=== almaisan-away is now known as al-maisan | ||
imbrandon | m_3: its ready for the charm store and been reviewed :) | 09:51 |
---|---|---|
imbrandon | i have it manually mirror both places atm, and then i setup a juju-drupal project so it would pull every few minutes, but it pulls to a diffrent branch | 09:51 |
imbrandon | soo | 09:52 |
=== al-maisan is now known as almaisan-away | ||
=== Guest394 is now known as jrgifford | ||
yolanda | hi, good afternoon | 13:04 |
yolanda | one question, how can i debug a db-relation-joined hook? it doesn't give any error, but the code in the hook doesn't seem to be executing | 13:05 |
sc-rm | Yolanda: Which charm are you trying to use? | 13:14 |
yolanda | sc-rm, i'm building a new one | 13:14 |
yolanda | i'm adding a relationship to postgresql | 13:14 |
yolanda | and i have a hook db-relation-joined | 13:14 |
sc-rm | yolanda: Which language do you code your charm im? | 13:15 |
yolanda | sc-rm, its' just bash | 13:15 |
sc-rm | youlanda: okay, then I would just do some echo in the places I suspect to not get hit | 13:16 |
yolanda | sc-rm, but how can i reach to that relationship in debug? | 13:17 |
sc-rm | youlanda: I guess you do something like juju add-relation postgresql you-charm? | 13:17 |
yolanda | sc-rm, hes, juju add-relation openerp postgresql:db | 13:18 |
yolanda | the relationship seems to be creating, doesn't give any error, but there is some code to change a config file and isn't triggered | 13:18 |
imbrandon | yolanda: anything in juju debug-log | 13:19 |
yolanda | juju add-relation openerp postgresql:db -e local | 13:20 |
yolanda | 2012-04-10 15:20:06,162 INFO Added pgsql relation to all service units. | 13:20 |
yolanda | 2012-04-10 15:20:06,163 INFO 'add_relation' command finished successfully | 13:20 |
yolanda | nothing in logs | 13:20 |
imbrandon | fun | 13:20 |
sc-rm | youlanda: have you looked at https://juju.ubuntu.com/docs/write-charm.html ? | 13:20 |
imbrandon | sc-rm: i'm fairly sure this is about yolanda's what the 3rd or 4th one :) | 13:21 |
yolanda | i've read it | 13:21 |
yolanda | i have successfully create the install, start and stop | 13:21 |
yolanda | but the db-relation-joined gives me problems | 13:22 |
* SpamapS stretches and yawns | 13:22 | |
imbrandon | moins SpamapS | 13:22 |
SpamapS | yolanda: have you used 'juju debug-hooks' yet ? | 13:22 |
SpamapS | yolanda: I think its what you're looking for | 13:22 |
yolanda | SpamapS, yes, i used it, but how can i retry a relationship? i never get into the db-relation-joined hook | 13:23 |
yolanda | i remove the relationship, i add it again, but i'm not placed into the hook | 13:23 |
SpamapS | yolanda: if the state is just 'up', you need to remove/add it again. If it is in error, you can use 'juju resolved --retry unit/X db' | 13:23 |
SpamapS | yolanda: are you sure you're not still *in* the hook? | 13:24 |
yolanda | the status of the relationship is that: | 13:24 |
yolanda | relations: | 13:24 |
yolanda | db: | 13:24 |
yolanda | - openerp | 13:24 |
yolanda | and that for openerp: | 13:24 |
yolanda | relations: | 13:24 |
yolanda | db: | 13:24 |
yolanda | - postgresql | 13:24 |
yolanda | is that ok?= | 13:24 |
SpamapS | yolanda: can you install 'pastebinit' and just pipe the whole output of 'juju status' to it | 13:25 |
SpamapS | ? | 13:25 |
SpamapS | bcsaller1: good morning? | 13:25 |
yolanda | ok | 13:25 |
yolanda | http://paste.ubuntu.com/923267/ | 13:27 |
SpamapS | yolanda: thanks reading | 13:29 |
yolanda | thx | 13:30 |
SpamapS | yolanda: so yeah I would expect remove/add to throw you back into db-relation-* | 13:30 |
SpamapS | yolanda: perhaps you need to exit debug-hooks and go back in ? | 13:30 |
yolanda | SpamapsS, so i just do a debug-hooks openerp/0 | 13:30 |
yolanda | and then remove add relation? | 13:31 |
yolanda | it doesn't place me in the hook, i just remain in the initial screen | 13:32 |
SpamapS | yolanda: it should definitely place you almost immediately in the joined hook | 13:33 |
yolanda | doesn't do it :( | 13:33 |
SpamapS | yolanda: can you run debug-log in another window and confirm that joined is being executed? | 13:33 |
yolanda | nothing in juju debug-log | 13:34 |
yolanda | SpamapS, so something is wrong with my juju charm? seems that nothing is triggered | 13:37 |
SpamapS | yolanda: can you pastebin metadata.yaml ? | 13:38 |
yolanda | sure | 13:40 |
yolanda | http://paste.ubuntu.com/923287/ | 13:41 |
SpamapS | yolanda: I suspect there is something more sinister going on. Can you try 'juju resolved --retry openerp/0 db' | 13:42 |
SpamapS | perhaps the new status output is not as obvious about errors | 13:42 |
yolanda | The relation 'db' has no unit state for 'openerp/0' | 13:42 |
yolanda | 2012-04-10 15:42:32,817 ERROR The relation 'db' has no unit state for 'openerp/0' | 13:42 |
SpamapS | yolanda: is this right after you added it? | 13:44 |
SpamapS | yolanda: can you run 'dpkg -l juju' on your machine, and on openerp/0 ? are they the same version? | 13:44 |
yolanda | SpamapS, yes, just after adding it | 13:45 |
yolanda | let me check the versions | 13:45 |
yolanda | mm, not same version | 13:45 |
yolanda | local is bzr504 | 13:45 |
yolanda | openerp/0 is bzr518 | 13:45 |
yolanda | but i'm using local containers in juju, why are they different? | 13:46 |
yolanda | i'm using ppa as juju-origin, perhaps i need to use repo? | 13:50 |
SpamapS | yolanda: well technically there should be no incompatibility between those versions | 13:50 |
SpamapS | but in reality, I've seen a lot of subtle breakage with stuff like this | 13:51 |
SpamapS | yolanda: just upgrade your local machine. 518 just landed yesterday. | 13:51 |
SpamapS | yolanda: if that fixes it.. please let me know. We really have to stop breaking things in backward incompatible ways. | 13:51 |
yolanda | let me see | 13:52 |
yolanda | do i have to regenerate all the environment? | 13:53 |
SpamapS | yolanda: hopefully no | 13:59 |
SpamapS | yolanda: you might want to destroy the existing services | 13:59 |
SpamapS | yolanda: but I'd try just with a new client binary | 13:59 |
yolanda | SpamapS, it gave some error, i'm destroying environment and creating new | 14:00 |
SpamapS | yolanda: could you pastebin the error? | 14:03 |
SpamapS | Thats actually quite critical since we plan to land one more version in precise .. I think | 14:03 |
yolanda | too late... | 14:03 |
yolanda | i just destroyed it | 14:03 |
yolanda | i'm running it again, let me check if the error shows now | 14:03 |
marcoceppi | I wonder would a "good" feature be to bootstrap a remote environment locally? So in the environments.yaml EC2 stanza say "local-bootstrap: true". Juju would spin up an LXC container instead of doing so in the intended cloud | 14:07 |
marcoceppi | I know you loose portability, but you could save money that way | 14:08 |
SpamapS | marcoceppi: you can save money with 'placement: local' | 14:11 |
SpamapS | marcoceppi: just remember to remove it after the first deploy. ;) | 14:12 |
marcoceppi | SpamapS: that deploys to the bootstrap, correct? | 14:12 |
SpamapS | marcoceppi: though I have thought a lot about how to do de-centralized juju.. and one way I was wondering about was to run the zookeeper locally and have juju agents just connect to 127.0.0.1:xxxx and have juju clients ssh in and "wake up" the agents with the port forwarded. Feels very hacky. | 14:13 |
SpamapS | marcoceppi: correct | 14:13 |
_mup_ | juju/ns-from-env r479 committed by kapil.thangavelu@canonical.com | 14:13 |
_mup_ | merge trunk | 14:13 |
marcoceppi | SpamapS: Right, I've been pondering different de-centralized ways to manage juju this morning, at least at a high level, and I couldn't really fathom another way than what's running now | 14:14 |
SpamapS | marcoceppi: I think ultimately zookeeper should be ditched for a map of the nodes kept somewhere (S3 for instance).. and agents should just be responsible for their own data, and use 0mq to talk to eachother and feed their status back to object storage per-unit. I'm not convinced anymore that zookeeper is needed, though it *does* make things simpler. | 14:14 |
SpamapS | marcoceppi: It makes me nervous how much vital information is stored in ZK. | 14:16 |
SpamapS | anyway, time to wake up the family | 14:17 |
