#!/bin/ksh
## IBM_PROLOG_BEGIN_TAG
## This is an automatically generated prolog.
##
##
##
## Licensed Materials - Property of IBM
##
## (C) COPYRIGHT International Business Machines Corp. 2009,2022
## All Rights Reserved
##
## US Government Users Restricted Rights - Use, duplication or
## disclosure restricted by GSA ADP Schedule Contract with IBM Corp.
##
## IBM_PROLOG_END_TAG
#
# sccsid = "@(#)53 1.85.1.1 src/rsct/cfg_access/ctcaactrl.sh, cfg.access, rsct_rady, radys004a 9/18/21 17:18:37"
#
# usage:
# ctcaactrl PRE_START validate this node can be added to a CAA cluster
# ctcaactrl [ -a | START | POST_START ] prepare/start RSCT CAA env
# (subsys etc)
# ctcaactrl [STOP|PRE_STOP] stop/remove RSCT CAA env (subsys etc)
# ctcaactrl PRE_STOP validate this node can be removed from a CAA
# cluster
# ctcaactrl { -O ADD_NODE|RM_NODE|STOP_NODE|START_NODE|JOIN_NODE|COMMIT_LVL|
# ADD_SITE|MOD_SITE|RM_SITE|MOD_TUNE }
# { -P CHECK | PRE | POST | UNDO_PRE }
# -T <trans id> -c <cluster id> -o <other targets>
# -t <target> -s <site> [-v variables]
# [-m] [-g] [-i]
#
# Soon-to-be-obsolete
# ctcaactrl -d StopAndDelete RSCT CAA env
# ctcaactrl -s start RSCT CAA subsystems
# ctcaactrl -k stop HA subsystems
# ctcaactrl -r refresh HA subsystems
#
# Set path to known value
PATH=/opt/rsct/install/bin:/opt/rsct/bin:/usr/bin:/bin:/usr/sbin:/sbin
export PATH
RSCTBIN=/opt/rsct/bin
CTROOT=/var/ct
# make sure ODMDIR to point /etc/objrepos
ODMDIR=/etc/objrepos
HA_CCAL_DIR=/var/ct/
HA_CCAL_LOG=${HA_CCAL_DIR}/caa.ctcaactrl.log
HA_CCAL_LOG_MAXSIZE_KB=500 # 500 KB
export HA_CCAL_LOG
# This is the file which talks to ConfigRM
# Don't change the filename
CAACMDCTRLFILE=/var/ct/cfg/caa-control.cmd
CAACOMMITCTRLFILE=/var/ct/cfg/caa-commit.cmd
#
# presence of this file means simulation of having
# powerha pre 721 version.
#
POWERHA_PRE_721="/var/ct/cfg/powerha_pre_721"
basecmd=$(basename $0)
#########################################################################
# Get the current Locale and determine whether English output is also needed
# Output: LOCALE
# PRINT_ENG_MSG
#########################################################################
get_current_locale()
{
# Check if current locale is a English
if [[ -f /usr/bin/locale ]]; then
LOCALE=`/usr/bin/locale`
else
LOCALE=`/bin/locale`
fi
# /usr/bin/locale may produce LC_MESSAGES=en_US or LC_MESSAGES="en_US".
# We use two sed "s" commands, the first one gets rid of LC_MESSAGES=,
# the second one removes the two surrounding '"' if they are there.
LOCALE=`echo "$LOCALE" | grep LC_MESSAGES | sed "s/LC_MESSAGES=//;s/\"//g"`
# Do not print English messages if the local language is English.
# C and POSIX are special locales. Treat them as en_US.
if [[ "$LOCALE" = "C" || "$LOCALE" = "POSIX" ]]; then
PRINT_ENG_MSG="NO"
else
# A locale may look like en_US or en_US.ISO8859-1. Use cut command
# to remove the language encoding.
LOCALE=`echo "$LOCALE" | cut -f 1 -d "."`
if [[ "$LOCALE" = "en_US" || "$LOCALE" = "En_US" || "$LOCALE" = "en_GB" || "$LOCALE" = "En_GB" ]]
