| adam_g | what is jujud serving on port 17070 and where does it log? | 01:03 |
|---|---|---|
| adam_g | attempts to create an lxc container in a kvm hang waiting for a response for the lxc rootfs | 01:03 |
| adam_g | oh, nvm. there it goes | 01:11 |
| jacekn | is it possible to write a charm that can be deployed as principal but in other deploymetns as subordinate? Or do I need to write two separate charms? | 10:39 |
| rick_h_ | jacekn: since you have to change metadata.yaml you have to have two | 11:00 |
| rick_h_ | jacekn: I think with the new layers system you could reduce a ton of duplicate code across the two charms though | 11:01 |
| rick_h_ | jacekn: but you'd probably end up with two branches, one for the subordinate and one for the non-sub charm so you can merge/back and forth | 11:01 |
| jacekn | rick_h_: hmm that's what I throught. It's a bit annoying because of two codebases to maintain | 11:02 |
| jacekn | thanks | 11:02 |
| rick_h_ | jacekn: yes, the new publish workflow allowing you to keep the two charms in branches of the same repo will help some | 11:02 |
| rick_h_ | jacekn: but agree, annoying. What service is it if I can ask? | 11:03 |
| jacekn | rick_h_: I'm trying to write a collectd charm that can work as subordinate but also as principal so that it can relate to graphite server | 11:05 |
| jacekn | rick_h_: I want subordiante because it's much more convenient in bigger environments, deploy once and just add relations | 11:05 |
| rick_h_ | jacekn: gotcha | 11:06 |
| sparkieg` | jacekn: if all you need is a relation to graphite, then you can keep it as "just" a subordinate? | 11:09 |
| sparkieg` | jacekn: subordinates can relate to more than just the unit their co-deployed on | 11:10 |
| jacekn | sparkieg`: oh I did not know that was possible. So how do I stop that subordinate service from being deployed on the graphite server when I add relation? | 11:12 |
| sparkieg` | eugh, what kinda of hell is going on with my nick | 11:13 |
| sparkieg` | brb | 11:13 |
| sparkiegeek | jacekn: so in metadata.yaml specify the interface without the scope set to container | 11:15 |
| sparkiegeek | jacekn: example: https://api.jujucharms.com/charmstore/v4/trusty/landscape-client-12/archive/metadata.yaml | 11:15 |
| sparkiegeek | that might be a bit misleading because we only had an experimental server charm that implemented the other side of the relation, but AFAIK it Worked™ | 11:17 |
| jacekn | oh wow if that works it would be nice, thanks | 11:18 |
| jacekn | I'll try this method | 11:18 |
| bloodearnest | sparkiegeek, jacekn: that totally works, we do it with logstash-forwarder charm I think | 11:38 |
| bloodearnest | we also relate subordinates together, which apparently was never supposed to work (I was surprised when it did), but is very useful | 11:38 |
| jacekn | bloodearnest: relating subordinates kind of works but I found it unreliable | 11:39 |
| bloodearnest | e.g. conn-check subordinate charm relates to nrpe subordinate charm in order to setup conn-check as a nrpe check | 11:39 |
| jacekn | yeah but if you change config that propagates data throug relations that can cause problems, I saw many stale services where I had to recreate relaions to get things fixed | 11:40 |
| bloodearnest | jacekn, orly? How so? We havn;t seen any problems (although upstream did break it dev release a while back, but the fixed it quick enough) | 11:40 |
| bloodearnest | jacekn, interesing. In the conn-check<->nrpe relation, we just use it for a) we know nrpe is installed/configured and b) grab the nagios_host_context for the nrpe script | 11:41 |
| jacekn | bloodearnest: it was very visible with nrpe-external master, change something in one subordinate that relates to n-e-m and checks are not always updated | 11:41 |
| bloodearnest | it's not really 2 way relation | 11:41 |
| bloodearnest | jacekn, interesting | 11:42 |
| jacekn | bloodearnest: so if it works for you great, maybe you don't trigger conditions that cause problems | 11:43 |
