cl_sig - signal programs on the cluster

cl_kill - send programs the kill signal


Synopsis:

usage: cl_sig [options] progname
     -1, -2              send SIGUSR1/SIGUSR2
     -9                  send SIGKILL (signal #9)
     -s name             send SIG-name
     -t                  test only, don't execute
     -v                  verbose output
     -V                  really verbose output


Description:
cl_sig is used to send Unix signals to all programs of a given name that are currently running on the cluster. In many cases, this will be the signal to kill stray jobs (after one of the nodes crashed). Note that there is no guarantee of synchrony in the signals, in fact, they can be guaranteed NOT to be delivered to all machine at the exact same time (the signal command is given inside a loop so it is necessarily one machine at a time). Thus, you should not use cl_sig if you need the signals to trigger synchronous activity on the nodes (ie. beware of using SIGUSR1 to trip a checkpoint).

cl_sig is a Perl script that simply automates the process of rsh-ing to the remote machines, doing a ps there, scanning for the named program, then issuing a kill on the remote machine.

cl_kill is a symbolic link to the cl_sig script.


Options:
By default, cl_sig sends SIGINT.

By default, cl_kill sends SIGKILL.

To override these signal names, you can select -1 to send SIGUSR1, option -2 sends SIGUSR2, and option -9 sends SIGKILL.

To select a named or numbered signal, use the -s option. To see a list of available signals, run the (serial) command "kill -l".


RCSID $Id: cl_sig.html,v 1.2 2002/03/12 14:11:18 jpormann Exp $