then
# Local language is English.
PRINT_ENG_MSG="NO"
else
# Local language is not English (including "$LOCALE"=""),
# which should not happen in normal cases)
PRINT_ENG_MSG="YES"
fi
fi
}
#########################################################################
# #
# Function: print_message #
# Description: wrapper for message printing. All messages go to stderr. #
# For now, all messages are printed in both English and the current #
# language unless current language is en_US, C, or POSIX. #
# #
# This subroutine makes it easy to print messages in either English#
# or current language or both. Also, it ensure all messages will be #
# printed to the same stream. #
#########################################################################
#Usage create_logfile_with_permission "logfile" "mode"
function create_logfile_with_permission
{
typeset logfile=$1
typeset permission=$2
# Create a log file and set the permission if defined
if [ -n "$logfile" ]; then
touch $logfile
[ -z "$permission" ] || chmod $permission $logfile
fi
}
#
function print_dbgmsg
{
if [[ -n $HA_CCAL_LOG ]]; then
print `date` ${basecmd} "($$)" : "$*" >> $HA_CCAL_LOG
fi
}
print_message()
{
MSGCMD2="hadspmsg cfgaccess ha_com.cat"
print -u2 "`${MSGCMD2} $*`"
print_dbgmsg "`${MSGCMD2} $*`"
}
#########################################################################
function print_process_tree
{
typeset OSNAME=$(uname -s)
if [[ "$OSNAME" = "AIX" ]]
then
PSTREE_MSG=$(proctree -T -p $$ 2> /dev/null)
[ -z "${PSTREE_MSG}" ] || print_dbgmsg "proctree:\n${PSTREE_MSG}"
else
PSTREE_MSG=$(pstree -sp $$ 2> /dev/null)
[ -z "${PSTREE_MSG}" ] || print_dbgmsg "\n${PSTREE_MSG}"
fi
}
#########################################################################
#
# backup the ccal log file if its size is big
function backup_ccal_log_file
{
if [[ -z $HA_CCAL_LOG ]]; then
return
fi
if [[ -n $HA_CCAL_DIR && ! -d $HA_CCAL_DIR ]]; then
print_dbgmsg "Create ccal dir $HA_CCAL_DIR"
mkdir -p ${HA_CCAL_DIR}
fi
if [[ ! -f $HA_CCAL_LOG ]]; then
# LOG file does not exist. Nothing to backup
print_dbgmsg "No log file exist. No backup is needed"
return 1;
fi
# Check size of "HA_CCAL_LOG"
if [[ -z $HA_CCAL_LOG_MAXSIZE_KB ]]; then
HA_CCAL_LOG_MAXSIZE_KB=500
fi
set -A lsout $(ls -s ${HA_CCAL_LOG})
if (( ${lsout[0]} > $HA_CCAL_LOG_MAXSIZE_KB ))
then
# Exceed
print_dbgmsg "Log file $HA_CCAL_LOG exceeds the size $HA_CCAL_LOG_MAXSIZE_KB Kb. Backup it"
mv -f ${HA_CCAL_LOG} ${HA_CCAL_LOG}.bak
fi
}
#########################################################################
# return RPD cluster names
function GetRPDClusters
{
clusters=""
while read cid cname ctype
do
if [[ "$cname" != "IW" ]]; then
if [[ -z "$ctype" || "$ctype" != "CAA" ]]; then
clusters="$clusters $cname"
fi
fi
done < /var/ct/cfg/clusters
# print outputs
echo $clusters
}
#
# Get CAA RPD clusters.
#
function Get_CAA_RPD_Clusters
{
clusters=""
while read cid cname ctype
do
if [[ "$cname" != "IW" ]]; then
if [[ "$ctype" == "CAA" ]];
then
clusters="$clusters $cname"
fi
fi
done < /var/ct/cfg/clusters
echo $clusters
}
#
# return (1 + 'OR'ed reason-codes) if an error happens
# reason code for cthagsctrl = 0x01 (1)
# input: (gs_rc)
function exit_script
{
rc=$gs_rc
exit $rc
}
################################################################################
# #
# exec_cmd_timeout - executes a command with a given timeout #
# #
# The function returns when the command returns or when the timeout expires, #
# whichever comes first. #
# #
# #
# Arguments: #
# #
# Timeout value in secs #
# Target program and its arguments #
# #
# #
################################################################################
function eval_cmd
{
print_dbgmsg "Background running: $*"
eval $* # run command in the background
rc=$?