| bloodearnest | reactive question: I have a test_requirments.txt file in my later, which I use in tox to run install deps for my tests. But charm build does not like that I have this file, says it should be in a base layer | 11:44 |
| bloodearnest | s/later/layer/ | 11:44 |
| bloodearnest | is there somewhere else it should be? | 11:45 |
| bloodearnest | I don't really want to add test deps to wheelhouse.txt? | 11:45 |
| marcoceppi | jacekn: that's how the current nrpe charms work, etc | 12:10 |
| jacekn | marcoceppi: yes I'm just saying that it's unreliable, on multiple occasions things got stale or hooks were not triggered for me | 12:12 |
| marcoceppi | jacekn: reactive framework should help with that | 12:13 |
| bloodearnest | reactive tactics question: how do I specify a tactic to use? I'm trying to use the wheelhouse tactic from the latest charm-tools release, but I seem to be using copy tactic, afaics | 13:08 |
| bloodearnest | s/reactive/build | 13:11 |
| bloodearnest | actually, I am totatally not sure which tactic I'm using, but no wheelhouse dir is being built, so I assume non wheelhouse | 13:23 |
| lazypower | hmm | 13:31 |
| lazypower | bloodearnest : lets follow up with cory_fu when he comes in. I haven't used the wheelhouse tactic yet | 13:31 |
| lazypower | but i know he put some work into that last week | 13:32 |
| bloodearnest | lazypower, it looks like it should auto-detect the tactic from the files present? | 13:32 |
| lazypower | bloodearnest: indeed - there are also tactic overrides you can specify | 13:33 |
| lazypower | but i'm not sure what the expected behavior is with wheelhousing deps. it sounded like it was automagic | 13:34 |
| bloodearnest | lazypower, ah! I did an upgrade, but charm-tools 1.9.3 is kept back for some reason. Manually installing... | 13:35 |
| lazypower | ah that'll od it | 13:36 |
| lazypower | *do | 13:36 |
| bloodearnest | lazypower, now we're cooking on gas | 13:38 |
| lazypower | bam! | 13:38 |
| cory_fu | bloodearnest: Did the update fix the "should be in a base layer" issue with test_requirements.txt? You should be able to have that alongside wheelhouse.txt | 14:16 |
| bloodearnest | cory_fu, I think so, yes | 14:26 |
| cory_fu | Ok, good | 14:35 |
| gnuoy | jamespage, do you have any feelings about https://code.launchpad.net/~gnuoy/charms/trusty/openvswitch-odl/tox/+merge/279126 | 14:36 |
| gnuoy | beisner, ^ | 14:36 |
| jamespage | gnuoy, that will do unit tests and lint by default | 14:39 |
| beisner | gnuoy, so the test runner didn't recognize that make target name as a unit test, because it disregards make target name and searches for one that looks like the unit test instead. | 14:46 |
| beisner | it is that way because make target names sometimes vary across charms, though we have now normalized at least the os-charms | 14:47 |
| Prabakaran | Hi mbruzek, This is regarding Platform Symphony Readme.md file, as per the recent comment received from you suggesting "Readme.md file should telll the juju user Juju users that they have to pay for the platform symphony software before they can download it” | 15:24 |
| Prabakaran | So I have mentioned in the README.md file “For details on Platform Symphony usage and its benefits please visit [Product Page.] [product-page] For information on purchasing, please visit IBM Fix Central][fixcentral]. To acquire and download Platform Symphony, login to fixcentral as an entitled user. If the user doesn’t have entitlement, they won't be able to see fixes in fixcentral.” | 15:24 |
| Prabakaran | Is it suffices to go or should I mention more details on this? | 15:25 |
| mbruzek | Hello Prabakaran. I have an IBM account and tried to follow the steps in the README. When I attempted to download Platform Symphony I was presented with an error message | 15:26 |
