Added command-line argument that allows the PID folder to be specified for the worker processes.
(Why? Even though a user can launch multiple god processes, all of them write pids to the same ~/.god/pids directory. Thus, each god process can clobber the PID files for the other god processes. Specifying a separate --managed_pid_dir for each god process ensures that they will not clobber one another)
PID dir can be specified for worker processes
I think this setting would be fine without a short option. Could you remove the -M option?
Also, the command line options use - between words instead of _.
Sounds good. I'm on it.
Updated command-line options for managed pids dir
Any ETA on merging this into master?
cronjob to restart sidekiq workers daily, might help with the memoryo…
Merge remote-tracking branch 'upstream/master'
Stops failing silently when the godfile is wrong.
Updates version number.
Merge pull request #1 from sdubinsky/master
god command checks that the daemon process actually launched.
God exits w/ status 1 when daemon fails to launch
Fixed another version string in the code
Is there an ETA on merging this into master? My company has been running this branch since May 2014, so we're fairly certain that it works as expected.
Merged in changes from upstream master
It looks like there's a separate change in here that is sleeping. Can you remove that?
We found that the god command fails silently if it encounters an error in its configuration file. The sleep statement is a bit of a hack, but it allows the god command to verify that the workers are actually running before it declares victory.
Do you want me to move the sleep hack to a separate merge request? Or, would you rather that I remove the sleep hack entirely (and, possibly replace it w/ something more resilient)?
I definitely think it should be a separate discussion. I'd be interested in talking about the sorts of errors you're talking about, but I'd rather if it was not mixed in with this.
Removed daemon launch verification logic
Alright, I removed the daemon launch verification logic. Sometime tomorrow, we can talk about the problem that it was intended to solve, and probably come up w/ a better way to address it.
Removed a newline character to force a travis-ci rebuild