Multi-process support, drain commands, etc...

[gshively11]

• Most commands now accept an optional process number as the first argument, (e.g. haproxyctl 2 enable all myserver, where 2 is the process number to target).
• Assumes all stats sockets are assigned an individual process number (ranges not supported).
• Defaults to process number 1 if nbproc < 2 OR the process number is not provided in the arguments.
• Removed state restrictions on enable/disable/drain commands. They don't appear to be needed.
• Updated help text
• Added drain all commands similar to the existing enable/disable all commands. These commands put servers in a drain state, which is useful for performing rolling deployments.

[zz9pzza]

+1 on wanting drain

[shey]

+1

[gshively11]

Added a reload_kill_zombies <wait_in_sec> command to deal with bugs in haproxy's reload (-sf) command. If you use peers, there are some race conditions that will leave connections in a CLOSE_WAIT state, keeping the old pids alive. It's hard to reproduce, but over the course of a week or two we'll end up having a couple of zombies. This new command will wait X number of seconds before terminating (and killing, if necessary) the old pids if necessary (but only if new processes start and the pidfile is updated with the new pids).

[fmluizao]

+1 for drain