Stop Jobs

A job can be stopped when it is running or queued. Stopping a job does not "forget it" from the vovserver database. A job can only be stopped by the owner or the Accelerator administrator.

nc stop


nc: Usage Message

NC STOP:
   Stop jobs.
   1. If the jobs are running, they are killed
      (unless you use option -dequeueonly).
   2. If the jobs are scheduled in the queue,
      they are removed from the queue.

   In either case, the jobs remain in the system.
   To remove them from the system, use the "forget" command.
   Jobs in the system can be rerun with the "rerun" command.

   When stopping a single job, the procedure checks for
   the properties NC_STOP_SIGNALS and NC_STOP_SIG_DELAY
   attached to the job to be stopped.

   The list of signals is also controlled
   by the environment variables VOV_STOP_SIGNALS and NC_STOP_SIGNALS.
   If both NC_STOP_SIGNALS and VOV_STOP_SIGNALS are present in the
   environment, the value of VOV_STOP_SIGNALS will be used.  Their
   functionality is otherwise identical.

   The default list of signals is TERM,HUP,INT,KILL and
   can be customized with the variable defaultStopSignalCascade in policy.tcl.

USAGE:
    % nc stop [OPTIONS] <jobId> ...

OPTIONS:
    -after   <s>        -- Start sending signals after specified seconds.
                           This is an initial delay, between 0 and 20s.
    -allusers           -- Stop all jobs (only ADMIN can do it).
    -d                  -- Same as -dequeueonly.
    -delay   <s>        -- Minimum delay between signals (in seconds),
                           between 0 and 20s. Default is 3.
                           This can also be set with the property
                           NC_STOP_SIG_DELAY, or with the environment
                           variables NC_STOP_SIG_DELAY or
                           VOV_STOP_SIGNAL_DELAY.  If both NC_STOP_SIG_DELAY
                           and VOV_STOP_SIGNAL_DELAY are present in the
                           environment, the value of NC_STOP_SIGNAL_DELAY
                           will be used.
                           Priority: 1. Option -delay
                                     2. job property NC_STOP_SIG_DELAY
                                     3. env variable VOV_STOP_SIGNAL_DELAY,
                                        NC_STOP_SIG_DELAY
                                     4. default
    -dequeueonly        -- Just remove jobs from the queue.
                           All currently running jobs are not affected.
                           Can be abbreviated to -d.
    -dir <directory>    -- Stop all jobs in the given directory.
    -exclude <PROCLIST> -- List of processes to exclude from receiving the
                           signal.
    -h                  -- This message
    -include <PROCLIST> -- List of processes to receive the signal.
    -J <jobname>        -- Stop all my jobs with given jobname.
    -mine               -- Stop all my jobs.
    -set <setname>      -- Stop all my jobs in the given set.
    -sig     <SIGLIST>  -- Same as -signals.
    -signals <SIGLIST>  -- Comma separated list of signals to send to the jobs
                           (default is the sequence TERM,HUP,INT,KILL )
                           This can be also set with property NC_STOP_SIGNALS
                           or with the environment variables NC_STOP_SIGNALS
                           or VOV_STOP_SIGNALS.
                           Priority: 1. Option -signals
                                     2. job property NC_STOP_SIGNALS
                                     3. env variable VOV_STOP_SIGNALS,
                                        NC_STOP_SIGNALS
                                     4. default (can be configured as
                                        defaultStopSignalCascade
                                        in policy.tcl)
                           See also:  vovshow -env VOV_STOP_SIGNALS
                                      vovshow -env NC_STOP_SIGNALS
    -skiptop <0|1>      -- Whether to kill the top process.  
                           This is normally the job wrapper (e.g. vw, vwi).
                           Default is 0.
    -why <reason>       -- Give a reason for the stop.
                           This is stored on the WHYSTATUS
                           field of the stopped jobs.

EXAMPLES:
    % nc stop 00123456
    % nc stop -d -mine
    % nc stop -after 3 -mine
    % nc stop -set Class:hsim
    % nc stop -mine -why "Jobs no longer needed"
    % nc stop -sig "TERM,KILL" -delay 4 0012345
    % env VOV_STOP_SIGNALS=TERM,INT,KILL nc stop 0012345

SEE ALSO:
    % vovshow -env VOV_STOP_SIGNALS
    % vovshow -env NC_STOP_SIGNALS
    % vovshow -env VOV_STOP_SIGNAL_DELAY
    % vovshow -env NC_STOP_SIG_DELAY

Override Signals to Stop a Job

A job can be stopped by overriding the sequence of signals that are sent for the job. To do so, set the properties NC_STOP_SIGNALS and NC_STOP_SIG_DELAY.

Automatic Stopping Based on Elapsed Time

If a job is submitted with the -autokill option, it will be stopped after the specified amount of time has elapsed. The check to stop the job is performed by the tasker itself at an interval of about one minute, which can be controlled with the -U option of vovtasker).

For example:

% nc run -autokill 30m sleep 1000000

Automatic Stopping Based on CPU Time

To stop a job that exceeds a specific duration of CPU time, set the variable VOV_LIMIT_cputime. A job that exceeds the limit will be killed by UNIX and will have status "Failed".

For example:

% nc run -e "BASE+D(VOV_LIMIT_cputime=10)" vovmemtime 10 100 0