=== freeflying_away is now known as freeflying | ||
=== freeflying is now known as freeflying_away | ||
omgponies | what's the format for deploying a charm that's not in the official charm store but is in a bzr repo ? | 03:39 |
---|---|---|
davecheney | omgponies: local ? | 03:40 |
omgponies | without having to grab it local | 03:41 |
omgponies | looks like this works - juju deploy cs:~paulcz/precise/elasticsearch | 03:41 |
davecheney | yup, that is the format for a private charm store branch | 03:42 |
davecheney | thing | 03:42 |
omgponies | is there a flag for setting a version ... which I assume is the equivalent of the 'revision' file ? | 03:43 |
davecheney | omgponies: version is the value of the revision file in the root of your charm | 03:47 |
davecheney | or 1, by default | 03:47 |
omgponies | right, I mean during deployment | 03:48 |
omgponies | for instance 'revision' for the charm above is currently 50 | 03:48 |
omgponies | can I specify that so if some jerk does a breaking change I don't find out by surprise | 03:49 |
davecheney | omgponies: make a file called revision in the root of your charm | 03:49 |
davecheney | put the numer 50 in ther | 03:49 |
davecheney | oh, hang on | 03:49 |
davecheney | i see what you are asking | 03:49 |
davecheney | cs:~paulcz/precise/elasticsearch:$REVISION | 03:50 |
davecheney | is the full name | 03:50 |
davecheney | best to commit a revision file | 03:50 |
davecheney | hm | 03:50 |
davecheney | actualy | 03:50 |
davecheney | no | 03:50 |
davecheney | that may | 03:50 |
davecheney | probably not work | 03:50 |
davecheney | i don't think private charmstore branches have a concept of revision as strong as real charms do | 03:50 |
davecheney | there is only one revision of the charm | 03:50 |
davecheney | and that is head | 03:50 |
omgponies | right | 03:51 |
omgponies | is there a doc somewhere that describes everything that is available from the 'unit-get' command ? | 04:16 |
davecheney | omgponies: not really | 04:18 |
davecheney | from memory | 04:18 |
davecheney | public-address and private-address are the only useful ones | 04:18 |
davecheney | yup, sauce says those are the only two commands | 04:20 |
omgponies | thinking about from a monitoring perspective ... be able to get a list of units/services deployed to a box and a) set a useful hostname, b) determine what metrics to care about | 04:20 |
davecheney | omgponies: unit-get probably isn't going to be what you want | 04:23 |
omgponies | yeah I don't think there's any good way to get what I want | 04:24 |
omgponies | probably need something to stick in the middle that can correlate 'ip-10-29-206-28' to data from `juju status` | 04:25 |
=== CyberJacob|Away is now known as CyberJacob | ||
=== _mup__ is now known as _mup_ | ||
=== axw_ is now known as axw | ||
=== rbasak_ is now known as rbasak | ||
=== dosaboy_ is now known as dosaboy | ||
gnuoy | I don't have a public bucket for juju tools and use bootstrap --upload-tools to get them into a new environment. I've upgraded my client to 1.14 and now I want to upgrade the juju tools in existing environments. How do I do that ? I don't see an --upload-tools option to sync-tools | 08:47 |
gnuoy | I do see --source for sync-tools and the comment suggests I can specify a local dir but I'm unable to spot the tools on the filesystem | 09:38 |
gnuoy | ok, facepalm. I didn't know about: juju upgrade-juju --upload-tools | 09:47 |
allenap | When talking about containers in https://juju.ubuntu.com/docs/authors-subordinate-services.html, is that a general term? Or is it referring to LXC, for example? | 09:53 |
=== freeflying_away is now known as freeflying | ||
evilnickveitch | allenap, I believe it is used in a general sense. That's another page that needs a good rewrite... | 10:51 |
