Open castarco opened 10 years ago
While I'll keep this issue open to investigate whether we should change supervisorctl, note that it's totally possible (and even encouraged really) to write a custom XML-RPC client that serially starts and stops some number of processes.
I would love to see this feature in supervisor and be able to configure the restart policy per process group.
At this time, the
do_restart
method in supervisorctl does ado_stop
and then ado_start
. If we have many workers for a process group, it would be nice to add the possibility to restart every worker in a sequential way, one worker after another worker (to easily achieve zero downtime).In my case, I have one worker per CPU, and I'm using a local HAProxy to "simulate" a single server.
I'd like to read the main developer's opinion. If they don't want to do it I'll be grateful to receive advice to contribute with my code... and if it's "impossible", i'll be gratefull to be informed about ^^U .
Thanks in advance!