marcoceppi | SpamapS: so, an alternative would be to just maintain a map in the bucket then all the relation based data that may or may not be sensative is on each node within the juju agent? | 14:17 |
SpamapS | marcoceppi: right | 14:17 |
marcoceppi | interesting | 14:18 |
SpamapS | basically juju client would just much with a tree of intent (much like it does now) but instead of having ZK there with all of the other resulting data (like relation-set produced data), the agents would just keep that to themselves and message it back and forth. | 14:19 |
SpamapS | s/much/muck/ | 14:19 |
SpamapS | that way there's no one server that you compromise to pwn the entire environment. | 14:20 |
* SpamapS goes upstairs to start shaking his toddler awake | 14:20 | |
marcoceppi | Is there any way to tell which unit is the "head" in a peer relation? | 14:42 |
hazmat | smoser, thanks for the clue yesterday re cloud-init, it appears that rsyslog fails to start in the lxc cloud-image | 14:43 |
hazmat | wonders what the heck /usr/bin/whoopsie is | 14:43 |
SpamapS | hazmat: just don't step on it | 14:44 |
jcastro | hey SpamapS | 14:49 |
jcastro | let's talk store for precise now that contest is nearly done | 14:49 |
SpamapS | yeah I started working on testing all the charms last week | 14:53 |
jcastro | SpamapS: anything sticking out as a problem? ie. do you think this will be a big deal or just flipping the switch? | 14:55 |
SpamapS | jcastro: no, initially most work fine. :) | 14:58 |
SpamapS | jcastro: we should have a period where we invite ~charmers to fix them all, and I need to ask the launchpad guys what will happen when there is already a branch present.. but i think we'll have this wrapped up by next week. | 14:59 |
jcastro | ok | 14:59 |
smoser | hazmat, why no syslog ? | 15:04 |
smoser | do you know ? | 15:04 |
smoser | as i might have been about to ask SpamapS on a similar question | 15:05 |
smoser | oh. | 15:05 |
smoser | lxc | 15:05 |
hazmat | smoser, haven't really debugged it yet.. it fails to start | 15:05 |
smoser | /dev/log | 15:05 |
hazmat | smoser, yup | 15:05 |
hazmat | probably | 15:05 |
smoser | lxc container does not probably have access to that device node | 15:05 |
smoser | does that make sense to you? would that be able to kick you in the right direction to fix it ? | 15:05 |
smoser | or you want more direction | 15:05 |
smoser | s/fix/test change/ | 15:06 |
hazmat | smoser, doesn't exist in the container | 15:06 |
smoser | right. because it can't get created | 15:06 |
smoser | because lxc wont let it. | 15:06 |
smoser | er... app armour or what not. | 15:06 |
avoine | hazmat have you found a bug about rsyslog on lxc? | 15:07 |
avoine | I saw that because it prevent cloud-setup to be fire | 15:08 |
hazmat | avoine, well it doesn't start in particular on the cloud-image which causes issues for cloud-init's config which juju relies on. | 15:08 |
hazmat | smoser, at this point i'm just going to punt till after precise, it was a big change to even try at this point, even though the code delta is small | 15:09 |
hazmat | smoser, as far as hints.. though i'm all ears | 15:09 |
smoser | oh. wait... maybe we can fix this via cloud-init data. | 15:10 |
smoser | wait. | 15:10 |
smoser | no | 15:10 |
smoser | rsyslog is going to use /dev/log | 15:10 |
smoser | we need access to that in the container | 15:10 |
hazmat | smoser, looks like just killing the app armor profile for rsyslog gets it to start | 15:11 |
hazmat | and then cloud-init config is happy | 15:11 |
smoser | hm.. | 15:11 |
smoser | iwould have thought youneeded access specifically to mknod from one of the lxc configs. | 15:11 |
smoser | but... ok. | 15:11 |
smoser | my suspicion is that /dev/log is not containerized | 15:12 |
smoser | hallyn or stgraber can probably provide insight on what to do here. | 15:12 |
hazmat | smoser, not sure, i don't see the messages from the container in the host.. | 15:13 |
smoser | oh. wait. | 15:13 |
smoser | /dev/log is just a socket. | 15:13 |
smoser | and rsyslog sets it up | 15:13 |
smoser | so, yeah, containerized. | 15:14 |
smoser | i was thinig it was a dev node. | 15:14 |
hazmat | smoser, so its just the app aromor pollicy | 15:16 |
hazmat | oh.. double typo, extra points | 15:16 |
bac | when i try to use 'juju bootstrap' with --constraints specifying more than one constraint i get errors: http://paste.ubuntu.com/923438/ | 15:20 |
bac | i'm following the example shown at https://juju.ubuntu.com/DeprecatedEnvironmentSettings | 15:21 |
avoine | hazmat: smoser: could you tell me if you fill a bug about the rsyslog issue? I would like to follow it. | 15:29 |
smoser | hazmat, ^ can you file one? open it against lxc. | 15:29 |
smoser | i'd do it, but you ahve more context. | 15:29 |
hazmat | smoser, avoine sure | 15:31 |
avoine | thank you | 15:31 |
hazmat | avoine, smoser its bug 978147 | 15:56 |
_mup_ | Bug #978147: rsyslogd fails to start in cloud template <lxc (Ubuntu):Confirmed> < https://launchpad.net/bugs/978147 > | 15:56 |
avoine | ok thanks again | 15:56 |
imbrandon | SpamapS: man , dunno if your a rock fan or not , but I just bought the new Halestorm album, if you are even a little, dont hesiate get it now! dude most sophmore albums suck, or at least dont stack up to the first, this one blows the first out of the water | 16:21 |
SpamapS | imbrandon: I'm disconnected from all new music unfortunately. I still just rock out to Pantera's Vulgar Display of Power when I need something with kick | 16:22 |
imbrandon | yea definaltly along the same lines, dude i broke my nose at that tour | 16:23 |
imbrandon | on* | 16:23 |
imbrandon | in the moshpit | 16:23 |
SpamapS | haha | 16:23 |
SpamapS | Its my hockey music | 16:23 |
imbrandon | :) | 16:23 |
SpamapS | ONETWOTHREEFOUR!!! | 16:23 |
imbrandon | hells yea there is some reason i knew i liked you hahahaha j/k | 16:23 |
SpamapS | Honestly lately I just listen to Billy Joel. I need to chill more than I need to get amped up. | 16:24 |
imbrandon | :) | 16:24 |
imbrandon | they got a little chill on the album, like "heres to us" more like an 80's powerbalid | 16:25 |
imbrandon | lemme see if i can find it on amazon, i think they let you preview it | 16:25 |
imbrandon | http://www.amazon.com/Heres-To-Us-Explicit/dp/B0070R94SK | 16:27 |
imbrandon | yea there is a preview | 16:27 |
imbrandon | its the slower side of the album | 16:27 |
imbrandon | this is the one that made me buy the album instantly http://www.amazon.com/American-Boys/dp/B007ONQC4Y | 16:28 |
imbrandon | and of course i have the first album too and seen em in concert 2 times now at Rockfest KC, bigest one day music fest in america ( 55k people, one day, 12 hours, 2 stages, hell yea ) | 16:30 |
marcoceppi | err, I think you guys broke compatibility between juju versions 0.5+bzr504-1juju4~oneiric1 and 0.5+bzr518-1juju5~oneiric1 | 16:36 |
marcoceppi | In this case, should I just run a apt-get upgrade on the bootstrap? juju status is giving a weird error ATM and I think it's because the juju versions mismatch | 16:37 |
SpamapS | marcoceppi: certainly worth a try. Can you maybe document the errors so we can try to avoid such madness in the future? | 16:39 |
* SpamapS really really wants to implementat version stability | 16:39 | |
marcoceppi | SpamapS: I understand I'm in the ppa version, so I expect some breakage. What would be best for you guys? | 16:41 |
marcoceppi | as far as documenting errors and giving enough details | 16:41 |
_mup_ | juju/must-specify-relation-name r516 committed by jim.baker@canonical.com | 16:42 |
_mup_ | Client side check | 16:42 |
m_3 | SpamapS: hey... so blacklisting charms from testing | 16:44 |
m_3 | I blacklist the openstack ones because they're already tested elsewhere | 16:44 |
m_3 | but take buildbot-slave for example... | 16:45 |
m_3 | it requires config to even install | 16:45 |
marcoceppi | m_3: should charms like that include a default.yaml for config then? | 16:46 |
m_3 | I can either blacklist such things or bump them back to the authors and require them to have enough default config to at least "start" | 16:46 |
m_3 | I'm tempted to bump them back, but just wanted peeps opinions on that... so we're requiring all charms to at least start when we're reviewing | 16:47 |