allenap | evilnickveitch: Cool, thanks. | 10:53 |
=== freeflying is now known as freeflying_away | ||
=== freeflying_away is now known as freeflying | ||
_mup_ | Bug #1236824 was filed: boostrap tries to build jujud <juju:New> <https://launchpad.net/bugs/1236824> | 12:16 |
=== freeflying is now known as freeflying_away | ||
=== TheRealMue is now known as TheMue | ||
_mup_ | Bug #1236900 was filed: tar: unrecognized option ''--numeric-uid'' <juju:New> <https://launchpad.net/bugs/1236900> | 14:49 |
drj11 | hello | 15:15 |
drj11 | we have changed our config.yaml for a service to add a new configuration option. How do we add that configuration option to an already running service? | 15:16 |
marcoceppi_ | drj11: you'll need to upgrade the charm | 15:22 |
drj11 | marcoceppi_: thanks. I thought so. and I thought we'd tried that. me and morty are working on it. we'll try again | 15:22 |
drj11 | marcoceppi_: thanks again | 15:23 |
marcoceppi_ | drj11: was the charm initially deployed from the charm store or from local? | 15:25 |
marcoceppi_ | adeuring: I've got a few comments on your merge for charm-tools | 15:26 |
adeuring | marcoceppi_: thanks, I'll look | 15:26 |
marcoceppi_ | adeuring: You're trying to find "maintainer" of the branch, but to be honest the owner of the branch is always going to be ~charmers for promulgated branches (or ~team) | 15:26 |
marcoceppi_ | adeuring: we don't do stacking anymore for promulgated branches | 15:26 |
drj11 | marcoceppi_: from local | 15:27 |
drj11 | marcoceppi_: we never use the charm store | 15:27 |
marcoceppi_ | drj11: gotchya, that shoudl work. You should be able to verify that the charm revision is bumped in the juju status | 15:28 |
marcoceppi_ | adeuring: so if a maintainer isn't in ~charmers (which is an expected case) then your function will fail | 15:28 |
adeuring | marcoceppi_: ok, so we might need a special rule for branches owned by charmers. But what if somebody deliberately want to fork a promulgated charm? In that case, this person should change the maintiner field. Otherwise, the official maintainers might receive undeserved "hate mail". | 15:29 |
marcoceppi_ | adeuring: It's an interesting case | 15:32 |
marcoceppi_ | I think an exception will need to be made for ~chamers for sure | 15:33 |
marcoceppi_ | adeuring: also, things like the the juju gui charm are maintained by "Juju GUI Team" not sure how this would handle that | 15:34 |
adeuring | marcoceppi_: let me look how this works with today data. | 15:35 |
adeuring | marcoceppi_: the GUI is acutally a good example: Sending a mail to the address given in the maintainer field (juju-gui@lists.launchpad.net) results in an error: "host polevik.canonical.com [91.189.95.64]: 550 unknown user". OTOH, the "real" mailing list (juju-gui@ubuntu.com) can't be checked either... Anyway, I#m open to sugegstion how else to check the sanity of the maintainer field. | 15:56 |
adeuring | ah, "juju-gui@lists.launchpad.net" would have worked, if the juju-gui team had set up this list on LP | 15:58 |
marcoceppi_ | adeuring: I think just making sure it follows "Full Name <properly-formated@email.tld>" would suffice. We don't take much responsibility for personal branches and these should be checked during finaly charm review | 15:59 |
adeuring | marcoceppi_: ok, I believe the check you suggest already exists, so let's abandon the MP | 15:59 |
marcoceppi_ | adeuring: I like the moving of the version to the package | 16:00 |
adeuring | marcoceppi_: yeah, that makes things easier for some changes, but taht's easy to include in as a drive-by fix in any real branch. | 16:02 |
marcoceppi_ | adeuring: I'll see how easy it is to just cherry pick that commit | 16:02 |
adeuring | marcoceppi_: ok, let me just revert the other changes, that's the most easy way, I beleive. | 16:03 |