| mbruzek | Prabakaran: so my account was not "enabled" to download the software, which if I were a user would be frustrating and I don't know how to solve that problem. | 15:27 |
| mbruzek | Prabakaran: I got the error message: "No applicable IBM support agreement found for one or more of the products you selected." | 15:27 |
| mbruzek | I don't know how to get an agreement or resolve that download error | 15:28 |
| mbruzek | only by reading the readme. | 15:28 |
| Prabakaran | Matt, if user don't have entitlement, they won't be able to see fixes in fixcentral for the updates so anyone who wants to try this out has to already be a paying symphony customer | 15:28 |
| Prabakaran | if they are a paid symphony customer, they should be able to see the fp on fixcentral | 15:29 |
| Prabakaran | So I have mentioned in the README.md file “For details on Platform Symphony usage and its benefits please visit [Product Page.] [product-page] For information on purchasing, please visit IBM Fix Central][fixcentral]. To acquire and download Platform Symphony, login to fixcentral as an entitled user. If the user doesn’t have entitlement, they won't be able to see fixes in fixcentral. | 15:30 |
| Prabakaran | Is it suffices to go? | 15:31 |
| mbruzek | Prabakaran: But I have a user ID on fix central so I assumed that I could download that software, I could not and there was no clear path for me to get that software. I don't think the current readme was sufficient | 15:31 |
| mbruzek | Prabakaran: Someone using Juju and interested in this Platform Symphony would need clear instructions on how to get it. | 15:32 |
| mbruzek | Prabakaran: my point is *I* was not able to figure it out and I don't think others will be able to. | 15:33 |
| Prabakaran | k then i will include more information on how to get entitlement to download these fixpacks when users is getting error while downloading fp. | 15:34 |
| Prabakaran | i will include these information in readme with required links | 15:35 |
| mbruzek | Prabakaran: I would mention this earlier near the top, so the reader has a chance to understand what is required here. REading from top down it was not clear. | 15:35 |
| mbruzek | Prabakaran: the rendered README is here: https://jujucharms.com/u/ibmcharmers/ibm-platform-symphony | 15:36 |
| mbruzek | Prabakaran: most of the charms either have the software included or they download the software for the user. It should be clear that this one is a little bit different and requires more set up. | 15:37 |
| mbruzek | Prabakaran: it was not accurate that a user will need to create an "apt repository" because the software needs to be hosted on a HTTP server such as apache. | 15:39 |
| Prabakaran | Still i have not pushed my readme changes to trunk branch to clarify with you i was asking wherther it is suffices to go...I will mention more detail on getting entitlement with the link as discussed | 15:41 |
| Prabakaran | now i will be removing the apache config part as i am implementing juju action on this. | 15:42 |
| mbruzek | a juju action to download the software? | 15:42 |
| Prabakaran | ya | 15:43 |
| lazypower | It seems like that should be charm config, not an action | 15:43 |
| lazypower | consider teh scenario | 15:43 |
| mbruzek | Prabakaran: How will an action prevent the need for an http server? | 15:43 |
| lazypower | you've deployed the service, and you get a message in juju status that you have to now `juju action do ibm-platform-symphony/0 download-software` | 15:44 |
| lazypower | what's this doing? why isn't the charm doing this for me automatically? I've input all the requested configuration, the rest should 'just work' (tm) | 15:44 |
| mbruzek | Prabakaran: I am not clear on the change to an action. | 15:44 |
| lazypower | mbruzek: i agree, there are unresolved technical questions, and workflow resons that is not a good path forward. | 15:45 |
