=== chuck__ is now known as zul [01:41] magicaltrout: heh, YOU POLLUTING THE CHARM DEVELOPER PROGRAM WITH ABANDONDED INSTANCES?! [07:58] sorry marcoceppi ! I don't know what is left in the disconnected world, if I have stuff running, feel free to shut it down [11:47] Anyone have any idea of when juju 2.0 RC1 or next Beta will rear it's head? [17:28] openstack-charmers: the issues I was experiencing with inability to modify the 30G default account quotas, and swift<->rgw api incompatibilities for rgw buckets has been remedied in Jewel :-) [17:35] I'm so distraught that I've a hammer deploy :-( .... I've successfully preformed an upgrade from hammer to jewel on test stacks ..... I'm trying to get a stable upgrade/migration procedure that I could apply to my production hammer deploy, I think I'm pretty close..... I'm wondering if there are any gotcha's the storage team might be able to unveil concerning upgradin hammer to jewel ? [17:39] bdx: what's Jewel? [17:42] marcoceppi, maybe refering to the Ceph LTS which is code named Jewel. not sure of the context [17:42] the latest stable release of ceph packaged in xenial [17:43] oooh nm. Its not in the xenial archives :-( [17:44] only ceph==10.1.2 .... grrr [17:44] openstack-charmers: is there a way we can bump ceph to jewel for xenial? [17:45] openstack-charmers: ceph v10.2.0 [17:45] storage-peeps: http://docs.ceph.com/docs/master/release-notes/#v10.2.0-jewel === natefinch__ is now known as natefinch [18:04] bdx: I think xenial includes a release candidate of jewel - and it will be upgraded later to the final release [18:06] ReSam: thanks, I see that now. Final release of 16.04 charms? [18:24] bdx: almost all the openstack-charmers are in Austin at ODS, so responses might be delayed [18:26] marcoceppi: totally, thx [18:52] mbruzek - i know you're occupied atm, but here's where we left off at standup https://github.com/juju-solutions/layer-tls/pull/21 [18:52] lets see if we can poke a hole in this, but this looks like its doing what i want it to do [19:13] urgh bloody interfaces [19:13] can I prod something to find out what states my charm has set? [19:14] magicaltrout charms.reactive get_states [19:15] ta [19:26] lazyPower: either I'm having a major brain fade or something is wonky [19:26] if I'm in juju debug-hooks how do I get access to charms.reactive? [19:26] its included in layer-basic, so its on $PATH if you're in a hook context [19:26] thats what I thought [19:27] root@ip-172-31-22-199:~# charms.reactive get_states [19:27] bash: /usr/local/bin/charms.reactive: Permission denied [19:30] marcoceppi you like this random stuff [19:30] what am I doing wrong/broke [19:30] wat [19:31] magicaltrout: obviously you have not acended past root to reach god level status which is the permission required [19:31] hehe [19:31] cory_fu: ^^ [19:31] magicaltrout: can you strace the command and paste the output? [19:32] that one is easy [19:32] root@ip-172-31-22-199:~# strace charms.reactive get_states [19:32] strace: Can't stat 'charms.reactive': No such file or director [19:32] +y [19:33] magicaltrout: strace /usr/local/bin/charms.reactive for thrills [19:33] ah that ones a treat [19:33] http://pastebin.com/EzRkWye5 [19:33] same old stuff [19:34] magicaltrout: can you ls -lah that file? [19:35] -rw-r--r-- 1 root root 360 Apr 25 19:00 /usr/local/bin/charms.reactive [19:35] magicaltrout: well there's ya problem [19:35] chmod +x that file [19:36] yeah that works [19:36] weird [19:36] magicaltrout: super weird [19:38] yeah its like that on my other unit as well [19:40] Wily & Juju 2.0-beta4 according to my status [19:42] problem is no one cares about Wily... its like the forgotten love child.... ;'( === matthelmke is now known as matthelmke-afk [19:43] okay [19:43] i have an interface question cause I can't get this to work [19:44] and you lot seem to think this stuff is easy ;0 [19:44] http://pastebin.com/HTfgbSbZ [19:44] dcosmaster.available is set by the master charm when its installed and running [19:45] if I debug the