| KjetilK | I got a bit further with my problem yesterday, but I could use a hand in translating this init.d-script to an upstart job: http://dolot.kipdola.com/wiki/Install_SASC-NG#Init_script (BTW, if anyone is in doubt, what I'm doing is legal here) | 18:12 |
|---|---|---|
| KjetilK | there is the runsasc script, which I suppose I could just exec | 18:18 |
| KjetilK | (the runsasc script is elsewhere on the same page) | 18:19 |
| KjetilK | there is a lot of additional complexity in there though, with timeouts and stuff, that is quite magical to me, and also I don't know if its needed under upstart, and also if there are better ways to deal with the insertion of the dvbloopback kernel module | 18:21 |
| ion | Is the module related to hardware? If yes, udev should autoload it. If not, it could be loaded with modprobe in a pre-start script. | 18:24 |
| ion | Something like this should work for starters: | 18:25 |
| ion | exec runsasc # no forking here | 18:25 |
| ion | post-start script | 18:25 |
| ion | until [ -e /tmp/SASC_COMPLETE ] ... | 18:26 |
| ion | (if timeout occurred) exit 1 | 18:26 |
| ion | end script | 18:26 |
| ion | Based on that, Upstart should handle the stopping when needed. | 18:27 |
| ion | Remember to rm -f /tmp/SASC_COMPLETE in pre-start script. | 18:27 |
| KjetilK | ion, ok, nice! | 18:27 |
| KjetilK | the module is only loosely related to hardware, in that it sets up a loopback device for a video stream that comes from hardware | 18:29 |
| KjetilK | so, I guess pre-start is better | 18:29 |
| KjetilK | hmmm, loading the driver is a large part of what runsasc does | 18:30 |
| KjetilK | perhaps I should try to do everything in the upstart job | 18:30 |
| KjetilK | hmmm, or perhaps not... | 18:32 |
| ion | Instead of the polling for the file, it might be better to use e.g. inotify. Incidentally, i wrote <http://github.com/ion1/wait-for-file> a couple of years ago, but that doesn’t support a timeout. OTOH, is the timeout really necessary? | 18:40 |
| ion | If runsasc fails, won’t it just return with a non-zero exit value? | 18:40 |
| KjetilK | I have no idea why they are using the timeout, this is too magic for me... | 18:47 |
| KjetilK | hmmm, it seems runsasc kills itself if it fails... | 18:47 |
| KjetilK | hmmm, perhaps I should try to write a minimal upstart job first, not looking too much at what runsasc does | 18:49 |
| KjetilK | and if that doesn't work, go from there... | 18:49 |
| KjetilK | runsasc has a lot of sleeps, is that a workaround for not having a proper event-based startup system? | 18:56 |
| KjetilK | i.e., something that could be done better with upstart? | 18:56 |
| ion | If it does any amount of preparation and then finally execs a single, independent server, there’s no need to split it. If it runs two independent services, it would be a good idea to create two jobs. | 19:05 |
| KjetilK | AFAICS, it just inserts a module, then starts a single daemon | 19:06 |
| KjetilK | but it seems to sit around and wait for things to come up | 19:07 |
| KjetilK | looks like those who wrote this had some problems they solved by doing this, but I don't know... | 19:07 |
| KjetilK | anyway, bbiab, gotta put the kid to bed :-) | 19:08 |
| * cwillu_at_work drops a hint about how he'd like to be able to run scripts as a user via "user <foo>" :p | 23:42 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!