m_3 | marcoceppi: yeah, we haven't been too strict with it though | 16:47 |
m_3 | marcoceppi: really it's enough default values in the config.yaml... not a default.yaml to pass | 16:48 |
m_3 | that's more work for the testing framework | 16:48 |
marcoceppi | m_3: I think it's a good idea to have enough defaults in config.yaml then | 16:48 |
m_3 | yup | 16:48 |
m_3 | it's strange for some charms though... buildbot-slave is a perfect example | 16:48 |
m_3 | it's a general tool... defaults don't make sense really, they'd be put there _just_ for our testing purposes | 16:49 |
m_3 | I'm cool with that tho | 16:49 |
marcoceppi | SpamapS: http://paste.ubuntu.com/923559/ | 16:50 |
marcoceppi | SpamapS: so upgrading the bootstrap didn't work, I guess I'm just going to have to destroy/re-bootstrap? | 16:52 |
* marcoceppi wonders how to best upgrade the omg env without downtime | 16:55 | |
m_3 | marcoceppi: dang, we've got to stop breaking the api between the cli and existing bootstrapped environments | 16:55 |
m_3 | it's almost like you've gotta balance across two environments | 16:55 |
marcoceppi | m_3: I have a question, it might be easy to hold the package on my local machine, but doesn't each deployment/add unit install the current juju package? | 16:55 |
marcoceppi | So holding the package on my machine wouldn't really save me in the future from the units being out of sync with each other | 16:56 |
m_3 | I've never gotten a variation of that to work... always ended up having to destroy/reboostrap | 16:56 |
m_3 | marcoceppi: are you using origin: ppa? | 16:57 |
m_3 | or distro? | 16:57 |
SpamapS | marcoceppi: right, the only way to make it truly repeatable is to stick juju in a bzr branch | 16:58 |
m_3 | well, even that's been changing | 16:58 |
SpamapS | marcoceppi: we *must* change that | 16:58 |
m_3 | right... gotta freeze juju | 16:58 |
SpamapS | I made a proposal for that | 16:58 |
SpamapS | but its seen basically no response | 16:58 |
marcoceppi | m_3: ppa ATM | 16:58 |
SpamapS | other tham +1's from hazmat here on IRC :) | 16:58 |
marcoceppi | with distro, will I be able to do placement: local and in subsequent upgrades of the distro what's to say it won't break the current deployment like the PPA did? | 16:59 |
SpamapS | marcoceppi: do you still have the affected environment? | 16:59 |
marcoceppi | SpamapS: yes | 16:59 |
marcoceppi | it's the omgubuntu env | 16:59 |
SpamapS | marcoceppi: can you do 'juju -v status' ? | 16:59 |
SpamapS | oh | 16:59 |
SpamapS | awesome! | 16:59 |
SpamapS | <sigh> | 17:00 |
imbrandon | marcoceppi: deploy on new ones and swap the ip is the only other option, but use the elastic ip | 17:00 |
imbrandon | for no downtime | 17:00 |
marcoceppi | SpamapS: it's okay, I can create a new env on the ohso AWS, bootstrap, deploy, add-relation, then move the elastic ip | 17:00 |
marcoceppi | imbrandon: yeah, that's the plan ATM | 17:00 |
SpamapS | I think its time for a stable PPA | 17:00 |
SpamapS | Been thinking about it for a long time | 17:00 |
marcoceppi | SpamapS: <3 | 17:00 |
imbrandon | :) | 17:00 |
SpamapS | with a ridiculous number of tests before upload | 17:00 |
marcoceppi | SpamapS: http://paste.ubuntu.com/923597/ | 17:01 |
SpamapS | and one of them will need to be "works with the last version" | 17:01 |
m_3 | marcoceppi: might be worth really balancing (dns) between the two different environments... instead of just a manual failover | 17:01 |
SpamapS | yeah | 17:01 |
m_3 | depends on the data | 17:01 |
imbrandon | SpamapS: can we add works on osx too ? /me ducks | 17:01 |
marcoceppi | m_3 ? | 17:01 |
SpamapS | looks like zookeeper was broken without bumping the version number | 17:02 |
SpamapS | imbrandon: don't duck, stand tall.. and donate a mini? ;) | 17:02 |
imbrandon | i actually probably could | 17:02 |
imbrandon | i got an extra | 17:02 |
m_3 | whoohoo! | 17:02 |
imbrandon | its an older one, but it will run 10.7 easy nuff ( intel core 2 duo ) | 17:03 |
imbrandon | marcoceppi: m_3 i think that would require dood to be using route 53 dns | 17:03 |
SpamapS | imbrandon: ideally we'd run a VM of Lion and Mountain Lion | 17:04 |
m_3 | marcoceppi: two different juju envs... A and B... name balances across elastic ips in both envs. Then you can make changes to one at a time and pull it out of rotation upon failure | 17:04 |
m_3 | it's more $$ tho | 17:04 |
imbrandon | SpamapS: oh thats easy, got any actual vmware products running, osx will legitly load in those | 17:04 |
imbrandon | from 10.7 on | 17:04 |
marcoceppi | m_3: then it's probably a no :) We're looking in to placement: local for mysql+bootstrap now to shave off another small | 17:05 |
m_3 | right | 17:05 |
marcoceppi | plus, not sure how the db sync would work | 17:05 |
SpamapS | marcoceppi: replication :) | 17:05 |
imbrandon | yea or rds | 17:06 |
SpamapS | rds would be overkill for their use wouldn't it? | 17:06 |
marcoceppi | now sure how to configure that in the existing charm infrastructure since bootstraps can't talk to each other (yet) | 17:06 |
imbrandon | yup | 17:06 |
imbrandon | awy overkill | 17:06 |
imbrandon | way* | 17:06 |
jcastro | SpamapS: dumb question. I am on precise, and using all "ootb" parts, like distro not ppa, etc. Since I can't deploy "juju deploy subway" since the precise store isn't open, should "juju deploy cs:oneiric/subway" work? | 17:07 |
m_3 | marcoceppi: no, the envs wwould be totally isolated | 17:07 |
* imbrandon really needs to fingure out how to bootstrap into hp cloud then i can test all kinda finky configs | 17:07 | |
m_3 | marcoceppi: oh wait, you're right... replication | 17:07 |
m_3 | we'd have to make it a config parameter... we want the environments as isolated as possible to provide availability | 17:08 |
m_3 | negronjl's done some stuff with this... it was mongo, not mysql replication, but something similar could be done for mysql | 17:09 |
imbrandon | SpamapS: but yea the mini dshould be able to handle that no problem, it will just be choked on ram, i thnk 4gb is the max for that generation | 17:09 |
imbrandon | SpamapS: but even today vmware has a special clause in their products with apple and can run osx, its how the big hosting companies get away with osx vps's | 17:10 |
imbrandon | i know fusion, esx , vsphere and workstation all work, not sure a bout plain old player | 17:11 |
negronjl | marcoceppi, m_3: The db sync would have to be a completely different approach ... they're diff. monsters | 17:11 |
marcoceppi | This is all hypothetically not going to happen for the omgubuntu install, since it's more $$, btw | 17:12 |
m_3 | negronjl: 1-way, right... but the basic idea of adding the replica's ip addr as a config param to the master... | 17:12 |
m_3 | marcoceppi: understood... sorry, just brainfarting out-loud | 17:13 |
negronjl | m_3: yeah ... that principle could work | 17:13 |
m_3 | well sorta-outlound :) | 17:13 |
imbrandon | heh | 17:13 |
marcoceppi | Feel free to keep farting, I just didn't want anyone to get too serious about ZOMG LETS DO THIS | 17:13 |
imbrandon | wow Kubuntu was "picked-up" i think that might be bigger news than the post leads on | 17:19 |
SpamapS | jcastro: you could just set default-series: oneiric too | 17:19 |
jcastro | SpamapS: yeah I was just wondering if I should just stick to all oneiric | 17:20 |
jcastro | but I want this precise hotness for LXC, etc. | 17:20 |
imbrandon | will the charms roll over like packages do ? | 17:20 |
jcastro | or does that not matter, my host is precise right? | 17:20 |
jcastro | imbrandon: yeah clint just needs to finish off some testing and stuff on it | 17:21 |
m_3 | we need to do the rollover sometime soon | 17:21 |
imbrandon | cool | 17:21 |
SpamapS | jcastro: we'll fix this by opening the precise charm store soon | 17:21 |
jcastro | SpamapS: yeah it's just hard to demo the contest winners but having to put a "by the way if you're running 12.04 this won't work." | 17:21 |
jcastro | so I announced the winners and then later on I will do the demo commands and stuff for each one | 17:22 |
SpamapS | jcastro: it will work, if they put 'default-series: oneiric' | 17:22 |
SpamapS | jcastro: or 'cs:oneiric/foo' | 17:22 |
SpamapS | jcastro: or you wait a week and we push hard to get it open | 17:22 |