| Prabakaran | I was referring SFTP implemention using download action to the install packages from an SFTP server that can be setup in your network. "juju action do <unit name> download username=<sftp user> password=<sftp password> packagedir=<full path to repository directory> host=<sftp host name>" | 15:49 |
| Prabakaran | so i will be changing readme also accordingly | 15:50 |
| mbruzek | Prabakaran: An action is a one time thing, and is not traditionally used to install the software. | 15:50 |
| mbruzek | Prabakaran: My understanding is the user would still need to host the Platform Symphony software *somewhere* after they purchased and downloaded it. | 15:51 |
| mbruzek | Prabakaran: and they would still need details like that in the charm. | 15:51 |
| mbruzek | how to set up sftp server | 15:52 |
| mbruzek | Prabakaran: I would keep that information in the configuration so the charm could be deployed from both the GUI and the command line. Currently the GUI does not provide a way to run actions on a charm. | 15:53 |
| Prabakaran | K ...i have one more question also... set –e was used as per your advise using “|| true” but unfortunately though I was getting non-existant error so in some places I was forced to used set +e. Is there any alternative way to alternative methods in order to resolved this non-existant error? | 15:59 |
| mbruzek | Prabakaran: the charm used "set +e" in the majority of the code, defeating the best practice for bash scripts. I would isolate the error commands down a little more and only use "|| true" where absolutely necessary. Check the commands for options to "force" or return without error (silent perhaps), but I understand some will not have the option. If you can't isolate those commands use if, then, else to prevent the command from being called. If [ f | 16:04 |
| mbruzek | Prabakaran: this is a very common pattern in the charms. The trick to writing good idempotent charms is to figure out what causes the error and use appropriate code blocks to avoid these errors/problem. | 16:05 |
| Prabakaran | Can anyone help me how to deploy openjdk layer charm available in the link https://github.com/juju-solutions/layer-openjdk as i am not able to find the deployment steps in the README.md file. | 16:17 |
| Prabakaran | Hi kwmonroe | 16:17 |
| Prabakaran | how to deploy openjdk layer charm available in the link https://github.com/juju-solutions/layer-openjdk as i am not able to find the deployment steps in the README.md file. | 16:18 |
| lazypower | Prabakaran: that's a layer, and is not necessarily deployable by itself | 16:18 |
| lazypower | Prabakaran: that's intended to be a runtime layer, as in a lower layer in a layered charm. You would include it in a layers.yaml file in your charm layer and subscribe to the events raised by the openjdk-layer | 16:18 |
| lazypower | Prabakaran: There's been a lot of info published about this recently, let me fetch you some links | 16:19 |
| lazypower | https://jujucharms.com/docs/stable/authors-charm-building | 16:19 |
| lazypower | https://insights.ubuntu.com/2015/10/29/now-youre-charming-with-layers/ | 16:20 |
| lazypower | http://blog.dasroot.net/2015-charming-2-point-oh.html | 16:21 |
| lazypower | Prabakaran ^ | 16:21 |
| lazypower | all three are excellent reads on the subject | 16:21 |
| Prabakaran | Thanks for the links lazypower .. i have already tried to deploy this openjdk using charm build command | 16:22 |
| Prabakaran | i was not able to check the openjdk service up and running | 16:22 |
| lazypower | Prabakaran: well its not a service, its just installing the JVM/JRE | 16:23 |
| Prabakaran | wondering wherther i followed correct steps while deploying | 16:23 |
| lazypower | Prabakaran: this isn't fully ready for release yet i dont think. there's noi Readme in the layer repository | 16:23 |
| lazypower | i know this is being worked on by folks over at IBM, WildFly, and a couple others to provide a java abstraction layer | 16:23 |
| lazypower | giving folks the ability to mix/match pick/choose their componentry when charming and target a JVM | 16:24 |
