Page MenuHomeClusterLabs Projects
Files F4185693

No OneTemporary
Actions

View Options

diff --git a/heartbeat/oracle b/heartbeat/oracle
index 1a2ab3d34..1b44e674b 100755
--- a/heartbeat/oracle
+++ b/heartbeat/oracle
@@ -1,681 +1,681 @@
#!/bin/sh
#
#
# oracle
#
# Description: Manages an Oracle Database as a High-Availability
# resource
#
#
# Author: Dejan Muhamedagic
# Support: linux-ha@lists.linux-ha.org
# License: GNU General Public License (GPL)
# Copyright: (C) 2006 International Business Machines, Inc.
#
# This code inspired by the DB2 resource script
# written by Alan Robertson
#
# An example usage in /etc/ha.d/haresources:
# node1 10.0.0.170 oracle::RK1::/oracle/10.2::orark1
#
# See usage() function below for more details...
#
# OCF instance parameters:
# OCF_RESKEY_sid
# OCF_RESKEY_home (optional; else read it from /etc/oratab)
# OCF_RESKEY_user (optional; figure it out by checking file ownership)
# OCF_RESKEY_ipcrm (optional; defaults to "instance")
# OCF_RESKEY_clear_backupmode (optional; default to "false")
# OCF_RESKEY_shutdown_method (optional; default to "checkpoint/abort")
#
# Initialization:
. ${OCF_ROOT}/resource.d/heartbeat/.ocf-shellfuncs
#######################################################################
usage() {
methods=`oracle_methods`
methods=`echo $methods | tr ' ' '|'`
cat <<-!
usage: $0 {$methods}
$0 manages an Oracle Database instance as an HA resource.
The 'start' operation starts the database.
The 'stop' operation stops the database.
The 'status' operation reports whether the database is running
The 'monitor' operation reports whether the database seems to be working
The 'dumpinstipc' operation prints IPC resources used by the instance
The 'cleanup' operation tries to clean up after Oracle was brutally stopped
The 'validate-all' operation reports whether the parameters are valid
The 'methods' operation reports on the methods $0 supports
!
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="oracle">
<version>1.0</version>
<longdesc lang="en">
Resource script for oracle. Manages an Oracle Database instance
as an HA resource.
</longdesc>
<shortdesc lang="en">oracle resource agent</shortdesc>
<parameters>
<parameter name="sid" unique="1" required="1">
<longdesc lang="en">
The Oracle SID (aka ORACLE_SID).
</longdesc>
<shortdesc lang="en">sid</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="home" unique="0">
<longdesc lang="en">
The Oracle home directory (aka ORACLE_HOME).
If not specified, then the SID along with its home should be listed in
/etc/oratab.
</longdesc>
<shortdesc lang="en">home</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="user" unique="0">
<longdesc lang="en">
The Oracle owner (aka ORACLE_OWNER).
If not specified, then it is set to the owner of
file \$ORACLE_HOME/dbs/*\${ORACLE_SID}.ora.
If this does not work for you, just set it explicitely.
</longdesc>
<shortdesc lang="en">user</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="ipcrm" unique="0">
<longdesc lang="en">
Sometimes IPC objects (shared memory segments and semaphores)
belonging to an Oracle instance might be left behind which
prevents the instance from starting. It is not easy to figure out
which shared segments belong to which instance, in particular when
more instances are running as same user.
What we use here is the "oradebug" feature and its "ipc" trace
utility. It is not optimal to parse the debugging information, but
I am not aware of any other way to find out about the IPC
information. In case the format or wording of the trace report
changes, parsing might fail. There are some precautions, however,
to prevent stepping on other peoples toes. There is also a
dumpinstipc option which will make us print the IPC objects which
belong to the instance. Use it to see if we parse the trace file
correctly.
