we'll get celery in a feature branch [20:15] lazyPower: yes, definitely [20:15] sounds awesome though, thanks for the dedication there [20:15] lazyPower: it needs some tests though [20:15] so does whats in teh charm store :) [20:15] also coming once this has landed will be moving all the feature tests to amulet [20:15] oh thank the developer, that will be great [20:16] instead of continuing the carry the pre-amulet tests forward [20:16] * thumper nods [20:16] as I mentioned before, I'd like this charm to become an example of a well written charm [20:16] an exemplar if you will [20:19] as we get workload status too, I'd like to hook this up [20:19] seeing the great work landing in juju master right now has me eager to use it [20:20] kinda hard to work out if django is fully "up" though since it is a framework charm [20:21] * thumper thinks... [20:21] lazyPower: how about this as a crazy arse idea... [20:21] lazyPower: we encourate a "special" url in the app [20:21] lazyPower: something configurable, but defaulting to something like '/.juju-status' [20:22] lazyPower: and use that for workload status [20:22] if it is configured [20:22] That doesn't seem very practical and very scoped to juju. [20:22] socialize juju into django apps :) [20:22] I think something along the lines of checking for the CSRF token you get out of a django site would be good [20:22] that doesn't tell me it is working [20:22] for example [20:22] it tells you the framework is responding, othewrise you'd get a 503 bad gateway error and no CSRF token in the tag [20:22] as far as I'm concerned, django isn't working if my app isn't there [20:23] my app provides https endpoints [20:23] lazyPower: this assumes that you are using the csrf middleware (which, you should be, but don't have to) [20:24] lazyPower: how about this... [20:24] lazyPower: as config [20:24] lazyPower: you specify which relations are needed [20:25] lazyPower: if all relations are made, service is up [20:25] lazyPower: that way, as the charm user, I could say "up means pqsql, amqp, django-settings" [20:26] although ideally we'd want to be able to specify the remote side name [20:26] that sounds strikingly similar to what the services framework is using as a basis for charming [20:26] I'd like to be able to say "you need a django-setting relation with foo" [20:26] lazyPower: I know nothing about the services framework [20:26] but what you're telling me you want to do sounds mor elike a health check of the service to a) ensure its online and b) has the associated components online and in a healthy state [20:26] exactly, this is what workload status is [20:26] and it shows up in juju status [20:27] (or it will soon...) [20:27] 1.25 I think [20:27] i think this needs more brainstorming than I have capacity for at the moment, because i see your point of "my service is an amalgamation of these components, just checking 200ok is not enough" [20:27] but theres a fair amount of overlap here that may or may not be present in other apps, and other healthchecks. [20:27] * thumper nods [20:28] s/apps/services/ [20:28] lazyPower: how about we schedule a chat some time this week [20:28] with perhaps another eco member or two [20:28] and just spit-ball some ideas [20:28] im booked this week in prep for DockerCon, can we do this after the conference if you need me there? [20:28] look, this isn't urgent, and unlikely to happen in the next two weeks [20:28] so we can do it after [20:29] I'm really just thinking out loud [20:29] so to speak [20:29] perfect, i dont want you to lose the enthusiasm/momentum you've got going though [20:29] so i'm willing to be flexible when its not stacking :) [20:29] lazyPower: unlikely, I'm using this charm in a production environment [20:29] and I want it to be good [20:31] sounds good, we have open office hours [20:31] and this would be brilliant to bring up during the office hours - it might be a bit early for you though [20:32] lazyPower: when is it? [20:32] 6 AM in Canberra ACT [20:33] lazyPower: good think I'm in NZ and it is 8am then [20:33] oh perfect [20:33] i had no idea so i guessed at your locale [20:33] well i knew you were a kiwi, but thats as far as that goes [20:33] with my wife leaving for work early, it isn't hard for me to start at 8am [20:33] ok cool, on the 18'th of this month at 8am [20:34] I'll pencil it in [20:35] lazyPower: I have to say, with all this charm dev work, it makes me want to add more to the charmhelpers and making juju better... [20:36] also, I have a patch for bundletester [20:36] orly? [20:36] but I haven't worked out how to add a test... [20:36] lazyPower: just upper cases the debug flag [20:36] thumper: does it change -l DEBUG to -l d ebug? [20:36] so it can handle 'debug' [20:36] haha [20:36] i love it [20:36] called it [20:36] :) [20:38] lazyPower: https://github.com/juju-solutions/bundletester/pull/19 [20:40] thumper: commented, one of the devx peeps will give that a thorough review / merge [20:40] sure, np [20:40] hi bilal, generally, the HA needs and topologies vary quite a bit from one environment to another. also as you note, the approach is different per service (percona-cluster vs. keystone vs. glance and such). [20:41] aisrael: 'preciate you [20:41] lazyPower: Right back atcha, dude. === natefinch is now known as natefinch-afk === kadams54 is now known as kadams54-away === brandon is now known as web === kadams54-away is now known as kadams54 === CyberJacob is now known as zz_CyberJacob === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 === kadams54 is now known as kadams54-away