adeuring | marcoceppi_: done | 16:07 |
matsubara | hi there, I bootstrapped a juju environment on openstack, then juju destroyed it. When I try to bootstrap again, juju says there's already a environment bootstrapped. Any ideas on how to fix this? Logs here: http://pastebin.ubuntu.com/6210041/ | 16:38 |
=== teknico1 is now known as teknico | ||
kurt_ | Hi Guys - I cannot get maas-tags to work with 14.1. According to the constraints documentation, it should. Any comments? | 17:05 |
kurt_ | http://pastebin.ubuntu.com/6210186/ | 17:07 |
marcoceppi_ | 1.14.1 doesn't have maas-tag support | 17:09 |
marcoceppi_ | It was just added in 1.15.1, https://lists.ubuntu.com/archives/juju/2013-October/003019.html | 17:10 |
marcoceppi_ | kurt_: ^ | 17:10 |
kurt_ | marcoceppi_: is 1.15.1 support on precise? | 17:10 |
marcoceppi_ | kurt_: all juju releases are available for all current supported ubuntu releases | 17:11 |
kurt_ | Ok, guess its time to upgrade :) | 17:12 |
CheeseBurg | no weekly update today? | 17:12 |
marcoceppi_ | kurt_: however, odd series, like 1.13, 1.15, etc are considered "devel" releases | 17:12 |
marcoceppi_ | so you need to get them from ppa:juju/devel | 17:12 |
kurt_ | marcoceppi_: Ok, thanks. I may have it already. | 17:13 |
=== jorge_ is now known as jcastro_ | ||
kurt_ | marcoceppi_: in reading the readme for 1.15.1, I'm confused by this statement: "As an unstable release we do not make guarantees about clean upgrade | 17:33 |
kurt_ | paths of running environments from one 1.13.x version to another." | 17:33 |
marcoceppi_ | kurt_: 1.<ODD>.X releases are development releases, they are not considered stable | 17:44 |
marcoceppi_ | For 1.<EVEN>.X repleases you should be able to safely run `juju upgrade-juju` to upgrade from one stable juju release to another | 17:45 |
kurt_ | marcoceppi_: right, but I am in fact going from 1.13.1 to 1.15.1. The statement appears to guarantee it will break. :) | 17:47 |
marcoceppi_ | kurt_: it shouldn't, it's just saying it might | 17:58 |
marcoceppi_ | once 1.16 comes out I recommend you sit on that release and move between 1.<even> releases | 17:58 |
kurt_ | marcoceppi_: ok, we'll see soon enough, I have it installed ;). And yep - how far is 1.16 out? | 17:58 |
marcoceppi_ | kurt_: it should be released sometime this month | 17:59 |
marcoceppi_ | iirc | 17:59 |
kurt_ | cool, thanks | 17:59 |
matsubara | Does juju keep any local state of a bootstrapped environment? I keep getting an error: ERROR juju supercommand.go:282 environment is already bootstrapped even though I destroyed the whole environment.http://pastebin.ubuntu.com/6210041/ | 18:39 |
marcoceppi_ | matsubara: is this a local environment? | 19:02 |
matsubara | marcoceppi_, canonistack | 19:07 |
matsubara | marcoceppi_, btw, I updated juju-core to 1.15.1 and now when I destroy-environment, the command returns but doesn't seem to destroy anything (i.e. If I run the command again, I get the question if I want to destroy the env) | 19:08 |
=== thomi_ is now known as thomi | ||
mhall119 | jcastro: marcoceppi_: halp! | 19:32 |
marcoceppi_ | mhall119: what's up? | 19:32 |
mhall119 | I'm trying to make my config-changed charm call my db-relation-changed charm to re-acquire my DB credentials | 19:32 |
mhall119 | http://bazaar.launchpad.net/~api-website-devs/ubuntu-api-website/api-website-canonical-is-charm/view/head:/hooks/config-changed#L49 | 19:32 |
mhall119 | but according to the charm log, that's failing: https://pastebin.canonical.com/98709/ | 19:33 |
marcoceppi_ | mhall119: yeah, because relation hooks get extra environment variables | 19:33 |
mhall119 | marcoceppi_: so how can I make this work? | 19:33 |
marcoceppi_ | you can't really call relation-get out-of-band without supplying a relation-id | 19:33 |
