10 minutes countdown while stopping plugin

[matthieumarrast]

In metwork v2.1, when I restart mfdata, I am frequently blocked by a 10 minutes countdown while stopping a plugin:

- Waiting for stop of plugin: foo
    - Waiting for stop of step.foo[[..… ━╺━━━━━━━━━━ [ RUNNING ] 0:08:52

Impossible to exit this countdown, I must disconnect from the server, then retry the stop/start process.

So today I decide to open an issue in order to understand this delay.

[matthieumarrast]

Here is the circus config for this plugin:

[watcher:step.foo.main]
cmd = log_proxy_wrapper
args =  --stdout /home/mfdata/log/step_foo_main.log --stderr STDOUT -- plugin_wrapper foo -- /home/mfdata/var/plugins/foo/main.py   --step-name=main --redis-unix-socket-path='/home/mfdata/var/redis.socket' step.foo.main
numprocesses = 1
working_dir = /home/mfdata/var/plugins/foo
copy_env = True
shell = True
async_kill = True
autostart = False
respawn = True
autorestart_kill_signal=2
graceful_timeout = 600
max_retry = -1
mfdata_plugin = opmtransmet
max_age = 3600
max_age_variance = 3600
rlimit_as = 1000000000
rlimit_nofile = 1000
rlimit_stack = 10000000
rlimit_fsize = 10000000000

The 10 minutes timeout seems to be related to graceful_timeout = 600s

[matthieumarrast]

And the related code here: https://github.com/metwork-framework/mfext/blob/master/adm/circus.py#L191