Three settings are possible:
- none: don't mess with IPC and hope for the best (beware: you'll
probably be out of luck, sooner or later)
- instance: try to figure out the IPC stuff which belongs to the
instance and remove only those (default; should be safe)
- orauser: remove all IPC belonging to the user which runs the
instance (don't use this if you run more than one instance as same
user or if other apps running as this user use IPC)
The default setting "instance" should be safe to use, but in that
case we cannot guarantee that the instance will start. In case IPC
objects were already left around, because, for instance, someone
mercilessly killing Oracle processes, there is no way any more to
find out which IPC objects should be removed. In that case, human
intervention is necessary, and probably _all_ instances running as
same user will have to be stopped. The third setting, "orauser",
guarantees IPC objects removal, but it does that based only on IPC
objects ownership, so you should use that only if every instance
runs as separate user.
Please report any problems. Suggestions/fixes welcome.
</longdesc>
<shortdesc lang="en">ipcrm</shortdesc>
<content type="string" default="instance" />
</parameter>
<parameter name="clear_backupmode" unique="0" required="0">
<longdesc lang="en">
The clear of the backup mode of ORACLE.
</longdesc>
<shortdesc lang="en">clear_backupmode</shortdesc>
<content type="boolean" default="false" />
</parameter>
<parameter name="shutdown_method" unique="0" required="0">
<longdesc lang="en">
How to stop Oracle is a matter of taste it seems. The default
method ("checkpoint/abort") is:
alter system checkpoint;
shutdown abort;
This should be the fastest safe way bring the instance down. If
you find "shutdown abort" distasteful, set this attribute to
"immediate" in which case we will
shutdown immediate;
If you still think that there's even better way to shutdown an
Oracle instance we are willing to listen.
</longdesc>
<shortdesc lang="en">shutdown_method</shortdesc>
<content type="string" default="checkpoint/abort" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="120" />
<action name="stop" timeout="120" />
<action name="status" timeout="5" />
<action name="monitor" depth="0" timeout="30" interval="120" start-delay="10" />
<action name="validate-all" timeout="5" />
<action name="methods" timeout="5" />
<action name="meta-data" timeout="5" />
</actions>
</resource-agent>
END
}
#
# methods: What methods/operations do we support?
#
oracle_methods() {
cat <<-!
start
stop
status
monitor
dumpinstipc
cleanup
validate-all
methods
meta-data
usage
!
