| codebeake | hi all, is there a way to allow upstart jobs to be started multiple times? I have a daemon (Unicorn) which takes care of gracefully restarting, if you start it whilst it's already running, seems reasonable - but I can't find any relevant service stanza that would permit "multiple instances" | 10:09 |
|---|---|---|
| plautrba_ | codebeake: try init(5) - part Instances | 10:11 |
| codebeake | thanks plautrba_ :) | 10:11 |
| codebeake | hrm, "task" stanza looks promising, but I'm afraid then the `respawn` et al wouldn't work as expected | 10:14 |
| codebeake | (reading the section on instances below that, but the `task` was nice to find out about) | 10:15 |
| codebeake | plautrba_: thanks for the suggestion, I don't want to pass variables, shame I can't shell out and see if the pid exists, already (and then start a differently identified instance) | 10:21 |
| codebeake | It's also not possible to specify how to restart the task, right? They always receive the same signal, regardless | 10:42 |
| === soren_ is now known as soren | ||
| SpamapS | codebeake: wait, why do you want to start the job/task multiple times? | 16:23 |
| SpamapS | codebeake: re restarting, restart is always stop/start | 16:24 |
| SpamapS | codebeake: reload is always SIGHUP | 16:24 |
| SpamapS | codebeake: you can change the kill signal in newer versions of upstart with 'kill signal XXXXXX' | 16:24 |
| codebeake | SpamapS, because Unicorn is a special case, where via it's configuration file you can add a block of code, which gracefully restarts | 16:35 |
| codebeake | when sending the process URS2, then it gratefully restarts | 16:36 |
| SpamapS | codebeake: right, thats why apache isn't upstartified either :) | 16:37 |
| codebeake | I think my upstart version is too old, though - I'm on whatever ships with 10.10/10.4 Ubuntu | 16:37 |
| codebeake | ah, ok - so I might want to look at regular init scripts | 16:37 |
| SpamapS | codebeake: well for apache there's apachectl to do gracefuls, and then the upstart job would, in theory, just manage the boot/shutdown. | 16:37 |
| SpamapS | codebeake: but yeah, simpler to just use an init script if there's already one | 16:38 |
| greylurk | I'm writing an upstart script that is hanging when I try to call it. How do I figure out where it's logging to? | 16:49 |
| greylurk | it just sits at start/killed forever. | 16:49 |
| codebeake | I wonder SpamapS, if it's worth making a restart scropt, that just checks for the pidfile, and signals that pid | 16:54 |
| codebeake | I can chmod+xs it to my web application user | 16:54 |
| codebeake | and use it in my deployments, then | 16:54 |
| JanC | greylurk: jobs != scripts, so you can't "call" them ;) | 17:00 |
| JanC | and logging depends on the upstart version | 17:01 |
| JanC | (there is some info about logging in the cookbook) | 17:01 |
| JanC | codebeake: you can specify how to start/stop a service by using pre/post-start/stop instead of plain exec/script stanzas | 17:04 |
| codebeake | JanC, thanks | 17:11 |
| codebeake | the long and short is - that I need to be able to call start, since I don't want to replicate this logic, but I'll see what I can find | 17:11 |
| codebeake | this is the unicorn behaviour - https://gist.github.com/5471a06ab93f192178c2 | 17:11 |
| codebeake | If I can find a way to do that in upstart, then I'm sorted | 17:12 |
| JanC | http://upstart.ubuntu.com/cookbook/ --> has lots of useful info and tips & tricks | 17:12 |
| codebeake | I'm stuck with an old version, alas | 17:13 |
| codebeake | and I *know* I'll screw something up, if I try and upgrade a running production system's init system :) | 17:13 |
| codebeake | 0.6.5, right now | 17:13 |
| codebeake | (and when I USR2 it myself, upstart loses track of the pid… so, yeah - I think it might be time to find an alt. solution) | 17:32 |
| codebeake | thanks for the help and advice | 17:32 |
| codebeake | is it wise to do a source install of upstart, with a recent version on ubuntu 10.4 ? | 17:42 |
| codebeake | … I notice there's not a new enough version of libnih in apt to satisfy the 1.0 (at least) ./configure script, and I remember last time I tried to do this, it involved a trip down the rabithole | 17:43 |
| greylurk | Is there any way to run "initctl start" with debugging info? I have a script that's hanging at start/killed, but it works when I run the command from the command prompt. | 18:32 |
| SpamapS | greylurk: initctl log-priority info | 22:04 |
| SpamapS | greylurk: that will make init dump logs into syslog | 22:04 |
| SpamapS | greylurk: you can make it spit even more info with 'initctl log-priority debug' | 22:04 |
| greylurk | Thanks, it turned out that the init process was holding on to one specific pid, and wouldn't let it go even though the pid had died. I ended up having to reboot but it's fine now. | 22:43 |
| SpamapS | greylurk: thats a known bug when using 'expect fork' with a process that does not really fork | 22:43 |
| SpamapS | greylurk: another workaround to rebooting is to write a program that forks a whole bunch of times until it hits the desired pid | 22:44 |
| SpamapS | I really wish we could add an 'initctl forget job_name' | 22:44 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!