m_3 | jcastro: charms are always developed against the current release, not dev | 17:22 |
imbrandon | hey while i got everyone in the same room, m_3 anything special i need to do besides manually mirror for now ? jcastro will you promgstrangualte drupal and where do i start down the path to get my 2 acks myself | 17:23 |
m_3 | working against precise should be gravy at this point for charms.. not required | 17:23 |
jcastro | m_3: yeah I just wanted my cake | 17:23 |
SpamapS | m_3: perhaps we should automatically push charms that have not diverged into the dev release when they pass tests on the dev release | 17:23 |
jcastro | SpamapS: hey so not doing a default-series:oneiric worked anyway | 17:23 |
SpamapS | that would be simple logic | 17:23 |
marcoceppi | imbrandon: you still need a final review on the drupal charm before promuglation | 17:23 |
jcastro | SpamapS: but we probably don't want to tell people to do that | 17:23 |
imbrandon | doh, marcoceppi sorry thought that was done | 17:24 |
imbrandon | sorry | 17:24 |
m_3 | SpamapS: but then the author's automatically maintaining two releases of the charm? | 17:24 |
SpamapS | m_3: not really. they only have to maintain when the tests start failing on the dev release. | 17:24 |
imbrandon | m_3: they do with packages, i dont see that as a problem | 17:24 |
marcoceppi | imbrandon: It looks like you fixed the mysql error thing, so I'll spin it up one more time (didn't get the bug reply for somereason) if it runs then it should be ready to go | 17:24 |
m_3 | ok | 17:24 |
jcastro1 | hello from the subway charm! | 17:24 |
imbrandon | marcoceppi: rockin | 17:24 |
SpamapS | jcastro: sweeeeet | 17:25 |
jcastro | man subway is so sexy, it's my new demo charm | 17:25 |
imbrandon | jcastro: how do you make subway persistant damn it, i had to go back to irssi | 17:25 |
m_3 | imbrandon: drupal'll be mirrored automatically in github.com/charms once it's in the store... then pull requests can go off of that mirrored branch | 17:25 |
m_3 | jcastro1 awesome! | 17:25 |
jcastro1 | imbrandon: run it on another machine you don't turn off | 17:25 |
imbrandon | sure then i close browser and it d/c | 17:25 |
SpamapS | no | 17:25 |
SpamapS | node.js keeps running | 17:25 |
SpamapS | its still online | 17:25 |
jcastro | oh, it's designed for multi user, you need to create an account on it | 17:25 |
jcastro | and just relogin next time | 17:26 |
imbrandon | it was on me, i couldent reconnect | 17:26 |
imbrandon | ohhh | 17:26 |
imbrandon | i never made acct | 17:26 |
SpamapS | yeah | 17:26 |
imbrandon | ok thats my missing part | 17:26 |
jcastro | imbrandon: that's nice because you can have multiple people on one deployment | 17:26 |
imbrandon | m_3: rockin, ok for now i've been manually pushing to both and have each in the others .ignore file | 17:26 |
jcastro | "here's my work IRC." done. | 17:26 |
imbrandon | jcastro: yea i'll likely redploy it on the box i run irssi from , i love it, just condent figure that bit out | 17:27 |
imbrandon | i was thinking it was a one user deal | 17:27 |
jcastro | alice irc in the charm store is the one user one | 17:27 |
imbrandon | yea | 17:28 |
m_3 | SpamapS: yeah, I guess that makes sense once we hit some milestone like beta2, then all charms get promoted and tested against the dev release | 17:28 |
m_3 | so we're behind :) | 17:28 |
imbrandon | SpamapS: btw i've had "walk" on repeat now for 45 minutes, lol , time to change the tunes | 17:28 |
imbrandon | your fault | 17:28 |
imbrandon | marcoceppi: yea it should have just been a matter of changing it to the mysql not mysqli driver on install | 17:30 |
imbrandon | i got some more goodness in store but testing it on brandonholtsclaw.com now, as it breaks some of the default drupal functionality | 17:30 |
imbrandon | as is | 17:30 |
imbrandon | but takes it from 20 r/s to like 160 r/s ootb | 17:31 |
imbrandon | :) | 17:31 |
SpamapS | mysqli is *awful* | 17:31 |
SpamapS | do not use it | 17:31 |
SpamapS | mysqlnd ftw | 17:31 |
imbrandon | yea i dont and will when 5.4 hits | 17:31 |
imbrandon | 5.3 nd is awefull too | 17:32 |
SpamapS | Tho I don't know if we have it enabled yet on 5.3 in precise :-P | 17:32 |
SpamapS | imbrandon: it still beats mysqli | 17:32 |
imbrandon | yea , just dont, its bad | 17:32 |
imbrandon | i've been using the 5.4 ppa | 17:32 |
imbrandon | apc is broken tho | 17:32 |
imbrandon | thats bad | 17:32 |
shazzner | hey everyone | 17:32 |
imbrandon | ello | 17:32 |
imbrandon | the debian 5.4 ppa that is | 17:33 |
imbrandon | something about it is breaking apc right now, not sure what though, i cant get past segfaults | 17:33 |
imbrandon | thus the charm still using 5.3 | 17:33 |
shazzner | a few days ago I talked about promulgating my kusabax charm, but I'm not sure how to do it | 17:33 |
imbrandon | but the 30% increase in perf would be nice | 17:34 |
m_3 | aargh! if I upgrade my cli, I can't talk to deployed envs... if I don't, I can't deploy new ones | 17:34 |
imbrandon | i officially changed that command to promgustrangulate | 17:34 |
shazzner | imbrandon: haha | 17:35 |
* m_3 groan | 17:35 | |
imbrandon | catch 22 | 17:35 |
marcoceppi | m_3: it's funtastic! | 17:35 |
m_3 | I'm gonna snapshot VMs to run the cli for each long-running env | 17:36 |
m_3 | and start freezing branches | 17:37 |
imbrandon | i'm thinkign we;re gonna start to see that lots in here | 17:37 |
imbrandon | the growing pains of rolling releases | 17:37 |
m_3 | actually, that sounds like a great role for wrap-juju in juju-jitsu... like a virtualenv | 17:41 |
imbrandon | oh :( | 17:42 |
m_3 | read environments.yaml to get the branch for each environment and update path accordingly... hmmm | 17:43 |
imbrandon | SpamapS: mmmm juju deploy osx/10.8 :) | 17:46 |
jcastro | SpamapS: m_3: you guys feel like hanging? | 17:55 |
jcastro | I have some questions about the documentation for review in that bug jdstrand filed | 17:55 |
SpamapS | jcastro: in 10 | 17:55 |
m_3 | jcastro: sure | 17:56 |
m_3 | lemme grab some more coffee, then I'll be on | 17:58 |
marcoceppi | Is there a way to get peer-relation information in the install hook? | 18:08 |
SpamapS | marcoceppi: with the new relation-ids thing, yes! | 18:08 |
marcoceppi | fantastic! | 18:08 |
hazmat | SpamapS, marcoceppi there's definite awareness of the foobar that is consistent versions... the bzr branch atm is the realistic option.. going forward, having bootstrap stick a code copy on provider storage, and deploying from provider storage is probably the right solution, package solutions get messy both for consistent versions and for cross release | 18:09 |
marcoceppi | SpamapS: can I bother you in the #juju-wordpress room for a few mins. Having a hard time figuring out the best way to proceed with something for omgwp | 18:09 |
marcoceppi | hazmat: that would be beneficial, if the bootstrap deployed the same juju version across all nodes. that way I can just hold the juju package on my desktop until a more "structured" release update schedule is created | 18:10 |
hazmat | imbrandon, that reminds there was a fun php article on hacker news today, in informed hater style, http://me.veekun.com/blog/2012/04/09/php-a-fractal-of-bad-design/ | 18:10 |
marcoceppi | Can you have more than one charm deployed to the bootstrap? | 18:16 |
marcoceppi | with placement: local | 18:18 |
SpamapS | marcoceppi: yes, but don't do it in parallel | 18:18 |
marcoceppi | how so? | 18:18 |
SpamapS | marcoceppi: if you deploy && deploy, apt-get will run in parallel and fail | 18:18 |
marcoceppi | So deploy, wait for started, then deploy again? | 18:19 |
marcoceppi | I'm trying to figure out if it'd be better to just deploy the memcache charm on the bootstrap with MySQL or if I should bake memcache in to the charm and use peer relations to manage the sessions between units | 18:20 |
marcoceppi | since there's a memcache charm, figured that would be easiest | 18:20 |
=== garyposter is now known as gary_poster | ||
yolanda | hi | 18:31 |
imbrandon | hazmat: heh /me looks | 18:31 |
yolanda | i found a problem running postgresql charm in juju local container | 18:31 |
yolanda | seems that isn't starting, and when i debug i receive that error: http://paste.ubuntu.com/923751/ | 18:31 |