print_dbgmsg "Cmd $* exited with rc=$rc"
return $rc
}
function exec_cmd_timeout
{
typeset timeOut=$1
shift # remove the first arg
print_dbgmsg "Invoking $*"
eval_cmd $* & # run command in the background
pid=$!
print_dbgmsg "Background pid=$pid"
typeset isTimeout=1
typeset sleepTime=2
typeset waitTime=0
while [[ $waitTime -lt $timeOut ]]
do
kill -s 0 $pid >/dev/null 2>&1
if [ $? -ne 0 ] # process no longer running
then
isTimeout=0
break;
fi
sleep $sleepTime
(( waitTime = $waitTime + $sleepTime))
done
if [[ $isTimeout == 1 ]]
then
print_dbgmsg "Command $1 ($pid) timed out after $waitTime secs"
else
print_dbgmsg "Command $1 ($pid) successful"
fi
}
#
# Tell ConfigRM and wait for the completion of operation (with timeout)
# Input: Op = AddAndStart | StopAndDelete
#
function tell_ConfigRM_resync
{
# Operation
typeset Op=$1
# Make sure ConfigRM is auto-startable
touch /var/ct/cfg/ConfigRM.autostart
# Just in case RMC is not running
rmcctrl -A 2> /dev/null >/dev/null
# Check whether ConfigRM is running
state=`LC_ALL=C /usr/bin/lssrc -s IBM.ConfigRM 2> /dev/null | LC_ALL=C /bin/sed -e"1d;s/.* \([a-zA-Z0-9]*\)$/\1/"`
if [[ -z $state ]]
then
# not defined in SRC ... add it here (instead going thru RMC)
print_dbgmsg "Refresh RMC to start ConfigRM"
/usr/bin/refresh -s ctrmc
elif [[ $state != "active" ]]
then
# Make sure ConfigRM running just in case ConfigRM was stopped
print_dbgmsg "start ConfigRM subsystem"
startsrc -s IBM.ConfigRM 2> /dev/null >/dev/null
fi
# refresh ConfigRM
print_dbgmsg "Tell ConfigRM to resync CAA...."
exec_cmd_timeout 15 /usr/bin/refresh -s IBM.ConfigRM
frc=$?
print_dbgmsg "refresh ConfigRM is done, rc=$frc"
}
################################################################################
# Test whether HAGS is still running for upto <sometime> secs #
# If it is too long, forcefully stopsrc -cthags #
################################################################################
function waitfor_hags_termination_or_stopsrc
{
#typeset timeOut=70 # 70secs (a little more than 60s [internal timeout])
typeset timeOut=15 # 70secs cause whole rmcluster operation delay. So, reduece to 15 seconds for now
# get 'cthags pid'
PARSE_SRC_PIDS="s/^ *[^ ]* *[^ ]* *\([0-9]*\) *active/\1/p"
TGT_SUBSYS="cthags"
pid=$(LC_ALL=C /usr/bin/lssrc -s "$TGT_SUBSYS" 2> /dev/null | grep "active" | sed -n -e "$PARSE_SRC_PIDS")
if [[ -n $pid ]]
then
print_dbgmsg "Wait for HAGS (pid=$pid) to stop"
typeset isTimeout=1
typeset sleepTime=2
typeset waitTime=0
while [[ $waitTime -lt $timeOut ]]
do
kill -s 0 $pid >/dev/null 2>&1
if [ $? -ne 0 ] # process no longer running
then
isTimeout=0
break;
fi
sleep $sleepTime
(( waitTime = $waitTime + $sleepTime))
done
if [[ $isTimeout == 1 ]]
then
print_dbgmsg "HAGS (pid=$pid) is still running. Stop it now"
stopsrc -cs cthags 2> /dev/null
fi
#check it again
kill -s 0 $pid >/dev/null 2>&1
if [ $? -ne 0 ] # process no longer running
then
print_dbgmsg "HAGS is now stopped"
fi
fi
}
################################################################################
# Check for cluster Offline status. If it is still Online wait for it until #
# it becomes offline. Wait for a maximum of 30 sec #
################################################################################
function waitfor_ConfigRM_Offline_state
{
print_dbgmsg "waitfor_ConfigRM_Offline_state "
CLSTNAME=`/opt/rsct/bin/ct_clusterinfo -c`
typeset sleepTime=1
typeset IWClusterName="IW"
typeset waitTime=0
typeset timeOut=30
while [[ $waitTime -lt $timeOut ]]
do
