ss_get_by_ssh.php does not timeout commands that hang

[Mike Benshoof]

When trying to fetch the remote command via ssh, there is a chance that the remote command may hang (df was the particular issue in this case).

This in turn causes the poller to back up and prevent other data sources from being populated and fills the cacti logs with warnings such as:

03/26/2013 03:00:01 PM - POLLER: Poller[0] WARNING: There are '1' detected as overrunning a polling process, please investigate

It would be helpful to have the ability to timeout the actual command as well as the base ssh call (with ConnectTimeout).

[Roman Vynar (roman-vynar)]

Mike, thank you for the patch!

[Roman Vynar (roman-vynar)]

Implementing two timeouts:
- $cmd_tout=10 - the timeout of ssh command itself or local cmd in case use_ssh is off;
- an option to enable SSH remote command timeout by prepending 'timeout $cmd_tout' to the actual command, i.e. `ssh HOST timeout 10 CMD`. Disabled by default because the timeout command can be missed on EL5 boxes (not a part of coreutils).

Why the second is important: when you run
ssh 192.168.8.1 sleep 9999
and then interrupts it, the ssh command itself stops but sleep 9999 will still be running on the target node.