hazmat | yolanda, the pg log should have the actual error | 18:32 |
hazmat | the locale stuff is looks like a warning.. | 18:33 |
hazmat | and the hba.conf errors look more interesting | 18:33 |
* hazmat tries it out | 18:33 | |
yolanda | http://paste.ubuntu.com/923758/ | 18:35 |
hazmat | yolanda, are you deploying that from the charm store or from a charm directory on disk? | 18:38 |
shazzner | mediagoblin looks pretty cool | 18:39 |
shazzner | let me see if I can't whip up a charm for it | 18:39 |
SpamapS | shazzner: I already made one | 18:40 |
SpamapS | shazzner: it uses pip installs heavily tho | 18:40 |
SpamapS | shazzner: got wrapped up in trying to package half of pypi to get it into a PPA | 18:40 |
shazzner | SpamapS: oh nice! :) | 18:41 |
yolanda | hazmat, i grabbed the charms with the charm get tool | 18:42 |
yolanda | it's on a directory on disk, local repository | 18:43 |
imbrandon | hazmat: haha yea i've seen that argument 1000 times, its fundmentally wrong, that parts i see that are good he see's as flaws, i actually think that its a strength of php that you can write bad code, infact one of his complaints sums it up perfect on why php is GOOD , PHP is built to keep chugging along at all costs. When faced with either doing something nonsensical or aborting with an error, it will do something nonsensical. Anything is better t | 18:43 |
imbrandon | /endquote :) | 18:43 |
m_3 | yolanda: is that against oneiric or precise? | 18:49 |
m_3 | (the postgresql problems) | 18:50 |
yolanda | m_3, precise | 18:50 |
m_3 | yolanda: it hasn't been tested against precise yet... that's later this week hopefully :) | 18:51 |
m_3 | we're about to test all charms against precise to prepare for bumping the release version | 18:51 |
yolanda | :( | 18:51 |
yolanda | i was trying to build a charm that uses postgres | 18:52 |
yolanda | so better wait until the test is done? | 18:52 |
jcastro | m_3:dang, that ilo card is $85 | 18:53 |
m_3 | jcastro: yup... I was kinda pissed when I realized it wasn't built-in | 18:53 |
m_3 | still rockin setup though... and pretty cheap even with the ilo card | 18:53 |
avoine | yolanda: can you show us the line that is broken? | 18:53 |
m_3 | yolanda: maybe I can test that one manually early... it shouldn't be too hard of a fix | 18:54 |
m_3 | the postgresql charm is kinda naive in some of what it does... needs some love from a pg expert :) | 18:54 |
yolanda | i pasted the postgres log here: http://paste.ubuntu.com/923758/ | 18:55 |
m_3 | sounds like the host-based-access template is barfing | 18:55 |
yolanda | let me check the config line | 18:55 |
m_3 | heck, could probably even be disabled depending on the app | 18:55 |
avoine | yeah it needs IP adresse I think | 18:56 |
avoine | or a valid domain name | 18:56 |
yolanda | that one: host all openerp 192.168.122.122 md5 | 18:56 |
m_3 | hmmm... that looks correct | 18:57 |
m_3 | lemme check the charm | 18:57 |
m_3 | oh wait... that's strange | 18:58 |
m_3 | the 192.168.122.122 should say 192.168.122.122/0 | 18:58 |
m_3 | yolanda: you can tweak this behavior on line 13 of hooks/db-relation-changed | 18:59 |
yolanda | let me test | 19:00 |
m_3 | yolanda: perhaps make the change manually on the instance first to test it, then if that works change the charm | 19:00 |
yolanda | ok | 19:00 |
m_3 | might even need to be 192.168.122.122/32 or 192.168.122.0/24 | 19:01 |
yolanda | line 13 shows just an "exit 0" for me | 19:02 |
yolanda | let me see if i have to update the charm | 19:03 |
shazzner | does anyone know of this server software that allows people to write/edit books collaboratively? | 19:05 |
* shazzner looking for charms to tackle | 19:05 | |
shazzner | it was called book/word-something | 19:06 |
shazzner | I thought I had saved the link :/ | 19:06 |
m_3 | yolanda: I was looking at an older version on my laptop... try line 17 | 19:06 |
shazzner | ah-ha Booktype that's it | 19:07 |
m_3 | yolanda: try to manually edit pg_hba.conf and add the /32 to the ip addr... see if pg will restart | 19:07 |
m_3 | if so, that can be added to the charm. the mask was there in the natty version, but was _removed_ from the oneiric version of the postgresql charm | 19:08 |
yolanda | m_3, seems that postgresq starts when i modify the 2 final lines: | 19:08 |
yolanda | host openerp openerp 192.168.122.122/0 md5 | 19:08 |
yolanda | host all openerp 192.168.122.122/0 md5 | 19:08 |
m_3 | shazzner: gitolite and pandoc works really well for that | 19:09 |
shazzner | m_3: ah cool | 19:09 |
* shazzner checking out pandoc | 19:09 | |
m_3 | yolanda: so the /0 needs to be added _back_ to the precise version of postgres (I'm confused) | 19:09 |
shazzner | this is booktype: http://www.sourcefabric.org/en/booktype/ | 19:10 |
avoine | shazzner: https://github.com/esetera/Booki.git | 19:10 |
yolanda | m-3, yes, seems that it's the solution | 19:10 |
m_3 | shazzner: http://johnmacfarlane.net/pandoc/epub.html | 19:10 |
avoine | oh booki is the old version of booktype | 19:10 |
yolanda | m_3, so something like ${remote_host}/0 must be there? | 19:10 |
m_3 | yeah, let me copy you the line in the older version.. "host $(get_database_name) ${user} ${remote_host_ip}/0 md5" | 19:11 |
m_3 | although /32 would make more sense honestly | 19:11 |
m_3 | s/remote_host_ip/remote_host/ | 19:12 |
m_3 | shazzner: I use an old-school makefile with this http://paste.ubuntu.com/923826/ | 19:14 |
yolanda | m_3, what's the last line? don't understand | 19:16 |
shazzner | m_3: pandoc is nice just saw there is an emacs mode too | 19:17 |
m_3 | yolanda: substitute "remote_host_ip" with "remote_host" in the line above... (remote_host_ip is old) | 19:18 |
m_3 | shazzner: and you can embed latex if you're into that sort of thing | 19:18 |
m_3 | shazzner: it's a little more manual... and you collaborate using revision control just like source... but I like it in practice | 19:19 |
m_3 | also easy to retarget to docbook or other such nonsense that publishers might want | 19:19 |
shazzner | m_3: yeah it's rad, I'll save that for later | 19:19 |
avoine | yolanda: are you using the last version of the openerp charm? | 19:20 |
* shazzner prefers local scripts to large server software | 19:20 | |
shazzner | still though I think booktype would make a good charm | 19:20 |
m_3 | absolutely | 19:21 |
shazzner | I'm reading that it requires postgresql, sounds like there are some bugs in the charm though? | 19:21 |
yolanda | m_3, there is not remote_host_ip line in the hook... | 19:22 |
yolanda | avoine, it's just a new openerp charm that i'm writing for 6.1 | 19:22 |
yolanda | we are also creating a new package for it, it's a kind of collaboration with openerp | 19:23 |
yolanda | so the existing openerp charm isn't the same | 19:23 |
shazzner | good heavens, the booktype install instructions is a book unto itself | 19:24 |
m_3 | yolanda: yes, sorry... I was pasting a line from an older version of the charm. That version used a variable that's no longer there, so I was trying to say "replace that variable with the new variable" in the line I pasted | 19:24 |
yolanda | oh, sounded very strange to me :) | 19:24 |
* m_3 just tripping over my words :) | 19:24 | |
yolanda | so i just add a /0 or a /32 at the end? | 19:24 |
m_3 | I'd try /32 because that makes the most sense for a netmask... but /0 seems to work | 19:25 |
m_3 | the /0 is safest b/c you've already tried that and it worked | 19:25 |
avoine | yolanda: I'm the author of the current openerp charms so don't be afraid to tell me if you want me to change anything to make it fits your new packages | 19:26 |
yolanda | avoine, the thing is that the charm is totally different, because openerp 6.1 has server and web integrated | 19:26 |
yolanda | and also it's relying on a new package that we are building | 19:26 |
yolanda | i wanted to push it as soon as i had it working | 19:27 |
m_3 | and I wrote the first postgresql charm... and haven't been able to find someone who knows something about postgresql to own it :) | 19:27 |
m_3 | happy to make whatever changes are necessary | 19:27 |
yolanda | i'll do that change and try | 19:28 |
avoine | m_3: I had a suggestion in this bug but I'm not sure if it could break something 902672 | 19:28 |