marcoceppi_ | mhall119: what I've done in charms is record the data in a dot file in the $CHARM_DIR and read that in other hooks | 19:34 |
mhall119 | marcoceppi_: line 48 of the config-changed hook gets a relation-id | 19:34 |
marcoceppi_ | mhall119: you need to record the $JUJU_RELATION_ID from the db-relation-changed hook | 19:34 |
mhall119 | marcoceppi_: FWIW, this is derived from the internal certificaiton website charm | 19:34 |
mhall119 | marcoceppi_: $(relation-ids db) won't work? | 19:35 |
mhall119 | or is it that I need to set that to a named env var in the db-relation-changed hook before calling relation-get? | 19:36 |
marcoceppi_ | mhall119: db-relation-changed, when fired during a relation event, will already have the proper JUJU_RELATION_ID set | 19:37 |
marcoceppi_ | one second | 19:37 |
mhall119 | ok, so if I fire it from config-changed I need to set JUJU_RELATION_ID=$DID when I call db-relation-changed? | 19:37 |
marcoceppi_ | mhall119: yes | 19:38 |
mhall119 | instead of as $1 | 19:38 |
marcoceppi_ | $1 ? | 19:38 |
* marcoceppi_ checks code | 19:39 | |
mhall119 | it was passing the relation id from $(relation-ids db) as the first argument to db-relation-changed when it called it | 19:39 |
mhall119 | https://pastebin.canonical.com/98709/ | 19:39 |
mhall119 | I mean | 19:39 |
mhall119 | DID=$(relation-ids db) | 19:39 |
mhall119 | [ -n "$DID" ] && hooks/db-relation-changed $DID | 19:39 |
marcoceppi_ | right, set it as an environment variable before executing the hook | 19:39 |
mhall119 | ok | 19:40 |
mhall119 | and is that the only one that needs to be set? | 19:40 |
marcoceppi_ | db-relation-changed, unless programmed to, won't know what to do with $2 | 19:40 |
marcoceppi_ | $1 | 19:40 |
marcoceppi_ | mhall119: you typically need JUJU_REMOTE_UNIT | 19:40 |
marcoceppi_ | but I think you can get away without it | 19:40 |
marcoceppi_ | I'm waiting for my desktop to come back online to double check | 19:41 |
mhall119 | ok | 19:41 |
mhall119 | marcoceppi_: could it have ever worked passing it as the first parameter? Like I said, this was taken from one of the IS charms | 19:44 |
marcoceppi_ | mhall119: depends on how the db-relation-changed hook is set up | 19:44 |
marcoceppi_ | mhall119: doesn't look like it | 19:45 |
mhall119 | ah, ok, I see how it was being used before | 19:47 |
mhall119 | marcoceppi_: on, new question | 19:48 |
mhall119 | suppose I have a config parameter called bzr_revno, and in config.yaml for my charm I have it default to 1 | 19:48 |
mhall119 | then, I update my app's code and I update the charm's config to make bzr_revno default to 2 | 19:48 |
mhall119 | now I just found out that this doesn't actually call config-changed hook | 19:49 |
mhall119 | is there *any* hook that would react to a change in the default config value? | 19:49 |
marcoceppi_ | mhall119: hooks/upgrade-charm - but that's only reacting to an upgrade-charm event. What I recommend you do is call hooks/config-changed from the upgrade-charm hook | 19:50 |
mhall119 | and then when IS deploys a new version, upgrade-charm will be called? | 19:50 |
marcoceppi_ | mhall119: I don't know how IS does it, but if they use upgrade-charm, then yes, hooks/upgrade-charm will run | 19:51 |
mhall119 | ok, I'll check | 19:51 |
mhall119 | oh, hey, hooks/upgrade-charm is a symlink to hooks/config-changed | 19:52 |
mhall119 | so it should be called anyway | 19:52 |
marcoceppi_ | :) | 19:52 |
mhall119 | but it wasn't... | 19:52 |
mhall119 | and I'm not sure why | 19:52 |
marcoceppi_ | check the logs? | 19:52 |
mhall119 | https://pastebin.canonical.com/98709/ | 19:52 |
marcoceppi_ | also, run juju get api-website | 19:53 |
marcoceppi_ | see what juju thinks the value/defaults are | 19:53 |
marcoceppi_ | mhall119: there might be a thing where juju doesnt' update default values for configs when charms are upgraded | 19:54 |
