| mitsuhiko | jodh: there is nothing wrong with the job | 09:51 |
|---|---|---|
| mitsuhiko | feel free to doubt that, but unfortunately that is the case | 09:52 |
| SpamapS | mitsuhiko: I would agree. The problem is upstart's fragility with regard to following forks. | 16:28 |
| SpamapS | mitsuhiko: the two options for avoiding this bug, are to not use expect fork/daemon (run in the foreground) or fore-go the process tracking of upstart and use pre-start/post-stop to create a pid file/kill said pid. | 16:29 |
| mitsuhiko | why does upstart track daemonizing behavior at all anyways? | 16:30 |
| mitsuhiko | there are much better ways than that | 16:31 |
| SpamapS | boot speed | 16:31 |
| SpamapS | mitsuhiko: for some of the early plumbing, we want to start running dependant tasks/jobs ASAP.. and often, on fork of a daemon is ASAP | 16:32 |
| mitsuhiko | what's wrong with cgroups? | 16:33 |
| mitsuhiko | then you can see what damage the proess did after the fact | 16:33 |
| SpamapS | nothing is wrong with cgroups | 16:37 |
| SpamapS | they did not exist when upstart was written | 16:37 |
| SpamapS | mitsuhiko: there is a mid-term plan to provide an alternative to the ptrace technique that expect * uses. | 16:46 |
| mitsuhiko | SpamapS: ignore my snark. upstart just caused me so many issues this week :-/ | 19:48 |
| SpamapS | mitsuhiko: no worries. It has caused me quite a few issues the last 2.5 years ;) | 22:04 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!