[03:31] bdx: that might be possible. I think that reference to a long model history was mentioned before. anastasiamac do you recall a bug/fix around a long model history causing issues recently? [03:32] * anastasiamac looking [03:33] rick_h: so, the issue dissapears when I create a new model [03:33] bdx: yea, which is why I don't see it in anything I can access [03:33] I've re-written half of my charms today [03:33] thinking its my charm code [03:34] bdx: https://github.com/juju/juju/commit/fd20944902f03e60e8477dd6969a9d42d1c50701 [03:34] anastasiamac: ^ [03:34] its been a very frustrating day .... I was suppose to do a production deploy for one of our apps .... but then the model started acting super wonky, and I started re-writing charm code ... really its been the wonky-ness of the model this whole time [03:35] rick_h: what does that do? [03:36] bdx: it adds an index to the db to make queries faster when collections get large [03:36] bdx: at least that's my read of that commit [03:36] rick_h: bdx: the bug refered in that commit was https://bugs.launchpad.net/juju/+bug/1668646 [03:36] Bug #1668646: Model migration fails [03:36] bdx: yea, https://github.com/juju/juju/pull/7059 [03:36] yeah [03:36] that it [03:37] rick_h: bdx: but it seemed to have been afecting migration only... [03:37] that *is it [03:37] ooh [03:37] anastasiamac: yea, just looking. I'm not sure if that change is useful or not here and bdx's perf problem [03:37] oh, possibly not [03:37] yeah [03:38] I mean an index on a collection is an index on the collection. [03:38] but things like juju status/etc I'm not sure if they hit this collection [03:38] rick_h: bdx: we hae cyclone tail here - exciting... i have missed the beginning of ur conversation... what is happenng? [03:38] anastasiamac: oh! are you by it? I guess you are. [03:39] anastasiamac: how was that? I've never been in a cyclone. [03:39] rick_h: it's a lot of water with a lot of wind :D [03:39] anastasiamac: bdx is noticing sluggishness in api response on a model in JAAS. He's verified that if he creates a new model it's fast. So an older model with a long history seems to slow down. [03:39] anastasiamac: heh, I wonder what's different in cyclone, hurricane, etc. [03:41] rick_h: no diff in terminology - except the location/where it happens: http://oceanservice.noaa.gov/facts/cyclone.html [03:41] rick_h: but schools, etc closed. ppl r told to saty indoors :) [03:42] anastasiamac: well yea. Here hurricanes are beastly natural diasters [03:42] I'm glad you're ok! [03:42] rick_h: bdx: JAAS is runing on 2.0x underneath [03:42] anastasiamac, rick_h: juju has cyclone tails happening in it right now too ... possibly I could grant you guys acls on the model so you can see/experience it first hand [03:42] rick_h: bdx: we have fixed every leak/performance issue we r aware of in 2.2... but it's at beta stage atm... [03:43] anastasiamac: yes, it is. [03:43] anastasiamac: ahh, so what I'm experiencing is probably fixed in 2.2 you are thinking ? [03:43] anastasiamac: right, I'm eager for 2.2 to come out and then to test/verify if the perf issue is effected [03:43] bdx: m hoping \o/ [03:43] you can't rush perfection ;) [03:44] magicaltrout: :D [03:48] rick_h, anastasiamac: do you think my loaded models with long history are bogging down the controllers though? [03:48] bdx: no, I can see the cpu/memory/disk IO/etc on all three of them [03:48] bdx: and nothing there is looking off. Load on all three are < 3 [03:49] rick_h: so, what gives then? [03:49] bdx: that's why I was thinking your hint that a new model was fast means that there's probably just a slow query/something in that model. [03:51] rick_h: yeah ... I mean ... its really chaotic [03:51] bdx: rick_h: here is the list of performance bugs that have been fixed in 2.2 :) it's possible that some of ur long living models have some leaking db connections.. bug # 1635311, [03:51] bug # 1671258, bug # 1651291, bug # 1634328, bug # 1587644, bug # 1581069, bug # 1649719 [03:51] rick_h, anastasiamac: I can deploy the same thing 10 times and everything gets borked/stops in a different state everytime [03:51] lol [03:51] I can make any sense of it [03:52] anastasiamac: thx, looking [04:36] rick_h, anastasiamac: on a new model -> http://paste.ubuntu.com/24279037/ [04:37] seamless [12:15] is there anyone around that i can chat with regarding the charm partner programme? [12:21] SimonKLB: I think you want arosales, who is in a US timezone [12:26] SimonKLB: Or SaMnCo , who is EU [12:29] I am EU [12:29] @SimonKLB ^ [12:30] And you can certainly talk to me about CPP :) [12:39] thanks! === frankban|afk is now known as frankban [13:32] hi Juju world, a new question, I found Grafana/InfluxDB very slowly these last days on our CDK clusters, I just noted that they have very short "limits" (100MHz / 100Mio per containers) sets on the influxdb-grafana pods, can I edit this value "freely" in /etc/kubernetes/addons/influxdb-grafana-controller.yaml and not be overrided by next Juju updates of CDK? [15:04] kwmonroe: ping, halp! I'm trying to figure out where to go here to file a bug on the zeppelin charm but lost in this site. [15:04] rick_h: working as designed. [15:04] :) [15:04] kwmonroe: bah [15:05] kwmonroe: I want to file a bug about relating to pgsql but :( [15:05] and the haproxy one [15:05] rick_h: since zepp is upstream at bigtop, the appropriate thing to do is open a JIRA here: https://issues.apache.org/jira/browse/BIGTOP/?selectedTab=com.atlassian.jira.jira-projects-plugin:summary-panel [15:05] rick_h: do you see a "Create" button at the top of that page? [15:06] kwmonroe: no, I probably need to create an account and all that [15:06] ah shoot rick_h. yeah, i just logged out and noticed "create" isn't an option without an account [15:06] yea, gotcha. Ok will do that then. [15:07] no no rick_h - that's a poor UX i think. we keep a fork of their repo. how about i update our bigtop charm bug links to point to https://github.com/juju-solutions/bigtop/issues [15:07] i think that'll be easier for people to work with [15:08] though it means i'll have to translate issues into a JIRA, but still, i think it's more likely that people will grok opening a gh issue vs a jira. [15:19] is there a way for me to tell juju which IP is the "public" one? [15:19] (preferably on the model level) [15:20] kwmonroe: up to you. I'm paid to go through pain on it if I need to but agree. When I went to the bugs-url for the charm I ended up getting lost in links trying to figure out how to find the charm and it wasn't clear to me what was charm vs other code. [15:24] understood rick_h -- unfortunately there's no way to link people directly to a sub-module of bigtop (like the charms) for filing bugs, so i do think it's an extra burden to expect people to figure out bigtop's bug tracker. please open your issue(s) at https://github.com/juju-solutions/bigtop/issues and i'll take on the pain of translating to jiras. i'll update the bugs-url next rev. [15:26] kwmonroe: all good ty [15:26] plus it's far easier for me to nak issues at github. === cmagina_ is now known as cmagina [16:02] lazyPower: hullo - do we have a good bundle reference for *beat + elasticsearch and kibana ? [16:50] is there something about reversations I'm missing? for the second time I've seen MAAS hand out reserved IP wreaking havoc in the network === frankban is now known as frankban|afk [17:16] 0/ [17:16] o/ juju world [17:37] centos7 bootstrap with the manual provider on juju 2.1.2 doesn't seem to work: http://paste.ubuntu.com/24282504/ Here's a paste of my virtual machine log [17:38] slightly more context: http://paste.ubuntu.com/24282509/ [17:40] axw: ^^ [17:41] axw: i'll let this setup running if you want to poke at it or gather any more logs [17:49] axw: i think the issue was centos7 comes out of the box with a firewalld enabled. i'm killing that and trying again [17:50] axw: yeah that was it. ignore me === kina___ is now known as kina__ === dames is now known as thedac === urulama is now known as urulama|afk [20:58] hi there, trying to get started with charms.. and found some odd inconstancies.. could the charm command be broken on zesty? [20:59] bug report - https://bugs.launchpad.net/ubuntu/+source/charm-tools/+bug/1675240 [20:59] Bug #1675240: charm-create crashed with ImportError in /usr/lib/python2.7/dist-packages/charmtools/utils.py: cannot import name path [21:05] kwmonroe: half of a fix for that matrix issue you found here: https://github.com/juju/python-libjuju/pull/100 (cc cory_fu) [21:05] (The other half is a commit to matrix, but it needs to be fixed in python-libjuju first.) [21:52] lazyPower, you go a con and this place goes quiet! [21:52] ARGHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH [22:22] what's up, magicalt1out? [22:23] nowt, it was just a bit quiet :) [22:23] lol [23:49] cholcombe: you can't use centos7 for the bootstrap machine. only ubuntu is supported for running the controller