if [ $CLSTNAME == $IWClusterName ]
then
break;
else
sleep $sleepTime
fi
(( waitTime = $waitTime + $sleepTime ))
CLSTNAME=`/opt/rsct/bin/ct_clusterinfo -c`
done
}
################################################################################
# #
# parse_arguments: parse input arguments #
# #
# Expected input format is #
# #
# <phase> <cluster id> <other target ids> <target id> <target indicator> #
# #
# <phase>: one of {CHECK| PRE | POST | UNDO_PRE} #
# <cluster id>: cluster UUID #
# <other target ids>: #
# comma separated list of other nodes to be targeted #
# (including target) #
# NO_OTHER_TARGETS if none #
# <target id>: #
# node add: Node UUID, canonical hostname, shortid #
# others: Node UUID #
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# #
# The following variables are filled in by parse_arguments #
# #
# PARSED_PHASE: script phase: one of the strings above #
# PARSED_CLUSTER_UUID: cluster UUID #
# PARSED_TARGET_UUID: node UUID of the target #
# PARSED_TARGET_HOSTNAME: hostname of the target, if available #
# (empty otherwise) #
# PARSED_TARGET_SHORT_ID: short node ID of the target, if available #
# ("INVAL_SHORT_ID" otherwise) #
# PARSED_AM_I_THE_TARGET: "yes" if local node is the target, #
# "no" otherwise #
# #
# #
# parse_arguments will cause the script to exit if the syntax is incorrect. #
# #
# #
################################################################################
parse_arguments() {
# check if 'phase' string is valid
PARSED_PHASE=$1
case $PARSED_PHASE in
CHECK | PRE | POST | UNDO_PRE)
;;
*) # invalid phase specification
print_message IMSG_ctcaactrl_Usage $basecmd
print_message EMSG100 $basecmd $PARSED_PHASE
exit 1
;;
esac
# Cluster ID
PARSED_CLUSTER_UUID=$2
# Target ID
PARSED_TARGET_SHORT_ID="INVAL_SHORT_ID"
PARSED_TARGET_HOSTNAME=""
typeset target_id_info=$4
if echo $target_id_info | grep "," > /dev/null
then
# Format assumed to be: Node UUID, canonical hostname, shortid
PARSED_TARGET_UUID=${target_id_info%%,*}
rest=${target_id_info#*,}
# echo "PARSED_TARGET_UUID=$PARSED_TARGET_UUID"
PARSED_TARGET_HOSTNAME=${rest%%,*}
rest=${rest#*,}
# echo "PARSED_TARGET_HOSTNAME=$PARSED_TARGET_HOSTNAME"
PARSED_TARGET_SHORT_ID=${rest%%,*}
# echo "PARSED_TARGET_SHORT_ID=$PARSED_TARGET_SHORT_ID"
else
# Format assumed to be: Node UUID
PARSED_TARGET_UUID=$target_id_info
fi
# Target indicator
typeset target_indicator=$5
case $target_indicator in
TARGET_NODE ) PARSED_AM_I_THE_TARGET="yes"
;;
*) PARSED_AM_I_THE_TARGET="no"
esac
# echo "PARSED_PHASE = $PARSED_PHASE"
# echo "PARSED_CLUSTER_UUID = $PARSED_CLUSTER_UUID"
# echo "PARSED_TARGET_UUID = $PARSED_TARGET_UUID"
# echo "PARSED_AM_I_THE_TARGET = $PARSED_AM_I_THE_TARGET"
}
################################################################################
# #
# check_for_legacy_cluster - checks whether there is a legacy (pre-CAA) #
# present on the local node. This function should be run on the CHECK #
# phase when the node is about to be added to the cluster or join to it. #
# #
# #
# The function will output an error message and exit 1 if a legacy cluster #
# is present. The function will NOT exit if there is no legacy cluster; #
# It's up to the caller to exit 0. #
# #
# Arguments: None #
# #
# Return value: 0: indicates that no legacy cluster has been found #
# #
################################################################################
check_for_legacy_cluster() {
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
if [[ "$caa_is_disabled_for_migration" == "true" ]]