avoine | hum, I thought the irc bot would catch it, the link: https://bugs.launchpad.net/charms/+source/postgresql/+bug/902672 | 19:29 |
_mup_ | Bug #902672: Postgresql pg_hba.conf needs IP addresses <postgresql (Juju Charms Collection):In Progress by mark-mims> < https://launchpad.net/bugs/902672 > | 19:29 |
m_3 | avoine: right... sorry that's been back-burner | 19:30 |
m_3 | avoine: I think the biggest problem with this is different providers behaving differently. I think relation-get private-address is now returning an actual ip addr on ec2 and lxc | 19:34 |
yolanda | well, leaving now, it's quite late here... | 19:34 |
avoine | m_3: yes | 19:34 |
yolanda | thanks for all | 19:34 |
avoine | and the patch won't work on IPv6 only network | 19:35 |
m_3 | avoine: I'll check to see if this was actually fixed in core or is still broken for other providers (openstack) | 19:36 |
bkerensa | jcastro: you should ask people to take photos with their juju shirts :) | 19:42 |
_mup_ | juju/must-specify-relation-name r517 committed by jim.baker@canonical.com | 20:16 |
_mup_ | Merged trunk | 20:16 |
robbiew | jcastro: http://gigaom.com/cloud/twitter-open-sources-its-mysql-secret-sauce/ | 20:33 |
jcastro | I saw, he told me that at posscon but didn't say which | 20:33 |
robbiew | https://github.com/twitter/mysql | 20:34 |
robbiew | SpamapS: ^ | 20:35 |
jcastro | I hope you are putting that into a PPA right now Clint, chop chop! | 20:35 |
jcastro | :) | 20:35 |
* robbiew spends time reading the news while his maas deploys | 20:35 | |
yagey | hello. embarrassing first question/problem: juju bootstrap Could not find AWS_ACCESS_KEY_ID 2012-04-10 20:28:41,619 ERROR Could not find AWS_ACCESS_KEY_ID | 20:36 |
SpamapS | twitter's mysql? | 20:36 |
SpamapS | nice | 20:36 |
yagey | I have set the values in my ~/.profile | 20:36 |
robbiew | I *think* I'll be ready for my first juju deploy into my maas...woot | 20:36 |
jcastro | SpamapS: they just OSSed all their internal tweaks to mysql | 20:36 |
jcastro | yagey: you need to put that in your .juju/environments.yaml file | 20:36 |
SpamapS | I do plan to gather some support tomorrow for a group of smart people getting together to have one packaging branch for all of the forks. | 20:36 |
jcastro | yagey: https://juju.ubuntu.com/docs/getting-started.html | 20:37 |
yagey | i've tried setting access-key and secret-key in environments.yaml too, same error | 20:37 |
jcastro | under "configuring your environment" | 20:37 |
jcastro | SpamapS: he told me they wanted them out there, they don't plan to maintain their own fork etc, they're just going to roll with whatever percona, maria, etc. and co do with them | 20:37 |
jcastro | yagey: hmm weird, that has always just worked for me | 20:38 |
yagey | i removed from .profile and added to env.y, different error sorry - ubuntu@ip-10-194-250-210:~$ juju bootstrap 2012-04-10 20:39:19,137 INFO Bootstrapping environment 'sample' (type: ec2)... SSH authorized/public key not found. 2012-04-10 20:39:19,291 ERROR SSH authorized/public key not found. | 20:39 |
SpamapS | "Update: Someone alerted me that Facebook does have a site for sharing its MySQL patches, although it’s hosted at Launchpad." | 20:40 |
SpamapS | although.. as in.. thats a reason not to acknowledge it? | 20:40 |
SpamapS | Does the author realize that *mysql* is hosted on launchpad? | 20:41 |
yagey | ah - I ran: ssh-keygen -t rsa -b 2048 that fixed it | 20:42 |
yagey | I should be able to run juju on an ec2 instance? | 20:52 |
hazmat | yagey, yes | 20:58 |
yagey | i see, it started an 'small' ec2 instance, which is not in the free tier :( | 21:04 |
imbrandon | SpamapS: yea but the world dont see that ( LP vs Github ) kinda like someone bitching about PHP again, php still powers the internet just as Github hosts more opensource code than any other place on the planet, even though Django or Plone or Launchpad or what ever may have the technical upper hand | 21:04 |
imbrandon | sucks but its the way of the west | 21:04 |
imbrandon | ( and yes github now has more code than sourceforge ) | 21:05 |
imbrandon | Launchpad was tied to Ubuntu only for too long, it will take a ton of momentum to upheave that in the collective mindset of the internets | 21:07 |
yagey | is there a way to have it use micro instances instead? | 21:07 |
imbrandon | yagey: you should be able to set the deafult int he env.y | 21:07 |
yagey | ok, rtfm https://juju.ubuntu.com/docs/constraints.html | 21:07 |
hazmat | yagey, yeah.. you can set default-instance-type: t1.micro in your aws environment section of environments.yaml.. or if your using a recent version (ppa) it can be specified via cli --constraints="instance-type=t1.micro" | 21:08 |
hazmat | yagey, be aware that the micro instances are severely penalized on cpu usage.. | 21:08 |
yagey | in terms of performance? but they are free :) | 21:08 |
imbrandon | 1 is free | 21:09 |
hazmat | we tried using them as the default for ec2 for a little while, but they were basically broken wrt to predictable performance. | 21:09 |
marcoceppi | You get 750 hours of free micros a month, which ends up being about 1 micro a month | 21:10 |
yagey | only 1? it doesn't mention that on the page | 21:10 |
yagey | ah that makes sense, but I won't run for a whole month yet | 21:10 |
marcoceppi | yagey: then you can probably get away with deploying a few things on micros and removing them when done, be aware that there are a lot of issues with bootstrapping and deploying on smalls because of their CPU constraints | 21:11 |
marcoceppi | It's a hit-or-miss situation | 21:11 |
yagey | ok, thanks | 21:11 |
imbrandon | you can just try things out too on local lxc containers | 21:12 |
yagey | i saw that page, didn't seem simple enough to start with however | 21:12 |
imbrandon | if your just kicking the tires | 21:12 |
yagey | I don't see a page showing the environment.yaml possible options in the dev documentation section | 21:14 |
marcoceppi | yagey: http://askubuntu.com/questions/65359/how-do-i-configure-juju-for-local-usage | 21:14 |
imbrandon | ok so hazmat , marcoceppi mentioned something ealier that i'd like to dig into deeper ( esp since hp cloud comes out of beta here very soon , as in we have a date now ) soooo instead of using s3 container for the bootstrap store the stuff on the bootstrap node and how do i go down that rabbit hole not being familiar with the juju internals at all at this point but more than willing to get my hands dirty , even with putting code where my mouth is | 21:16 |
* marcoceppi has secretly been talking to HP. Tuns out their storage doesn't have S3 compatibility turned on but has EC2 compatibility enabled for compute. | 21:16 | |
imbrandon | yea i got the same thing from pete just now | 21:16 |
hazmat | yeah.. i saw that re storage is native, and compute has ec2 compat.. | 21:18 |
yagey | LXC doesn't appear to work on a virtualbox ubuntu VM, as the network is already virtual | 21:18 |
imbrandon | the nic on the vm should be seen as normal by the os | 21:19 |
imbrandon | if its bridged | 21:19 |
hazmat | imbrandon, so the client uses the storage to identify where the bootstrap node is.. | 21:20 |
hazmat | so running the storage on the bootstrap node becomes a bit more challenging given that | 21:20 |
hazmat | another option would be to play mix and match storage and compute with an ostack native storage | 21:20 |
imbrandon | right, so a) i write it to work with swift, or b) make the bootstrap node info avail some other way , least that how i understand it, not having looked at code | 21:21 |
hazmat | imbrandon, yup that's the nutshell.. or c) ask hp to turn on the swift s3 middleware | 21:21 |
imbrandon | hehe been there, i dont think that will happen iirc | 21:21 |
imbrandon | leaste not comming from me | 21:22 |
marcoceppi | I had them open a ticket on it for tracking | 21:22 |
marcoceppi | not sure if it'll happen | 21:22 |
hazmat | yeah.. probably not.. but you never know.. i'm why they would do one but not the other.. i take it the s3 swift middleware perhaps isn't as battle tested at scale perhaps.. | 21:23 |
imbrandon | so just how intracate is the s3:// code, i mean can either of you point me to the proper digging point i can see just how much of a pain it would be two accomidate swift | 21:23 |
