worker_threads not honored entirely? pgrp signals?

[teichopsia]

hello, with salt 0.9.6 via pip:

when I start the salt-master nondaemonized, I have 8 processes.

when starting the master with worker_threads=1 in the master config file, I still have 8 processes.

when changing the default opt in config.py to 1, I now have 4 processes.

1. why isn't the configuration setting honored?
2. is there more recent code to offer a pidfile for the pid that handles signalling the entire process group for graceful shutdown? I recall some pidfile work but didn't see the extent yet.

[thatch45]

Looks to me like it is being honored, the master starts 3 processes and then "n" worker threads. It looks like you have a config error:
worker_threads=1 # not yaml
worker_threads: 1 # correct
please take a look at the config (there should also be an error in the logs)

[thatch45]

I missed #2
I am having issues with zeromq not dying correctly, in the current git HEAD the pidfile for the master records all PIDS

[teichopsia]

well I was implying assignment, not a direct yaml quote. the configuration is valid and accepted. it's just not being used.

[thatch45]

Hmm, so far I am unable to reproduce the issue, when I change my config I am consistently getting the correct number of processes. Would it be possible to post your master configuration file? I trust that it is correct, but another configuration might be producing a conflict (I have no idea how this might happen, but it is worth taking a look)

[teichopsia]

actually, closing. found the culprit. the environment pulled in a minion config instead of a master config due to salt unrelated virtualenv setup. due to the shared keywords it wasn't immediately obvious until we did get_opts dumps. sorry about that.

in general the pidfile is intended to be the single process that handles signals for the entire process group. most scripts expect invocations (/proc/cmdline) on subsequent lines and not additional pids.

multiple processes for salt-master is fine but there needs to be a clean way to signal exit of all processes via one pid.

[thatch45]

Thanks, I was wondering what could cause this issue. As for the pidfile, yes, I agree, and I am working on cleaning it up, thanks for your help!