dbprocs, at least in our environment, has two flaws:

o "dbprocs stop" tends to leave one or two processes still running.
o Because dbprocs is monolithic, it's hard to tell what those processes are.

In order to correct these flaws, Alexis rewrote dbprocs as four small
programs.  I like his rewrite a lot and I'd like you to consider it as
a replacement for the existing dbprocs.

This suite should exactly replicate the behavior of the original,
with two small exceptions:

o When dbp_logrm exits due to $DB_HOME not being set, it now emits an
  error message and exits with a non-zero status.
o dbp_logrm now checks to see if it's already running, as the other
  two already did.

Comments?

-- 
Ben Rosengart     (212) 741-4400 x215

1. A robot may not injure entertainment industry profits, or, through inaction,
   allow entertainment industry profits to come to harm.     --Matt McLeod


-- Attached file included as plaintext by Ecartis --

#! /bin/sh
. /news/lib/innshellvars

# dbp_chkpt_watch
#   Start the checkpointer, write its PID to a lockfile, and remove it
#   once the checkpointer exits

if [ -z "$DB_HOME" ]; then
   exit
fi

cd $PATHTMP

if [ -f $PATHRUN/db_checkpoint.pid ] ; then
  if kill -0 `cat $PATHRUN/db_checkpoint.pid` ; then
    echo "$0 (dbprocs): db_checkpoint already running, not started."
  else
    echo "Warning: stale db_checkpoint lock detected (and ignored)."
  fi
fi

db_checkpoint -p 1 -k 2000 &
echo $! > $PATHRUN/db_checkpoint.pid
wait
rm -f $PATHRUN/db_checkpoint.pid


-- Attached file included as plaintext by Ecartis --

#! /bin/sh
. /news/lib/innshellvars

# dbp_deadlk_watch
#   start the deadlock detector, write its PID to a lockfile, and remove it
#   once the detector exits

if [ -z "$DB_HOME" ]; then
   exit
fi

cd $PATHTMP

if [ -f $PATHRUN/db_deadlock.pid ] ; then
  if kill -0 `cat $PATHRUN/db_deadlock.pid` ; then
    echo "$0 (dbprocs): db_deadlock already running, not started."
  else
    echo "Warning: stale db_deadlock lock detected (and ignored)."
  fi
fi

db_deadlock -t 30 &
echo $! > $PATHRUN/db_deadlock.pid
wait
rm -f $PATHRUN/db_deadlock.pid


-- Attached file included as plaintext by Ecartis --

#! /bin/sh
. /news/lib/innshellvars

#
# dbp_rmlogs
#   Remove db logs that are no longer needed

if [ -z "$DB_HOME" ]; then
   echo "\$DB_HOME is not set, exiting."
   exit 1
fi

cd $PATHTMP

if [ -f $PATHRUN/logremover.pid ] ; then
  if kill -0 `cat $PATHRUN/logremover.pid` ; then
    echo "$0 (dbprocs): logremover already running, not started."
  else
    echo "Warning: stale logremover lock detected (and ignored)."
  fi
fi

(
  trap "rm -f $PATHTMP/dba.$$ $PATHRUN/logremover.pid ; exit" 1 2 3 15

  while [ 1 ];
  do
    sleep 120
    db_archive -a >$PATHTMP/dba.$$
    if [ -s $PATHTMP/dba.$$ ]; then
      xargs rm <$PATHTMP/dba.$$
    fi
  done
) & 
echo $! > $PATHRUN/logremover.pid


-- Attached file included as plaintext by Ecartis --

#! /bin/sh
. /news/lib/innshellvars

#
# dbprocs
#   start or stop processes that support the BerkeleyDB ovdb environment
#

if [ -z "$DB_HOME" ]; then
   exit
fi

cd $PATHTMP

case "$1" in

stop)

  if [ -f $PATHRUN/db_deadlock.pid ]; then
    kill `cat $PATHRUN/db_deadlock.pid`
  fi
  if [ -f $PATHRUN/logremover.pid ]; then
    kill `cat $PATHRUN/logremover.pid`
  fi
  if [ -f $PATHRUN/db_checkpoint.pid ]; then
    kill `cat $PATHRUN/db_checkpoint.pid`
    sleep 1

    # db_checkpoint may not exit right away if it is in the
    # middle of a checkpoint.  So we wait for it here.
    i=12
    while [ $i -gt 0 ];
    do
      [ -f $PATHRUN/db_checkpoint.pid ] || break
      sleep 5
      i=`expr $i - 1`
    done

    # run one more checkpoint just to make sure, but not if
    # the old one is still running
    [ -f $PATHRUN/db_checkpoint.pid ] || db_checkpoint -1
  fi
  ;;

start)

  # Run checkpointer and deadlock detector
  ( exec dbp_chkpt_watch ) &

  ( exec dbp_deadlk_watch ) &

  # Remove logs that are no longer needed
  exec dbp_rmlogs
  ;;

*)
  echo "Usage: $0 <start|stop>"
  ;;

esac