| lazypower | kwmonroe is currently out of office, but should be returning later today | 16:24 |
| Prabakaran | I am working on IBM Java SDK for charming with layers. For implementation i referred OpenJDK layer, i want to deploy and check IBM Java layer charm is working or not. So for my understanding i want to test how openjdk works. Could pls help me on this?... | 16:31 |
| mbruzek | Prabakaran: Do you get a jvm when you deploy the openjdk charm after building it? | 16:33 |
| mbruzek | Prabakaran: it may not be *running* but is it on the system somewhere? | 16:33 |
| Prabakaran | After deployment in the container i was able to see openjdk was up and running but when sshd into the container java -version command is not working | 16:36 |
| mbruzek | Prabakaran: Perhaps java is not set up for the user. Search for the java binary file | 16:37 |
| Prabakaran | k | 16:40 |
| roadmr | heya folks! I'm playing with the leadership feature and hooks. Is there a way to force a leader change without destroying the currrent leader unit? | 16:43 |
| roadmr | I don't care which unit ends up being leader, just want a new one to be elected | 16:43 |
| rick_h_ | fwereade: ^ any idea there please? | 16:43 |
| fwereade | roadmr, you could ssh in and stop the unit that's currently leader (`sudo stop jujud-unit-foo-123`); wait for failover; and start it again? | 16:47 |
| fwereade | roadmr, you can see when that's happened with `juju run --service foo is-leader` | 16:48 |
| fwereade | roadmr, let me know if that helps | 16:49 |
| roadmr | fwereade: sure thing, trying it right now | 16:50 |
| roadmr | fwereade: so it mostly works with one exception: I wanted to test the current leader getting the leader-elected hook so, so it can remove some files that should only exist on the leader | 16:54 |
| roadmr | fwereade: if I stop the agent and restart it, the former leader comes back thinking it's still leader (I mean, is-leader says False, which is correct, but it never got the leader-changed event so my hook which manages the files never fired) | 16:55 |
| === sg is now known as Guest56867 | ||
| fwereade | roadmr, maybe I misunderstand: shouldn't you do that on the leader-settings-changed hook? getting that should indicate that you're definitely not leader | 16:56 |
| roadmr | fwereade: maybe; I'm just learning how/when these hooks fire :) so you're saying only the leader will get leader-elected? | 16:57 |
| fwereade | roadmr, exactly | 16:57 |
| fwereade | roadmr, https://jujucharms.com/docs/1.25/authors-charm-leadership might be helpful | 16:57 |
| roadmr | fwereade: right, I probably misread that to mean that all units would get leader-elected and should call is-leader accordingly to see if they are or not | 16:58 |
| roadmr | fwereade: so in my scenario, when the former leader rejoins, it should get leader-settings-changed? are events queued for unresponsive units? | 16:59 |
| fwereade | roadmr, yeah, leader-settings-changed will be one of the first hooks that executes | 17:00 |
| roadmr | fwereade: great, that should do it then! ok, mangling my charm, give me a couple of mins... | 17:00 |
| fwereade | roadmr, FWIW, if you're interested: all the queueing actually happens on the unit agent -- when it wakes up it asks the state server how the world should look, and then decides what hooks need to run to bring your unit up to date | 17:01 |
| roadmr | fwereade: that sounds nicer than a queue of events in the state server | 17:02 |
| fwereade | roadmr, so if a counterpart changes relation settings 10x while you were down, you'd just see a single change when you woke up | 17:02 |
| fwereade | roadmr, yeah, exactly | 17:02 |
| roadmr | haha, sounds like me slamming "juju set blah" :) | 17:02 |
| geetha | Hi, I am trying to use HAProxy as reverseproxy for my charm, but my charm is using SSL and will automatically redirect to https, when I try to access my charm through HAProxy, its giving me HTTPS connection error. | 17:03 |
