Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • Parallel MPI jobs can require 2 "bkill" commands issued within 5 seconds
  • If a job doesn't respond to the above, execute "bjobs -w" and find the name of the first node in the list.  Then execute "ssh <nodename> pkill --signal 9 -u <yourUserName>".  You may have to type your password.  This will kill all your processes on the head node (typically bkill will take care of processes on any MPI slave nodes)..
  • If the above doesn't work do "bkill -r <jobid>" where "jobid" can be seen with the "bjobs" command.  But use this only as a last resort, since it will leave old processes running on the batch nodes.

Getting An Interactive Session

NOTE: This is only permitted for the experiment that currently has beam.  You can get an interactive session using one of the nodes in psnehhiprioq/psfehhiprioq by executing the following from a psana node:  

bsub -q psnehhiprioq -n 16 -Is /bin/tcsh

(or /bin/bash if you use bash). Note that the above option is "Is" (capital-eye-esss) not "ls" (ell-ess).

One caveat: X11 graphics don't work in this mode (e.g. matplotlib).  To get X11 graphics, do the above command (to reserve the node), then "ssh psanaXXXX" (from pslogin or psana) on another window.

Remember to logout of all sessions when you are done with them (e.g. when you don't have beam).

Additional LSF References

...