}
# Gather up information about our oracle instance
ora_info() {
ORACLE_SID=$1
ORACLE_HOME=$2
ORACLE_OWNER=$3
# get ORACLE_HOME from /etc/oratab if not set
[ x = "x$ORACLE_HOME" ] &&
ORACLE_HOME=`awk -F: "/^$ORACLE_SID:/"'{print $2}' /etc/oratab`
# there a better way to find out ORACLE_OWNER?
[ x = "x$ORACLE_OWNER" ] &&
ORACLE_OWNER=`ls -ld $ORACLE_HOME/. 2>/dev/null | awk 'NR==1{print $3}'`
sqlplus=$ORACLE_HOME/bin/sqlplus
lsnrctl=$ORACLE_HOME/bin/lsnrctl
tnsping=$ORACLE_HOME/bin/tnsping
}
testoraenv() {
# Let's make sure a few important things are set...
[ x != "x$ORACLE_HOME" -a x != "x$ORACLE_OWNER" ] ||
return 1
# and some important things are there
[ -x "$sqlplus" -a -x "$lsnrctl" -a -x "$tnsping" ] ||
return 1
return 0
}
setoraenv() {
LD_LIBRARY_PATH=$ORACLE_HOME/lib
LIBPATH=$ORACLE_HOME/lib
TNS_ADMIN=$ORACLE_HOME/network/admin
PATH=$ORACLE_HOME/bin:$ORACLE_HOME/dbs:$PATH
export ORACLE_SID ORACLE_HOME ORACLE_OWNER TNS_ADMIN
export LD_LIBRARY_PATH LIBPATH
}
#
# Run commands as the Oracle owner...
#
execsql() {
if [ "$US" = "$ORACLE_OWNER" ]; then
$sqlplus -S /nolog
else
su - $ORACLE_OWNER -c "$sqlplus -S /nolog"
fi
}
#
# Run commands in the oracle admin sqlplus...
#
dbasql() {
typeset func
(
echo "connect / as sysdba"
echo "set feedback off"
echo "set heading off"
echo "set pagesize 0"
for func; do $func; done
) |
execsql | grep -v '^Connected'
}
#
# various interesting sql
#
dbstat() {
echo 'select status from v$instance;'
}
dbmount() {
echo 'alter database mount;'
}
dbopen() {
echo 'alter database open;'
}
dbstop_immediate() {
echo 'shutdown immediate'
}
dbstop_checkpoint_abort() {
echo 'alter system checkpoint;'
echo 'shutdown abort'
}
dbstop() {
case "${shutdown_method}" in
"immediate")
dbstop_immediate
;;
"checkpoint/abort")
dbstop_checkpoint_abort
;;
esac
}
dbstart() {
echo 'startup'
}
dbstart_mount() {
echo 'startup mount'
}
dbendbackup() {
echo 'alter database end backup;'
}
db_backup_mode() {
echo "select 'COUNT'||count(*) from v\$backup where status='ACTIVE';"
}
is_clear_backupmode_set(){
[ x"${clear_backupmode}" = x"true" ]
}
is_instance_in_backup_mode() {
typeset count
count="`dbasql db_backup_mode | sed 's/COUNT//'`"
[ x"$count" != x"0" ]
}
clear_backup_mode() {
typeset output
output="`dbasql dbendbackup`"
ocf_log info "Oracle instance $ORACLE_SID alter database end backup: $output"
}
getdumpdest() {
#echo 'select value from v$parameter where name = \'user_dump_dest\';'
echo "select value from v\$parameter where name = 'user_dump_dest';"
}
getipc() {
echo "oradebug setmypid"
echo "oradebug ipc"
}
#
# IPC stuff: not overly complex, but quite involved :-/
#
# Part 1: Oracle
dumpinstipc() {
typeset dumpdest=`dbasql getdumpdest`
[ "x$dumpdest" != x -a -d "$dumpdest" ] || return 1
typeset -i fcount=`ls -rt $dumpdest | wc -l`
dbasql getipc >/dev/null 2>&1
typeset lastf=`ls -rt $dumpdest | grep -v '^\.*$' | tail -1`
typeset -i fcount2=`ls -rt $dumpdest | wc -l`
[ $((fcount+1)) -eq $fcount2 ] || return 1 # more than one file created
echo $dumpdest/$lastf
}
parseipc() {
typeset inf=$1
test -f "$1" || return 1
awk '
$3 == "Shmid" {n=1;next}
n {
if( $3~/^[0-9]+$/ ) print $3;
n=0
}
' $inf |
sort -u | sed 's/^/m:/'
awk '
/Semaphore List/ {insems=1;next}
insems {
for( i=1; i<=NF; i++ )
if( $i~/^[0-9]+$/ ) print $i;
}
/system semaphore information/ {exit}
' $inf |
sort -u | sed 's/^/s:/'
}
# Part 2: OS (ipcs,ipcrm)
filteroraipc() { # this portable?
grep -w $ORACLE_OWNER | awk '{print $2}'
}
ipcdesc() {
typeset what=$1
case $what in
m) echo "shared memory segment";;
s) echo "semaphore";;
q) echo "message queue";;
esac
}
rmipc() {
typeset what=$1 id=$2
ipcs -$what | filteroraipc | grep -w $id >/dev/null 2>&1 ||
return
ocf_log info "Removing `ipcdesc $what` $id."