imbrandon | hazmat: well i'm told that even the compute will go away is their hopes, i dont see that, amazon has won this round | 21:24 |
imbrandon | but you know | 21:24 |
hazmat | imbrandon, its pretty simple.. 3 methods (get/put/get_url) -> juju/providers/ec2/files.py | 21:25 |
imbrandon | sweet, ok that dont sounds terrible | 21:26 |
marcoceppi | hazmat: so would you recommend making juju/providers/openstack/ for this? | 21:26 |
hazmat | the get_url needs to create a url that can be used sans credentials.. with s3 its done via their url signing mechanism.. swift also has this capability.. but its in a middleware piece | 21:26 |
marcoceppi | I feel it would be better to create a native openstack provider rather than just modifying the current | 21:26 |
hazmat | i don't know that its enabled on hp's storage | 21:26 |
hazmat | ie.. this one.. http://swift.openstack.org/misc.html#module-swift.common.middleware.tempurl | 21:27 |
imbrandon | i was thinking about just translating the ruby gem hpcloud cli parts to python | 21:27 |
hazmat | imbrandon, its just ostack native api.... ostack distributes native python clients | 21:28 |
hazmat | afaik re hp's impl of it | 21:28 |
imbrandon | oh then even better | 21:28 |
imbrandon | yea' | 21:28 |
imbrandon | if its just pure | 21:28 |
SpamapS | hazmat: couldn't we also just use groups to find the bootstrap machine? | 21:28 |
hazmat | imbrandon, the juju code base is twisted though.. its like python.. but sorta of not :-( | 21:28 |
hazmat | SpamapS, that would work as well | 21:29 |
SpamapS | I mean, do we really need file storage for *that* ? | 21:29 |
hazmat | SpamapS, good point | 21:29 |
hazmat | SpamapS, we just need some discovery mechanism | 21:29 |
SpamapS | I'd love to see an 'hpcloud' provider which just hack-xtends the ec2 provider to implement the file storage. | 21:29 |
imbrandon | heh | 21:30 |
hazmat | SpamapS, at that point why not just a native ostack impl? | 21:30 |
SpamapS | which would then be useful for any cloud that doesn't have S3 | 21:30 |
SpamapS | hazmat: because thats a lot more code! | 21:30 |
hazmat | SpamapS, ie. providers/openstack per marcoceppi's suggestion | 21:30 |
SpamapS | Well yes call it whatever you want | 21:30 |
SpamapS | call it "halfec2" or "franken-ec2" | 21:31 |
imbrandon | well os would work on rackspace too even though they have s3 middleware rtight | 21:31 |
hazmat | SpamapS, not nesc. it could be incomplete if we allow for storage spec independently.. but yeah.. that kinda of gets ugly | 21:31 |
marcoceppi | so, just do providers/hpcloud until someone comes out with a full openstack native provider? | 21:31 |
SpamapS | I wouldn't call it openstack unless you do the full OSAPI implementation | 21:31 |
hazmat | didn't someone start one of those.. | 21:31 |
SpamapS | hazmat: the openstack guys were interested | 21:31 |
imbrandon | i thought that swift was | 21:31 |
SpamapS | one of the nova core devs | 21:31 |
imbrandon | osapi | 21:31 |
marcoceppi | I wish I was better at python or knew go | 21:32 |
imbrandon | why go ? | 21:32 |
marcoceppi | that's what juju will be written in soon, iirc | 21:32 |
imbrandon | ahh, somone got google fever, why not just full nodejs then | 21:33 |
SpamapS | imbrandon: uh.... nodejs and go have like, nothing in common. | 21:33 |
imbrandon | go compiles to js | 21:33 |
SpamapS | dart | 21:33 |
SpamapS | thats dart | 21:33 |
SpamapS | go compiles into native statically linked binaries | 21:34 |
imbrandon | ... >.> /me might be mixed up | 21:34 |
SpamapS | with a big runtime | 21:34 |
SpamapS | well | 21:34 |
SpamapS | a tiny runtime actually | 21:34 |
SpamapS | anyway, yeah, go, not dart | 21:34 |
marcoceppi | php \o/ | 21:34 |
SpamapS | imbrandon: lp:juju/go | 21:34 |
imbrandon | :) | 21:34 |
* imbrandon high fives marcoceppi | 21:34 | |
imbrandon | SpamapS: yea i'll have to look, no idea how i equated the two | 21:35 |
SpamapS | imbrandon: well apparently google likes to invent languages :) | 21:35 |
imbrandon | maybe dart in written in go ? /me is even more confused now | 21:35 |
SpamapS | and go was such an odd choice.. like, how about call it 'be' or 'have' | 21:35 |
imbrandon | lol | 21:36 |
imbrandon | marcoceppi: and languages are just syntax, once you learn one, picking up new ones is easy :) its what keeps us relevant | 21:37 |
SpamapS | yeah, I like go actually | 21:37 |
imbrandon | is it whitespace depdant ? | 21:37 |
SpamapS | makes me feel like I'm reading C, but I don't have to get all sideways on pointers | 21:37 |
SpamapS | no | 21:37 |
SpamapS | imbrandon: its very C-like | 21:38 |
SpamapS | imbrandon: but it has an automatic formatter thingy that keeps all go indented the same | 21:38 |
imbrandon | sweet i like c like | 21:38 |
marcoceppi | I went through the go demo thing, got to chapter 20 something | 21:38 |
imbrandon | yea as long as its not whitspace sensative i can learn to love it, its my ONLY problem with python | 21:39 |
SpamapS | whats so bad about whitespace sensitive? It makes all python readable, even written by the most junior coders. | 21:39 |
imbrandon | auto formatting is good, i am anal about that in my own code | 21:40 |
imbrandon | even php | 21:40 |
imbrandon | but i dont want code to break because of it | 21:40 |
imbrandon | 2 spaces per indent, no tabs | 21:40 |
imbrandon | ever | 21:40 |
imbrandon | brackets are a must, even for one liners | 21:40 |
imbrandon | :) | 21:40 |
imbrandon | people hate when i nit pick their php | 21:41 |
imbrandon | but it makes it so i can maintain shit i wrote 5 years ago easy | 21:41 |
SpamapS | imbrandon: right, which is why people like python. There is no nit picking. | 21:42 |
SpamapS | it works, or it doesn't | 21:42 |
imbrandon | heh | 21:42 |
SpamapS | anyway, language debates are outlawed for the next 30 minutes so I can get something worthwhile done | 21:43 |
imbrandon | roger :) | 21:43 |
SpamapS | (like save all these daisies for these pirates) | 21:43 |
hazmat | http://www.pirateslovedaisies.com/ :-) | 21:45 |
imbrandon | zomg | 21:46 |
imbrandon | ok sooo back to the main question, do i start a hp cloud provider and just copy most of the code from ec2 | 21:48 |
imbrandon | as bloilerplate | 21:48 |
imbrandon | and fix the storage or what was that aobut groups | 21:48 |
SpamapS | imbrandon: I would import the ec2 provider's machineprovider object, and just write my own filestore | 21:48 |
SpamapS | imbrandon: please don't copy/paste! ;) | 21:49 |
imbrandon | or import | 21:49 |
imbrandon | :) | 21:49 |
SpamapS | imbrandon: I know you're used to PHP and namespaces are all new and sexy.. but they are built in to python :) | 21:49 |
imbrandon | hahaahh | 21:49 |
imbrandon | i know i know :) i dont even have to name them, just have the file names ;) | 21:49 |
imbrandon | just cuz i dont like python dont mean i cant hobble some togather :) | 21:50 |
imbrandon | man i could get used to ubuntu app design again, css3 for skins now go instead of python | 21:51 |
imbrandon | its like your after my own heart | 21:52 |
imbrandon | pypm install launchpadlib | 21:56 |
imbrandon | gah | 21:56 |
robbiew | SpamapS: so...if I have juju installed from the ppa and do a bootstrap, my node should have the same version of juju...right? | 21:56 |
robbiew | testing maas...and my node is getting juju from the archive :/ | 21:57 |
robbiew | even though I have the ppa setup on my local machine | 21:57 |
imbrandon | what about in the env.y | 21:57 |
robbiew | that's empty | 21:57 |
robbiew | wait..no | 21:58 |
robbiew | I actually added juju-origin: ppa | 21:58 |
imbrandon | ahh then smells like a bug | 21:58 |
robbiew | b/c w/o that line, it still did the same :/ | 21:58 |
robbiew | http://www.sadtrombone.com | 21:58 |
hazmat | ugh.. pypm.. active state ugh.. | 21:59 |
imbrandon | hazmat: easy way to get launchpadlib on osx | 21:59 |
hazmat | robbiew, so checked your apt-cache output, the code parses it correctly.. which odd | 21:59 |
hazmat | imbrandon, pip/easy_install are builtin on osx as well | 21:59 |
SpamapS | robbiew: the PPA auto-detect thing may not be working right | 22:00 |
hazmat | robbiew, so even with the explicit juju-origin its installing from the archives? | 22:00 |
robbiew | yes | 22:00 |