then
#
# If "disabled for migration", the PowerHA/RPD domain is being
# migrated in coordination with RSCT, so having a previously
# defined domain is not an error.
#
print_dbgmsg "SUCESS: RSCT-to-CAA is in progress"
return 0
fi
clusters=$(Get_CAA_RPD_Clusters)
if [[ -n "$clusters" ]]
then
#
# The existence of a CAA RPD cluster means we have a previously
# defined domain, which is not an error.
#
print_dbgmsg "SUCCESS: CAA RPD cluster already exists: $clusters"
return 0
fi
#
# See if a PowerHA cluster is configured which includes this node.
# If so, the odm entry for the HACMPcluster object class will have
# a non-zero id attribute value, and a non-empty name attribute value.
#
power_ha=`/usr/bin/odmget HACMPcluster 2>/dev/null`
rc=$?
if [[ 0 == $rc ]]
then
#
# grep for the id.
#
power_ha_id=`echo "$power_ha" | /usr/bin/grep -w "id"`
rc=$?
if [[ 0 == $rc ]]
then
#
# grep for the id.
#
power_ha_name=`echo "$power_ha" | /usr/bin/grep -w "name"`
rc=$?
if [[ 0 == $rc ]]
then
#
# The id is an integer. Remove the "large left" pattern up
# through a space to leave only the integer itself.
#
power_ha_id=${power_ha_id##* }
#
# The name is a string. Remove the "small left" pattern up
# through the first double quote, then the "small right"
# pattern from the second double quote on, leaving only the
# string itself.
#
power_ha_name=${power_ha_name#*\"}
power_ha_name=${power_ha_name%\"*}
if [[ "0" == "$power_ha_id" ]]
then
print_dbgmsg "SUCCESS: power_ha_id is 0"
return 0
fi
if [ -z "$power_ha_name" ]
then
print_dbgmsg "SUCCESS: power_ha is not defined"
return 0
fi
#
# The id and the name are legitimate. Consider PowerHA
# configured.
#
# If PowerHA cluster version is at least 7.1 then
# this cluster creation is being made on behalf of a
# CAA-aware instance of PowerHA. Allow cluster.
PowerHA7_version=12
cluster_version=`/usr/es/sbin/cluster/utilities/clmixver`
rc=$?
if [[ ($rc == 0) || ($rc == 1) ]]
then
if (( $cluster_version >= $PowerHA7_version ))
then
print_dbgmsg "PowerHA 7.1 or later present"
return 0
else
print_dbgmsg "PowerHA pre-7.1 present: cluster version $cluster_version"
fi
else
# A problem with clmixver: should not happen, but it does,
# we assume that a propoer PowerHA cluster is not present
print_dbgmsg "clmixver error: $rc. Assume cluster not present"
return 0
fi
print_message EMSG111 $basecmd $power_ha_id $power_ha_name
exit 1
fi
fi
fi
#
# See if an RPD is configured which includes this node. If so,
# there will be more than just an IW entry in /var/ct/cfg/clusters.
#
clusters=$(GetRPDClusters)
if [[ "" != "$clusters" ]]
then
#
# An RPD is configured on this node.
#
migration_realm=`ct_caa_get_migration_realm`
print_dbgmsg "migration_realm=$migration_realm"
if [[ "$caa_is_disabled_for_migration" == "false" || "$migration_realm" != "RPD" ]]
then
# RPD is not in migration.
print_dbgmsg "Fail: RPDs: $clusters are already defined"
print_message EMSG112 $basecmd
exit 1
fi
fi
return 0
}
################################################################################
# #
# check_for_inconsistent_cluster_ID - if CAA is up, check if the cluster ID #
# matches the one sent as parameter to the script. #
# #
# The function will output an error message and exit 1 if the cluster IDs #
# don't match. The function will NOT exit if the cluster IDs match. #
# It's up to the caller to exit 0. #
# #
# Depends on the following variables already being set: #
# * REALM: if 'CAA', indicates CAA is configured on the node #
# (obtained via ct_clusterinfo -r) #
# * CAA_UUID: if CAA is configured, its cluster ID #
# * PARSED_CLUSTER_UUID: cluster ID retrieved from the parameters to the #
# the script #
# #
# Arguments: None #
# #
# Return value: 0: indicates that either CAA is not up or the cluster IDs #
# did match #
# #
################################################################################
check_for_inconsistent_cluster_ID() {
if [[ "$REALM" == "CAA" ]]
then
# This indicates that CAA is configured on the node
# just in case: if our cluster ID is invalid then assume CAA is not
# really configured
if [[ -z "$CAA_UUID" || "$CAA_UUID" == "00000000-0000-0000-0000-000000000000" ]]
then
print_dbgmsg "Local Cluster ID ($CAA_UUID) is invalid"
return 0
fi
if [[ -z "$PARSED_CLUSTER_UUID" ]]
then
print_dbgmsg "Cluster ID received as paramter not set"
return 0
fi
if [[ "$CAA_UUID" != "$PARSED_CLUSTER_UUID" ]]
then
print_dbgmsg "Cluster ID ($CAA_UUID) does not match ID passed as parameter ($PARSED_CLUSTER_UUID)"