ipcrm -$what $id
}
ipcrm_orauser() {
typeset what id
for what in m s q; do
for id in `ipcs -$what | filteroraipc`; do
rmipc $what $id
done
done
}
ipcrm_instance() {
typeset ipcobj
for ipcobj; do
rmipc `echo $ipcobj | sed 's/:/ /'`
done
}
#
# oracle_status: is the Oracle instance running?
#
# quick check to see if the instance is up
is_oracle_up() {
ps -ef | grep -wiqs "[^ ]*[_]pmon_${ORACLE_SID}"
}
# instance in OPEN state?
instance_live() {
[ "`dbasql dbstat`" = OPEN ]
}
ora_cleanup() {
#rm -fr /tmp/.oracle #???
rm -f `ls $ORACLE_HOME/dbs/lk* | grep -i $ORACLE_SID`
#return
case $IPCRM in
none)
;;
instance)
ipcrm_instance $*
;;
orauser)
ipcrm_orauser $*
;;
*)
ocf_log warn "bad usage: ipcrm set to $IPCRM"
;;
esac
}
#
# oracle_start: Start the Oracle instance
#
# NOTE: We handle instance in the MOUNTED and STARTED states
# efficiently
# We *do not* handle instance in the restricted or read-only
# mode, i.e. it appears as running, but its availability is
# "not for general use"
#
oracle_start() {
typeset status output
if is_oracle_up; then
status="`dbasql dbstat`"
case "$status" in
"OPEN")
: nothing to be done, we can leave right now
ocf_log info "Oracle instance $ORACLE_SID already running"
return $OCF_SUCCESS
;;
"STARTED")
output=`dbasql dbmount`
;;
"MOUNTED")
: we proceed if mounted
;;
*) # status unknown
output=`dbasql dbstop dbstart_mount`
;;
esac
else
output="`dbasql dbstart_mount`"
# try to cleanup in case of
# ORA-01081: cannot start already-running ORACLE - shut it down first
if echo "$output" | grep ORA-01081 >/dev/null 2>&1; then
ora_cleanup
output=`dbasql dbstart_mount`
fi
fi
# oracle instance should be mounted.
status="`dbasql dbstat`"
case "$status" in
"MOUNTED")
;;
*)
: error!!
ocf_log error "Oracle $ORACLE_SID can not mount."
return $OCF_ERR_GENERIC
;;
esac
# It is examined whether mode is "online backup mode",
# and if it is true, makes clear the mode.
# Afterwards, DB is opened.
- if is_clear_backupmode_set && is_instance_in_backupmode; then
+ if is_clear_backupmode_set && is_instance_in_backup_mode; then
clear_backup_mode
fi
output=`dbasql dbopen`
if is_oracle_up && instance_live; then
: cool, we are up and running
ocf_log info "Oracle instance $ORACLE_SID started: $output"
return $OCF_SUCCESS
else
ocf_log err "Oracle instance $ORACLE_SID not started: $output"
return $OCF_ERR_GENERIC
fi
}
#
# oracle_stop: Stop the Oracle instance
#
oracle_stop() {
typeset status output ipc=""
if is_oracle_up; then
[ "$IPCRM" = "instance" ] && ipc=$(parseipc `dumpinstipc`)
output=`dbasql dbstop`
else
ocf_log info "Oracle instance $ORACLE_SID already stopped"
return $OCF_SUCCESS
fi
ora_kill # kill any processes left
if is_oracle_up; then
ocf_log err "Oracle instance $ORACLE_SID not stopped: $output"
return $OCF_ERR_GENERIC
else
ocf_log info "Oracle instance $ORACLE_SID stopped: $output"
sleep 1 # give em a chance to cleanup
ocf_log info "Cleaning up for $ORACLE_SID"
ora_cleanup "$ipc"