| beisner | jamespage, gnuoy - i can't file a bug against https://bugs.launchpad.net/charms/+source/openvswitch-odl/+filebug -- probably due to ingestion/namespace thing. | 17:10 |
| beisner | jamespage, gnuoy - openvswitch-odl test is failing with: http://paste.ubuntu.com/13602264/ | 17:10 |
| beisner | jamespage, gnuoy - added as a card in place of a bug for now. | 17:12 |
| gnuoy | beisner, I've seen that bug before | 17:13 |
| gnuoy | maybe it was in another charm | 17:13 |
| beisner | gnuoy, indeed, tcp:172.17.113.167:6633 tcp:172.17.113.167:6633 != tcp:172.17.113.167:6633 | 17:15 |
| beisner | i've not dug into it, just wanted to raise a bug to get it on the radar. not sure if it's a test bug or a charm bug yet. | 17:16 |
| mbruzek | geetha: It should be possible to set up haproxy with https | 17:16 |
| mbruzek | geetha: I have not done this personally but I know there are configuration options for the ssl_key | 17:17 |
| mbruzek | geetha: I would read up on haproxy and ssl or HTTPS for more information such as: https://www.digitalocean.com/community/tutorials/how-to-implement-ssl-termination-with-haproxy-on-ubuntu-14-04 | 17:19 |
| mbruzek | geetha: If the charm does not do something you need you can branch it, fix the code and submit the improved code to help make the haproxy charm better, but I suspect it does work with https, just needs to be configured properly. | 17:21 |
| geetha | mbruzek, there is no configuration option for https in HAProxy charm. could you please suggest me how to configure https in HAProxy charm. | 17:26 |
| mbruzek | geetha: As I said earlier, I have not done this before. You will have to read up on the charm's readme, a quick glance shows ssl_key and ssl_cert configuration options. | 17:30 |
| mbruzek | geetha: a Google search shows me another doc http://seanmcgary.com/posts/using-sslhttps-with-haproxy | 17:33 |
| === JoshStrobl is now known as JoshStrobl|AFK | ||
| roadmr | fwereade: it works \o/ using leader-settings-changed was the missing link. On coming back, the former leader correctly updates its files to be just a follower :) Thanks so much for your help! | 17:48 |
| lazypower | geetha: It sounds like we need to file a bug against eth HAProxy charm | 17:56 |
| lazypower | geetha: I don't beleive the haproxy charm supports ssl out of the box in its current form | 17:56 |
| lazypower | ah disregard, mbruzek si correct | 17:58 |
| lazypower | it accepts a base64 encoded certificate / key in the haproxy charm configuration. Which will ssl terminate the proxy frontend. | 17:58 |
| fwereade | roadmr, excellent, glad to help | 18:24 |
| adam_g | using the local kvm provider, is it possible to use a locally modified image instead of one fetched from simplestreams? | 19:26 |
| === JoshStrobl|AFK is now known as JoshStrobl | ||
| rick_h_ | adam_g: 'local kvm provider'? you mean lxc? | 19:28 |
| adam_g | rick_h_, no, i mean the local provider /w 'container: kvm' | 19:28 |
| rick_h_ | adam_g: ah ok | 19:28 |
| rick_h_ | adam_g: no, it's not built in to do that atm | 19:29 |
| adam_g | rick_h_, bummer | 19:29 |
| rick_h_ | adam_g: there's plans to use lxd and enable some of that for dev in the future but not here yet | 19:30 |
| rick_h_ | adam_g: and tied to lxd vs kvm | 19:30 |
| adam_g | ah | 19:30 |
| lazypower | rick_h_: correct me if i'm wrong, but if you're using MAAS you can specify one of the custom images? | 21:32 |
| rick_h_ | lazypower: if you build streams from that image list | 21:32 |
| lazypower | s/you can/cant you specify/ | 21:32 |
| rick_h_ | lazypower: see https://jujucharms.com/docs/1.24/howto-privatecloud | 21:33 |
| lazypower | Yeah, i thought so. And thats right, gotta write streams for it | 21:33 |
| * lazypower nods | 21:33 | |
| lazypower | ok, cool cool, thanks rick_h_ :) | 21:33 |
| rick_h_ | lazypower: np | 21:33 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!