## ==> need new message after EMSG112 (?)
exit 1
else
print_dbgmsg "Cluster ID check succeeded"
fi
fi
return 0
}
################################################################################
# #
# add_and_start_local - creates the RPD cluster tree, if not present, #
# create the GS subsystem (if not already created), and instruct ConfigRM #
# to go online. The actions are intended to be idempotent: it should cause #
# no harm to invoke this function when some of the actions to bring the #
# node online have already previously been taken. #
# #
# Assumes that the following variables are already set up: #
# * REALM #
# * CLUSTER_ID #
# * CLUSTER_NAME #
# * NODE_NUMBER #
# #
# This function will exit 1 (error) if some fatal error is encountered. #
# #
# Arguments: None #
# #
# Return value: None #
# #
################################################################################
add_and_start_local() {
#
# Start with assumptions about what this block should accomplish.
#
print_dbgmsg "Processing AddAndStart ... "
if [[ ! -e "$CT_CAA_STATE_DEFINED_FILE" ]]
then
/usr/bin/touch $CT_CAA_STATE_DEFINED_FILE
print_dbgmsg "CAA defined state file created"
fi
if [[ -e "$CT_CAA_STATE_STOPPED_FILE" ]]
then
rm -f $CT_CAA_STATE_STOPPED_FILE
print_dbgmsg "CAA stopped state file removed"
fi
# Important: Write CAA cmd file so that ConfigRM can understand the operation
echo "START $CLUSTER_NAME $CLUSTER_ID" > $CAACMDCTRLFILE
if [[ ! -z $REALM && $REALM != "CAA" ]]; then
# non CAA environment
print_message EMSG110 $basecmd
exit 1
fi
#
create_group_services_subsystem=true
start_group_services_subsystem=true
refresh_ConfigRM=true
#
# Take migration into account. Start by discovering whether operating in a
# CAA environment should be disabled because migration is taking place.
#
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
if [[ "$caa_is_disabled_for_migration" == "true" ]]
then
start_group_services_subsystem=false
refresh_ConfigRM=false
migration_realm=`ct_caa_get_migration_realm`
print_dbgmsg "migration_realm=$migration_realm"
if [[ "$migration_realm" == "RPD" ]]
then
#
# Don't need to create the Group Services subsystem in an RPD
# environment, because it should already have been created in
# that environment.
#
create_group_services_subsystem=false
fi
else # not disabled for migration (but migration may still be under way)
# ctcaactrl may have been invoked via ct_caa_gs_migrate_ctrl.
# If the GS daemon undegoing migration is still running then
# it should not be started. Especially critical in PowerHA migration,
# where the subsystem being running is grpsvcs, not cthags.
# In addition, if the migrated GS daemon is still running, the
# cthags subsystem will have been already created, at the time
# ctcaactrl POST_START was invoked upon CAA cluster creation. So no
# need to create GS subsystems again
migrated_GS_daemon_still_running=`ct_caa_migrated_GS_still_running`
if [[ $migrated_GS_daemon_still_running == "true" ]]
then
print_dbgmsg "GS daemon which underwent migration is still running: do not start cthags"
create_group_services_subsystem=false
start_group_services_subsystem=false
fi
fi
print_dbgmsg "refresh_ConfigRM=$refresh_ConfigRM"
print_dbgmsg "start_group_services_subsystem=$start_group_services_subsystem"
print_dbgmsg "create_group_services_subsystem=$create_group_services_subsystem"
