dmtcpckpt.1

NAME

       dmtcpckpt - manual page for dmtcpckpt 1

SYNOPSIS

       dmtcpckpt [options] [[--] program args]

DESCRIPTION

       DMTCP  checkpointing  support  for  SGE,  covering running, checkpoint,
       migrate and clean.

ACTIONS

       Only one can be specified, default -r:

       -c command

              Behave like "-r sh -c 'command'", e.g. for  use  with  qsub  -S.
              Quoting may be problematic due to an extra shell expansion.

       -k     Clean up (stop processes and delete dir)

       -m     Migrate (checkpoint and exit with code 99)

       -p     Make a checkpoint

       -r     Run program args under checkpointing (default action).  Tries to
              restart an existing checkpoint if RESTARTED is not 0.  Due to an
              apparent  bug  in DMTCP 2.4, at least, <program> is not found on
              PATH, and so must be a file name.

OPTIONS

       -d dir Directory in which to write checkpoints (default '/'.)

       -s signal
              Signal to cause a checkpoint (e.g. "10", "USR1"), probably as in
              checkpoint(5).

       Signals caught:

              USR1, and signal, as above, cause a checkpoint;

              USR2  initiates  the  migrate  action,  and  returns 99 to cause
              rescheduling.

       Files ~/.dmtcpckpt And .dmtcpckpt are sourced in that  order,  if  they
       exist,    to    supply    hook    variables.    Of   these,   variables
       {pre,post}_{migrate,checkpoint}_hook and pre_launch_hook are  evaluated
       as  commands  before/after  the  relevant  actions, and $launch_opts is
       expanded as extra options for dmtcp_launch when  program  is  run  ini-
       tially,  e.g.  to  use plugins, or add --ckpt-open-files or --interval.
       The migrate and checkpoint actions call dmctp_command with  the  appro-
       priate coordinator commands as arguments.

       The  RESTARTED  environment  variable  must have a valid value per sub-
       mit(1), and is used to decide whether or not to start from scratch.

       Example checkpoint(5):

              ckpt_name          dmtcp
              interface          application-level
              ckpt_command       $sge_root/site/dmtcpckpt -p
              migr_command       $sge_root/site/dmtcpckpt -m
              restart_command    NONE
              clean_command      $sge_root/site/dmtcpckpt -k
              ckpt_dir           /scratch/$job_owner/checkpoints
              signal             NONE
              when               xs

SEE ALSO

       dmtcp(1), checkpoint(5), sge_ckpt(5).



dmtcpckpt 1                      November 2015                    DMTCPCKPT(1)