| SpamapS | sbuild does not.. something w/ the aufs I think | 00:45 |
|---|---|---|
| hunterloftis | So - I'm trying to issue a 'killall' command to clean stuff up before a job is started. But if a matching process isn't killed, I guess killall returns a failed exit code (?) because then the job "fails to start" | 02:54 |
| ion | What are you killing in pre-start? | 02:55 |
| hunterloftis | ion - killing old instances | 02:55 |
| hunterloftis | This process sometimes doesn't clean up after itself | 02:56 |
| hunterloftis | And it creates child processes | 02:56 |
| ion | Shouldn’t that kluge be placed in post-stop? | 02:56 |
| hunterloftis | It also binds to a port, so if any of the worker processes still exist on the next attempt at restarting it, you get an address in use error | 02:56 |
| hunterloftis | I plan on putting it in both places | 02:56 |
| ion | Or perhaps pre-stop. | 02:56 |
| hunterloftis | Why not pre-start? | 02:57 |
| ion | Isn’t that redundant? | 02:57 |
| hunterloftis | Well, if I had to choose, pre-start would be the best place in that case | 02:57 |
| hunterloftis | Because I don't care about the workers except when I'm trying to start a new process | 02:57 |
| hunterloftis | So cleaning up after is much less important than preparing at the beginning | 02:57 |
| ion | When Upstart thinks a job is in the stopped state (which is a prerequisite for starting something), it *should* be. | 02:58 |
| ion | It’s very nasty to leave processes just laying around even though the job claims to be stopped and then try to clean them up just before starting it again. | 02:58 |
| hunterloftis | Then why not just do it on both ends? | 02:58 |
| frewsxcv | to get a script working, do i just copy my conf file into /etc/init? and then just do 'start script'? | 06:26 |
| codebeaker | is it intentional that daemons started with upstart (ubuntu 10.x) don't inherit limits from /etc/security/limits.conf ? | 14:45 |
| frewsxcv | so if i create an upstart conf, shouldn't i be able todo 'start myscript' immediately? | 14:56 |
| jhunt | codebeaker: upstart doesn't use pam currently. | 14:57 |
| codebeaker | ok, so it's planned/unplanned/etc - but the limits stanza is just as sane (once you know about it) | 14:58 |
| jhunt | frewsxcv: yes, assuming the script has valid syntax - try running "init-checkconf file.conf" | 14:58 |
| codebeaker | I've inherited a somewhat badly written upstart script via a ppa package, and was surprised that the software basically doesn't work out of the ppa, didn't know that there was a limits option | 14:58 |
| jhunt | l | 15:00 |
| frewsxcv | jhunt: i don't have that command | 15:01 |
| jhunt | assuming the file is in /etc/init/, run "initctl log-priority debug && touch /etc/init/myfile.conf". Then, tail you system log and look for the name of your .conf file. | 15:02 |
| jhunt | Upstart will have logged if it has read it or if it is syntactically incorrect. | 15:02 |
| jhunt | frewsxcv: alternatively, do this: http://upstart.ubuntu.com/cookbook/#older-versions-of-upstart | 15:03 |
| jhunt | frewsxcv: (just to check the code blocks) | 15:03 |
| frewsxcv | thanks jhunt | 15:06 |
| jhunt | frewsxcv: np. | 15:06 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!