#
# Setup the cluster info & dir.
#
if [[ ! -d /var/ct/${CLUSTER_ID} || ! -d /var/ct/${CLUSTER_NAME} ]]
then
# cluster does not exist...then create it
print_dbgmsg "Configure CAA ${CLUSTER_NAME} ${CLUSTER_ID} ${NODE_NUMBER}"
cfghact -c ${CLUSTER_NAME} -n ${NODE_NUMBER} -y ${CLUSTER_ID} -t CAA -f
else
print_dbgmsg "/var/ct/${CLUSTER_ID} is already found...but make sure to have the correct ones"
# 'cfghact' will recreate if nodedef.cfg is empty
cfghact -c ${CLUSTER_NAME} -n ${NODE_NUMBER} -y ${CLUSTER_ID} -t CAA -f
fi
if [[ "$create_group_services_subsystem" == "true" ]]
then
print_dbgmsg "Calling $RSCTBIN/cthagsctrl -a ..."
ctrl_result=`$RSCTBIN/cthagsctrl -a 2>&1`
gs_rc=$?
if (( $gs_rc != 0 ))
then
print_dbgmsg "cthagsctrl -a exited with rc $gs_rc:\n$ctrl_result"
fi
fi
if [[ "$start_group_services_subsystem" == "true" ]]
then
print_dbgmsg "Calling $RSCTBIN/cthagsctrl -s ..."
ctrl_result=`$RSCTBIN/cthagsctrl -s 2>&1`
gs_rc=$?
if (( $gs_rc != 0 ))
then
print_dbgmsg "cthagsctrl -s exited with rc $gs_rc:\n$ctrl_result"
fi
fi
if [[ "$refresh_ConfigRM" == "true" ]]
then
tell_ConfigRM_resync $Op
else
# Make ConfigRM auto-startable
touch /var/ct/cfg/ConfigRM.autostart
fi
print_dbgmsg "Initialization done rc=$gs_rc"
}
################################################################################
# #
# stop_and_delete_local_pre - stops the RPD on the local node and remove it. #
# If ConfigRM is running (it should) then ConfigRM is responsible for #
# stopping GS. #
# #
# #
# Assumes that the following variables are already set up: #
# * REALM #
# * CLUSTER_ID #
# * CLUSTER_NAME #
# #
# This function will exit 1 (error) if some fatal error is encountered. #
# #
# Arguments: None #
# #
# Return value: 0: success #
# 1: failure which should be reflected in the ctcaactrl exit val #
# #
################################################################################
stop_and_delete_local_pre() {
# Should avoid returning non-zero, since that will cause the script to
# return non-zero, resulting in the entire cluster operation (likely the
# node or cluster being removed) to fail.
print_dbgmsg "Processing stop_and_delete_local_pre ... "
# Write CMD file (STOP)
echo "STOP $CLUSTER_NAME $CLUSTER_ID" > $CAACMDCTRLFILE
echo "STOP-COMMIT $CLUSTER_NAME $CLUSTER_ID" > $CAACOMMITCTRLFILE
if [[ $REALM != "CAA" && $CAAForceOpt != "FORCE" ]]; then
# non CAA environment and Not force
print_message EMSG110 $basecmd
print_dbgmsg "Assume stop_and_delete_local as successful even if REALM is not CAA"
return 0
fi
# If ConfigRM is running
state=`LC_ALL=C /usr/bin/lssrc -s IBM.ConfigRM 2> /dev/null | LC_ALL=C /bin/sed -e"1d;s/.* \([a-zA-Z0-9]*\)$/\1/"`;
if [[ -z $state || $state != "active" ]]
then
# ConfigRM is not running, remove CAA cluster structure
print_dbgmsg "Unconfig CAA cluster..."