master charm i can see it set [19:45] but I don't seem to be picking it up on the requires end [19:45] any tips? [19:45] I can see the various hooks being triggered in the debug log [20:06] magicaltrout: That file should be +x when it's installed from the wheelhouse by pip. If it's not, then I can only guess that it's a problem with pip on wiley? === redir_ is now known as redir [20:13] i like the way you answer that with a question mark cory_fu as if I have a any idea how pip works! ;) [20:13] I do have a question though [20:13] @hook('{requires:dcos}-relation-{joined,changed}') [20:13] i have that in my interface [20:14] how can i find out what it things requires:dcos is [20:14] because when i hard code it to dcosmaster my hook runs [20:14] and when its templated [20:14] it doesn't [20:15] i assumed it picked it up from metadata.yaml [20:15] magicaltrout: it gets that from the metdata.yaml [20:16] magicaltrout: what's the name of the interface? [20:16] It does pick it up from metadata.yaml. It should match any relation in metadata.yaml in the "requires:" section that has "interface: dcos" [20:16] dcos or dcosmaster? [20:16] "requires": [20:16] "dcosmaster": [20:16] "interface": "dcos" [20:16] magicaltrout: that will be triggered anytime dcosmater-relation-joined/changed is called [20:16] yeah well thats what i thought [20:17] but that only happens when i remove the templating [20:17] i realise you'll think i'm the crazy mad man at this point [20:17] but its actually true :P [20:18] magicaltrout: I believe you and defer to my associate, cory_fu ;) [20:18] magicaltrout: linky to the full requires.py file? [20:19] http://pastebin.com/nZWQ1brm its just a hacked up one from the docs [20:19] doesn't do much at the mo [20:19] ignore line 11 I deliberately broke it to check its execution :) [20:20] magicaltrout: is it because of the class name? [20:21] doubt it [20:21] i shall check [20:22] oh wtf [20:23] looks like you're correct marcoceppi [20:23] spent bloody ages looking at that wondering why filling in the templates worked fine [20:23] teaches me to copy and paste [20:24] magicaltrout: copy and paste - it's the reason we built layers ;) [20:24] magicaltrout: lesson learned, we need to have a charm create -t interface [20:25] * marcoceppi opens bug [20:25] yeah that would be really good [20:25] oh well i can return to actually implmenting the busines end now :P [20:25] \o/ another blocker unblocked [20:26] yeah i'm well in a hole in terms of time and apachecon, the less blockers the better please :P [20:27] magicaltrout: well please feel free to aggressively ping me [20:27] magicaltrout: happy to make sure thta's a success for you [20:27] magicaltrout: and to basically defer all the hard stuff to cory_fu [20:28] hehe [20:29] I think the juju stuff is mostly in place. For the demo I plan to spin up Apache OODT for Data Management in DC/OS and use the Big Data bundles to spin up a hadoop platform and have some interaction between them all, probably similar to cory_fu's ssh failed attempts [20:29] suck them into the platform via Apache OODT for catalog and search stuff and Zeppelin for the analysis [20:30] which is a pseudo real world scenario we have, although we don't currently use DC/OS clearly as its rather new [20:30] and honestly, OODT, Mesos, Zookeeper, Hadoop, Zeppelin, how much more Apache tech could you have in a talk thats powered by Juju? [20:31] i don't think its possible [20:34] oh and Tomcat and Solr and Tika [20:34] blimey [20:35] magicaltrout: it's like the entire apache catalog ;) [20:36] yeah its not far off, but also is real life, so its not like i'm trying to cram it all into just to make a point [20:36] but it certainly helps ;) [20:39] :D [20:51] marcoceppi: whats the name of the stuff that houses downloads in 2.0 and is there any documentation for the feature I've forgotten? :) [20:59] hmm looks like classname might not be the problem [20:59] fresh install back to broken [21:02] fscking templates [21:04] cory_fu: http://pastebin.com/4LWd3s8q