return $OCF_SUCCESS
fi
}
# kill the database processes (if any left)
# give them 30 secs to exit cleanly (6 times 5)
killprocs() {
typeset sig=$1
shift 1
kill -$sig $* >/dev/null 2>&1
}
ora_kill() {
killprocs TERM `eval $procs | awk '{print $1}'`
for i in 1 2 3 4 5; do
killprocs 0 `eval $procs | awk '{print $1}'` ||
return
sleep 5
done
killprocs KILL `eval $procs | awk '{print $1}'`
}
#
# oracle_monitor: Can the Oracle instance do anything useful?
#
oracle_monitor() {
if is_oracle_up && instance_live
then
#ocf_log info "Oracle instance $ORACLE_SID is alive"
return $OCF_SUCCESS
else
ocf_log info "Oracle instance $ORACLE_SID is down"
return $OCF_NOT_RUNNING
fi
}
#
# 'main' starts here...
#
if [ $# -ne 1 ]
then
usage
exit $OCF_ERR_ARGS
fi
# These operations don't require OCF instance parameters to be set
case "$1" in
meta-data) meta_data
exit $OCF_SUCCESS;;
usage) usage
exit $OCF_SUCCESS;;
methods) oracle_methods
exit $?;;
*);;
esac
clear_backupmode=${OCF_RESKEY_clear_backupmode:-"false"}
shutdown_method=${OCF_RESKEY_shutdown_method:-"checkpoint/abort"}
case "${shutdown_method}" in
"immediate") ;;
"checkpoint/abort") ;;
*) ocf_log error "unsupported shutdown_method, please read meta-data"
esac
if [ x = "x$OCF_RESKEY_sid" ]
then
ocf_log err "Please set OCF_RESKEY_sid to the Oracle SID !"
exit $OCF_ERR_ARGS
fi
ora_info "$OCF_RESKEY_sid" "$OCF_RESKEY_home" "$OCF_RESKEY_user"
LSB_STATUS_STOPPED=3
if ! testoraenv; then
ocf_log info "Oracle environment for SID $ORACLE_SID does not exist"
case "$1" in
stop) exit $OCF_SUCCESS;;
monitor) exit $OCF_NOT_RUNNING;;
status) exit $LSB_STATUS_STOPPED;;
*)
ocf_log err "Oracle environment for SID $ORACLE_SID broken"
exit $OCF_ERR_ARGS
;;
esac
fi
setoraenv # important: set the environment for the SID
procs="ps -e -o pid,args | grep -i \"[o]ra.*$ORACLE_SID\""
US=`id -u -n`
if [ $US != root -a $US != $ORACLE_OWNER ]
then
ocf_log err "$0 must be run as root or $ORACLE_OWNER"
exit $OCF_ERR_PERM
fi
if [ x = "x$OCF_RESKEY_ipcrm" ]
then
IPCRM="instance"
else
IPCRM="$OCF_RESKEY_ipcrm"
fi
# What kind of method was invoked?
case "$1" in
start) oracle_start
exit $?;;
stop) oracle_stop
exit $?;;
status) if is_oracle_up
then
echo Oracle instance $ORACLE_SID is running
exit $OCF_SUCCESS
else
echo Oracle instance $ORACLE_SID is stopped
exit $OCF_NOT_RUNNING
fi
;;
dumpinstipc)
is_oracle_up && parseipc `dumpinstipc`
exit $?;;
cleanup)
if [ "$IPCRM" = "instance" ]; then
ora_cleanup $(parseipc `dumpinstipc`)
else
ora_cleanup
fi
exit $?;;
monitor) oracle_monitor
exit $?;;
validate-all) # OCF_RESKEY_sid was already checked by testoraenv(),
# just exit successfully here.
exit $OCF_SUCCESS;;
*) oracle_methods
exit $OCF_ERR_UNIMPLEMENTED;;
esac
#
# vim:tabstop=4:shiftwidth=4:textwidth=0:wrapmargin=0

File Metadata

Mime Type
text/x-diff
Expires
Thu, Jun 5, 11:14 PM (4 h, 26 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1862595
Default Alt Text
(16 KB)

Event Timeline