shutdown timeout

9 thoughts on “shutdown timeout”

1. Nicolas Gasparotto September 17, 2012 at 17:14

   Well, you may have some 9i databases somewhere, but a new 9iR2 feature ?

   Nicolas.
2. Laurent Schneider Post authorSeptember 17, 2012 at 17:20

   according to metalink :
   This is applicable only to databases running on 9.2.X and above.

   but, no, I do not have 9i databases to check
3. Nicolas Gasparotto September 17, 2012 at 17:31

   Woaw, thanks for the ID reference.

   Nicolas.
4. Eric Grancher September 17, 2012 at 21:48

   Laurent,
   (if needed to shutdown abort) after shutdown abort, I normally do startup *restrict* (** for emphasis) then shutdown immediate.
   best
   Eric
5. Laurent Schneider Post authorSeptember 17, 2012 at 22:03

   yes, this is cleaner, approve and will do
6. DKB September 19, 2012 at 14:44

   In cases where the database MUST be shutdown at a specific time, there is no alternative to a “shutdown abort”, startup, shutdown immediate.
   Can I set _shutdown_completion_timeout_mins to a value of one?

   DKB
7. Laurent Schneider Post authorSeptember 19, 2012 at 16:30

   sure, but you could try a shutdown immediate at 2:30 and a shutdown abort at 2:59 to have your db shut down at 3:00 (not related to my timemout however)
8. Laurent Schneider Post authorSeptember 19, 2012 at 16:31

   putting a timeout of 1 would not be a good idea, not only it is not supported to set underscore parameters, but the chance to have an interupted shutdown will be to high and an interupted shutdown leave your db in a bad state
9. Ludovico Caldara November 27, 2012 at 17:08

   Hi Laurent,
   I know it looks a little complicated but that’s the way I did it for years…
   Lokk at the code to see how it works and feel free to take some ideas.
   Bests
   —
   Ludovico

   #!/bin/bash

   #################################################################################
   # – Enables shutdown tracing (events 10046 and 10400)
   # – Tries a shutdown IMMEDIATE in background
   # – If after X minutes the db is down, it exits 0
   # – If after X minuti the smon is still there:
   # – select * from x$ktuxe where ktuxecfl = ‘DEAD’;
   # – optionally kills all LOCAL=NO processes and wait 10 more seconds
   # – shutdown abort
   # – startup restrict
   # – shutdown immediate
   # – At the end, it checks again if smon is really dead.
   # – exit 0 if the db is really down

   #################################################################################

   #########
   #MOFIFIABLE VARS
   # log can be stdout to be catch by the father
   LOG=$HOME/stop.log
   #wait 300 seconds for immediate, then abort
   # sleep seconds is the interval between polls
   sleep_seconds=10
   # sleep loops is the number of polls before the abort
   sleep_loops=30
   #########

   ORACLE_SID=$1
   ORACLE_HOME=$2
   export ORACLE_SID ORACLE_HOME

   shutdown_immediate () {
   #tries shutdown immediate.
   # launched in background while the “MAIN” checks for effective shutdown
   $ORACLE_HOME/bin/sqlplus /nolog <>$LOG
   connect / as sysdba
   alter system checkpoint;
   alter system flush shared_pool;
   alter system checkpoint;
   alter session set events ‘10046 trace name context forever,level 12’;
   alter session set events ‘10400 trace name context forever, level 1’;
   shutdown immediate;
   exit
   EOF
   }

   shutdown_abort () {
   # after the “grace period” the MAIN triggers this function
   # it does shutdown abort + start/stop for clean shutdown
   $ORACLE_HOME/bin/sqlplus /nolog <>$LOG
   connect / as sysdba
   shutdown abort;
   startup restrict;
   shutdown immediate;
   exit
   EOF
   }

   check_smon_down () {
   ## if smon is up returns 0 else 1
   ps -eaf | grep -v grep | grep ora_smon_${ORACLE_SID}\$ >>$LOG
   }

   echo `date +%Y-%m-%d_%H:%M`” — Starting Shutdown instance $ORACLE_SID” >>$LOG
   shutdown_immediate &
   sh_i_pid=$!
   #sh_i_pid set to the PID of the shutdown immediate. (not used but nice to have)

   #looping sleep_loops times, waiting sleep_seconds each iteration.
   # if the instance shuts down in the meantime, exit 0. if at the end it’s still alive , abort
   while [ ${sleep_loops} -gt 0 ] ; do
   sleep ${sleep_seconds}
   check_smon_down
   if [ $? -eq 0 ] ; then
   echo `date +%Y-%m-%d_%H:%M`” — ($sleep_loops) Instance $ORACLE_SID active.” >>$LOG
   else
   echo `date +%Y-%m-%d_%H:%M`” — Instabce $ORACLE_SID shutdown.” >>$LOG
   exit 0
   fi
   sleep_loops=$(($sleep_loops-1))
   done

   echo `date +%Y-%m-%d_%H:%M`” — WARNING: Shutdown of $ORACLE_SID in HANG. Trying kill sessions LOCAL=NO.” >>$LOG
   ps -eaf | grep oracle${ORACLE_SID} | grep -v grep | grep LOCAL=NO >>$LOG
   ps -eaf | grep oracle${ORACLE_SID} | grep -v grep | grep LOCAL=NO | awk ‘{print $2}’ | xargs kill -9
   sleep 10
   check_smon_down
   if [ $? -eq 0 ] ; then
   echo `date +%Y-%m-%d_%H:%M`” — WARNING: Shutdown of $ORACLE_SID in HANG. Forcing shutdown abort.” >>$LOG
   shutdown_abort
   fi

   sleep 2
   check_smon_down
   if [ $? -eq 0 ] ; then
   ## wtf!!! 🙂
   echo `date +%Y-%m-%d_%H:%M`” — ERROR: Instance $ORACLE_SID still alive” >>$LOG
   exit 1
   else
   echo `date +%Y-%m-%d_%H:%M`” — Instance $ORACLE_SID shutdown.” >>$LOG
   exit 0
   fi

Comments are closed.