that runs and logs fine [21:04] http://pastebin.com/6WVCExFc [21:04] that sinks without trace [21:05] http://pastebin.com/x1yUwwCz [21:05] thats my metadata [21:05] suggestions please! :) [21:09] ack ack [21:09] bloody typo [21:09] always the users fault [21:13] magicaltrout: Don't use -broken in reactive [21:13] Also, your hook line should be @hook('{requires:dcos}-relation-{joined,changed}') [21:14] Other than that, what's the problem? [21:14] yeah i saw it ;( [21:14] I was in a meeting, and not really able to pay attention before. :( [21:14] if you shouldn't use broken you should probably get evilnick to touch up the docs: https://jujucharms.com/docs/devel/developer-layers-interfaces [21:16] Yeah, we should update that. It will only actually cause problems if you use set_state in there, but it doesn't work like you'd expect so it's better to avoid it entirely [21:16] cool [21:17] as you're here and marco isn't whats the name for that charm storage stuff? [21:17] not "storage" but where you store artifacts [21:17] Are you asking about resources? [21:17] my google-fu is failing me [21:17] that be the one [21:17] i knew it began with r [21:17] :) [21:17] magicaltrout: So I'm still confused what your issue is with the interface layer [21:18] cory_fu kwmonroe admcleod1: this PR on hue does a lot of cleaning and adds integration to ZK and Hive, integration with YARN & HDFS was already there (as far as I remember) https://github.com/juju-solutions/layer-hue/pull/6 [21:18] You definitely should not hard-code the name of the relation in the @hook line [21:18] i don't it was a typo and lack of understanding :) [21:18] Ah, ok [21:18] not the hard coding was just because i couldn't see why it wasn't running [21:18] and to prove to myself i wasn't completely mad [21:19] arosales: As kjackal pointed out above, we think that Hue is basically ready (pending the PR) and we can hopefully have it cleaned up and done by EOW [21:23] arosales: Just tested upgrade from spark 1.6.0 to 1.6.1 and it works [21:28] marcoceppi: with juju resources, on the first execution of a charm download an archive [21:28] then in the charm dump it into resources so in future it doesn't have to fetch it again [21:29] cory_fu: kjackal interesting and good to hear. [21:29] kjackal: is that with spark only or with hadoop too? [21:30] this is for spark 1.6.1 to 1.6.0 and a fixed version of Hadoop 2.6.0 [21:31] kjackal: gotcha, and I think the hadoop admcleod1 is working on is 2.7.x so perhaps a little work there, but good to hear that is close. [21:32] kjackal: burning the mid night oil too I see ;-) [21:33] arosales: The binary was for Hadoop 2.6.0+ . I do not think there is a Spark binary for Hadoop 2.7.0 only [21:34] arosales: regarding Hue the misunderstanding (at least from my part) was that we do not need the entire functionality of Hue, we need integration with YARN. We can gradually integrate Hue with the rest of the Hadoop ecosystem [21:35] magicaltrout: that's not how resources work [21:35] kjackal: for sure we can always build from there [21:45] marcoceppi: maybe, but it should be how resources work! :P [21:47] for example, DC/OS it has to download a 500mb file to bootstrap === matthelmke-afk is now known as matthelmke [21:47] if it was a resource, that would be pretty instant, not that its that slow over the net, but you get my drift [21:48] so I could define it as a resource and have people upload a bootstrap tarball to the Juju server resource pool [21:48] or, I could have some cool hook that grabs it on the first run, and dumps it into the resource pool so future runs could grab it from there instead of the web [22:13] magicaltrout: well... [22:22] well indeed! [22:53] boom [22:53] semi working DC/OS [22:53] what a hack! [22:53] ;) [23:32] magicaltrout: all great production deployments start as hacks :) [23:33] https://ibin.co/2f1dfqzROiqs.png [23:33] thanks marcoceppi nice charm developer program 12 node cluster ;) [23:42] (don't worry marco, I switched it off :P )