robbiew | 0.5+bzr504-0ubuntu1 on the bootstrap node | 22:01 |
robbiew | 0.5+bzr518-1juju5~precise1 on my machine | 22:02 |
hazmat | robbiew, if you login into the instance can you pastebin /var/log/cloud-init-output.log and /var/lib/cloud/instance/user-data.txt | 22:02 |
robbiew | sure thing | 22:03 |
hazmat | i put some code in the latest trunk on r517 to log origin on bootstrap, but with the apt-cache policy juju you sent me earlier, the auto detect code did the right thing.. but even ignoring that.. getting juju-origin switched out is bogus.. | 22:04 |
hazmat | if its explicitly specified | 22:04 |
hazmat | robbiew, hmm.. i wonder.. does maas reinstall when machines get recycled on destroy-environment/bootstrap.. or just stop/start the machine. | 22:05 |
robbiew | hazmat: good question...i don't believe it reinstalls | 22:07 |
robbiew | hazmat: http://pastebin.ubuntu.com/924048 | 22:09 |
hazmat | robbiew, hmm.. 2012-04-10 17:46:34,398 - cc_apt_update_upgrade.py[WARNING]: Source Error: ppa:juju/pkgs:add-apt-repository failed | 22:09 |
robbiew | hmm | 22:10 |
robbiew | maybe network blip | 22:10 |
hazmat | so the ppa fails, and when it goes to install it latter it gets distro.. that makes more sense | 22:11 |
SpamapS | doesn't reinstall? | 22:12 |
SpamapS | thats a violation of the basic expectations that juju offers charm authors IMO. | 22:13 |
SpamapS | Otherwise charms need to get even more careful about idempotency | 22:13 |
imbrandon | Why is Petyr Baelish called "LittleFinger" ? | 22:19 |
hazmat | idempotency doesn't nesc. help here, without the reinstall the disk just becomes an accumulated dumping ground, juju expects thats releasing a machine will make it go poof, destroy-env doesn't attempt to cleanup disks, so you'd have potentially old agents still running on the newly booted machine. | 22:19 |
robbiew | ok...I'm going to blow away my maas and start fresh...luckily I've done that a bazillion times now, so it's painless :P | 22:21 |
hazmat | imbrandon OT but.. http://gameofthrones.wikia.com/wiki/Petyr_Baelish | 22:22 |
imbrandon | hazmat: haha yea, i seen it on a adv banner, thought it was clever , winter is def comming | 22:23 |
SpamapS | imbrandon: because he's from the fingers and he's little | 22:34 |
* SpamapS has found himself getting more and more obsessed with GoT | 22:34 | |
imbrandon | same here. ive actualy read ahead in the books now, just waiting on the story to play out on tv | 22:35 |
SpamapS | yeah, I'm on book 5 | 22:35 |
imbrandon | i'm not quite that far, i think starting 4 its been a few months | 22:36 |
imbrandon | i'm 1 past where last season ended | 22:36 |
imbrandon | and that was 3 i think | 22:36 |
SpamapS | yeah, from the previews, it looks like season 2 may be a bit more than book 2 | 22:37 |
imbrandon | yea, i just watched the first eps of season 2 lasty night on dvr | 22:37 |
marcoceppi | Holy crap, that's quite a loud depreciated warning | 22:45 |
marcoceppi | I guess no more default-instance-type | 22:46 |
SpamapS | what | 22:46 |
SpamapS | oh its all wrapped up in constraints now? | 22:46 |
marcoceppi | http://paste.ubuntu.com/924092/ | 22:46 |
SpamapS | I'm guessing constraints is what broke existing bootstrapped envs from using the PPA | 22:46 |
marcoceppi | yeah, I was just surprised how verbose that warning was | 22:47 |
marcoceppi | NO! It's not in autocomplete | 22:47 |
marcoceppi | unless, you can't bootstrap with constraints? | 22:47 |
SpamapS | we need to update the bash completion to call juju the way bzr does | 22:47 |
marcoceppi | can I run juju set-constraints before bootstraping, or is that something that needs to run after you boostrap? | 22:49 |
SpamapS | marcoceppi: I think it has a --constraints | 22:49 |
marcoceppi | ah, the wiki page has the info | 22:50 |
marcoceppi | sweet | 22:50 |
marcoceppi | I searched everywhere but the obvious place | 22:50 |
marcoceppi | SpamapS: so, since I have the latest ppa version, how do I do placement: local? | 22:53 |
jcastro | aha! | 22:53 |
marcoceppi | is that still in the environment.yaml | 22:53 |
jcastro | updated documentation! | 22:53 |
jcastro | so used to having it be out of date that now you have no idea what to do! | 22:53 |
imbrandon | SpamapS: dude, GO is typed PHP 5.4 | 22:56 |
imbrandon | like for real | 22:56 |
SpamapS | http://cdn.memegenerator.net/instances/400x/18261941.jpg | 22:56 |
imbrandon | 404 | 22:57 |
SpamapS | not here. :-P | 22:59 |
lifeless | 500 | 22:59 |
imbrandon | ERROR | 22:59 |
imbrandon | The request could not be satisfied. | 22:59 |
lifeless | indeed | 23:00 |
imbrandon | btw moins lifeless :) | 23:00 |
imbrandon | mmm i could get used to this though, /me downloads the dev kit | 23:01 |
SpamapS | imbrandon: golang 1 is in precise | 23:03 |
imbrandon | i know i know, i need to install precise, i tried this morning actually, and failed, almost took out my osx partition | 23:04 |
imbrandon | was quite miffed , will try again in the monring | 23:04 |
SpamapS | I install using rEFIt and unetbootin .. never have problems that way | 23:05 |
robbiew | hazmat: sorted it....networking related...routing between my maas setup and the internet was broken | 23:05 |
yagey | when I deploy a mongodb instance, and another via add-unit - it should configure a replica set right? | 23:07 |
imbrandon | SpamapS: yea i couldent get the precise image to boot from refit | 23:08 |
imbrandon | just kept recycling | 23:08 |
imbrandon | but i grabbed the one "made for mac" i'm grab the normal x64 one and try again | 23:09 |
SpamapS | yagey: yes it should | 23:10 |
* SpamapS bumps negronjl to write a README for mongodb | 23:10 | |
SpamapS | negronjl: ^^ | 23:10 |
yagey | it doesn't. I'll try debug-log to see what up | 23:11 |
negronjl | SpamapS: I'll get to it in just a minute :) | 23:11 |
yagey | does juju need to run as root or sudo? | 23:12 |
SpamapS | yagey: the juju client should be fine running as your user | 23:14 |
SpamapS | yagey: the agents run as root, so the charms run as root | 23:14 |
yagey | 2012-04-10 16:14:14,908 Machine:0: twisted ERROR: Copy /var/cache/lxc/oneiric/rootfs-amd64 to /var/lib/lxc/yagey-sample-0-template/rootfs ... 2012-04-10 16:14:14,912 Machine:0: twisted ERROR: Copying rootfs to /var/lib/lxc/yagey-sample-0-template/rootfs ...Please change root-password ! | 23:15 |
lifeless | SpamapS: so whats the memgenerator url (page, not jpg) | 23:16 |
SpamapS | http://memegenerator.net/instance/18261941 | 23:16 |
SpamapS | so not as funny 10 minutes later | 23:16 |
lifeless | :) | 23:16 |
m_3 | yagey: I tested mongo replicasets using three nodes at a time (juju deploy ... -n3 local:mongodb) | 23:18 |
m_3 | yagey: didn't do a lot of starting one and then adding others one at a time... I'll bet that takes longer | 23:18 |
yagey | thanks, it appears to be a lxc issue with dependencies | 23:18 |
m_3 | yagey: should still work eventually though | 23:18 |
m_3 | ah, gotcha | 23:19 |
SpamapS | yagey: charm missing some deps? | 23:19 |
yagey | 2012-04-10 16:14:14,920 Machine:0: twisted ERROR: The following packages have unmet dependencies: 2012-04-10 16:14:14,920 Machine:0: twisted ERROR: ubuntu-minimal : Depends: netcat-openbsd but it is not going to be installed 2012-04-10 16:14:14,920 Machine:0: twisted ERROR: E: Unmet dependencies. Try 'apt-get -f install' with no packages (or specify a solution). 2012-04-10 16:14:14,928 Machine:0: twisted ERROR: failed to execute template | 23:19 |
SpamapS | weird | 23:20 |
* m_3 suspects incomplete lxc images... let it finish caching the first image (?) | 23:21 | |
negronjl | SpamapS: MongoDB README file done ... let me know if it's good enough | 23:29 |
SpamapS | negronjl: you type fast ;) | 23:29 |
yagey | *m_3 any suggestions? what kind of image? | 23:29 |
negronjl | SpamapS: I had it done already just forgot to push it ;) | 23:30 |
negronjl | SpamapS: err .. I mean .. yeah I typ really fast :) | 23:30 |
yagey | spamaps - can I see that README? | 23:30 |
SpamapS | yagey: yes, just pull the latest version of the charm | 23:32 |
SpamapS | yagey: should be visible here soon: http://jujucharms.com/charms/oneiric/mongodb | 23:32 |
m_3 | crap yagey left | 23:40 |
m_3 | hope that got worked out | 23:43 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!