marcoceppi_ | I could see that being the case | 19:54 |
mhall119 | so you'd have to still call uju set api-website-app bzr_revno=37 | 19:55 |
marcoceppi_ | mhall119: if what I described above is true, then yes | 19:55 |
mhall119 | ok | 19:55 |
mhall119 | between that and needing to set JUJU_RELATION_ID, I think that explains the failure I was having | 19:55 |
mhall119 | thanks marcoceppi_ | 19:56 |
marcoceppi_ | mhall119: np, let me know if you bump in to any other oddities | 19:56 |
mhall119 | don't worry, I will :) | 19:59 |
kurt_ | marcoceppi_: odd things going on with using maas-tags and deploying services http://pastebin.ubuntu.com/6210935/ | 20:20 |
kurt_ | my tag matches an existing the host, the one the bootstrap node is on, but juju still wants to spin up a second node | 20:21 |
jamespage | kurt_, specifying a tag does not force two services to deploy onto the same machine | 20:45 |
jamespage | kurt_, you have to use --to <machineid> todo that | 20:45 |
kurt_ | jamespage: but if those hosts are not yet known to juju, how do you accomplish deploying services to those nodes? | 20:47 |
jamespage | kurt_, deploy one service first, and then deploy using --to for subseqent services | 20:47 |
kurt_ | jamespage: how is handling moving on to the next node done then? | 20:48 |
jamespage | by co-locating services you are specifically telling juju that you are taking charge of where stuff lands | 20:48 |
jamespage | to add a new node just don't specify --to | 20:48 |
kurt_ | right. but if I want specific services to land on specific nodes, and they are unknown to juju, this sounds like a problem | 20:49 |
kurt_ | I get that I use --to. But when I need the next set of services on the the next node... | 20:49 |
kurt_ | how do I force juju to use a particular node? I thought that's what the purpose of tagging was | 20:50 |
kurt_ | tagging + constaints | 20:50 |
jamespage | kurt_, tagging is just a way of grouping servers | 20:52 |
jamespage | kurt_, juju will ask maas for a new unit which matches the provided tag | 20:52 |
kurt_ | ok, but it doesn't appear that was happening | 20:52 |
jamespage | kurt_, thats not what it sounded like above "juju still wants to spin up a second node" | 20:53 |
jamespage | thats the correct result | 20:53 |
kurt_ | jamespage: so if the node is already running that matches the tag, juju will look to deploy elsewhere without the "--to" function? | 20:55 |
jamespage | kurt_, yes | 20:55 |
kurt_ | that doesn't seem intuitive | 20:55 |
jamespage | the node is already allocated to a service | 20:55 |
jamespage | so juju won't by default place another service on it | 20:55 |
jamespage | think about a deployment with 1000 nodes | 20:56 |
jamespage | where 400 are for storage and 600 are for compute | 20:56 |
jamespage | I can assign tags for 'compute' and 'storage' to the different server types | 20:56 |
jamespage | and then use juju to target nova-compute to 'maas-tag=compute' and ceph to 'maas-tag=storage' | 20:56 |
jamespage | or servers could be tagged per physical availability zone, or switch or whatever | 20:57 |
kurt_ | Ok, when we talk about more servers in the tag group than a single server it makes sense | 20:57 |
kurt_ | but in the context of a single node, it didn't | 20:58 |
jamespage | tags don't really make sense in that context | 20:58 |
kurt_ | ok. makes sense. I am really looking for the one-size fits all hammer for juju. It just doesn't exist. | 21:00 |
kurt_ | my "--force-to" option :D | 21:00 |
=== natefinch is now known as natefinch-afk | ||
omgponies | is there a way to get the unit name from inside a hook | 23:15 |
omgponies | i.e. from juju status where I see 'services:\n elasticsearch\n ... units: elasticsearch/0 | 23:16 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!