# put "IW" to default_cluster so that ConfigRM won't react to online
IWClusterID=$(grep -w "IW" /var/ct/cfg/clusters | awk '{print $1}')
if [[ -n ${IWClusterID} ]]; then
print "${IWClusterID}" > /var/ct/cfg/default_cluster
fi
# Stop HAGS if ConfigRM is not running
cthagsctrl -k
fi
# Tell ConfigRM to resync
tell_ConfigRM_resync $Op
# Wait a few seconds so that ConfigRM can process the active RMC requests (e.g. rmrpdomain)
# This will wait up to 60secs
waitfor_hags_termination_or_stopsrc
# Wait a few seconds so that ConfigRM finishes up Offline processing.
# This will wait up to 30 secs
waitfor_ConfigRM_Offline_state
print_dbgmsg "Processing stop_and_delete_local_pre Done. "
}
################################################################################
# #
# stop_local - stops the RPD on the local node, which includes stopping #
# the GS daemon. ConfigRM is responsible for stopping GS #
# #
# #
# Assumes that the following variables are already set up: #
# * REALM #
# * CLUSTER_ID #
# * CLUSTER_NAME #
# #
# This function will exit 1 (error) if some fatal error is encountered. #
# #
# Arguments: None #
# #
# Return value: 0: success #
# 1: failure which should be reflected in the ctcaactrl exit val #
# #
################################################################################
stop_local() {
# Should avoid returning non-zero, since that will cause the script to
# return non-zero, resulting in the entire cluster operation (likely the
# node or cluster being removed) to fail.
print_dbgmsg "Processing stop_local ... "
# Write CMD file (OFFLINE) & NONE for CommitLvl
echo "OFFLINE $CLUSTER_NAME $CLUSTER_ID" > $CAACMDCTRLFILE
echo "NONE $CLUSTER_NAME $CLUSTER_ID" > $CAACOMMITCTRLFILE
if [[ $REALM != "CAA" && $CAAForceOpt != "FORCE" ]]; then
# non CAA environment and Not force
print_message EMSG110 $basecmd
print_dbgmsg "Assume stop_local as successful even if REALM is not CAA"
return 0
fi
# If ConfigRM is running
state=`LC_ALL=C /usr/bin/lssrc -s IBM.ConfigRM 2> /dev/null | LC_ALL=C /bin/sed -e"1d;s/.* \([a-zA-Z0-9]*\)$/\1/"`;
if [[ -z $state || $state != "active" ]]
then
# put "IW" to default_cluster so that ConfigRM won't react to online
IWClusterID=$(grep -w "IW" /var/ct/cfg/clusters | awk '{print $1}')
if [[ -n ${IWClusterID} ]]; then
print "${IWClusterID}" > /var/ct/cfg/default_cluster
fi
# Stop HAGS if ConfigRM is not running
cthagsctrl -k
fi
# Tell ConfigRM to resync
tell_ConfigRM_resync $Op
# Wait a few seconds so that ConfigRM can process the active RMC requests
# (e.g. rmrpdomain)
# This will wait up to 60 secs
waitfor_hags_termination_or_stopsrc
# Wait a few seconds so that ConfigRM finishes up Offline processing.
# This will wait up to 30 secs
waitfor_ConfigRM_Offline_state
print_dbgmsg "Processing stop_local Done."
}
################################################################################
# #
# stop_and_delete_local_post - The RPD was supposed to have been deleted #
# via stop_and_delete_local() (run in the "PRE" phase), but this function #
# is invoked via "POST" just as a "catch all", to ensure that, for example, #
# GS has been stopped. #
# #
# #
# Arguments: None #
# #
# Return value: None #
# #
################################################################################
stop_and_delete_local_post() {
# show the current states
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
migration_realm=`ct_caa_get_migration_realm`
print_dbgmsg "migration_realm=$migration_realm"
#
# See if an RPD is configured which includes this node. If so,
# there will be more than just an IW entry in /var/ct/cfg/clusters.
#
clusters=$(GetRPDClusters)
if [[ "" != "$clusters" ]]
then
# An RPD is configured on this node.
print_dbgmsg "RPDs: $clusters are already defined. HAGS subsystem will not be stopped."
else
print_dbgmsg "Stopping cthags if it is still up"
# Final cleanup steps in case StopAndDelete did not get to complete them
cthagsctrl -k
fi
# Tell ConfigRM to resync
tell_ConfigRM_resync $Op
# Make sure HAGS to be stopped
waitfor_hags_termination_or_stopsrc
waitfor_ConfigRM_Offline_state
print_dbgmsg "Processing stop_and_delete_local_post Done."
}
commit_lvl_check() {
#
# Starting from 0DY support onwards CAA will generate new style of site
# events different from linked cluster. We need to reject the commit
# level operation if PowerHA SystemMirror is having older version (< 17)
# so that CAA can generate LC style of site events.
#
#
#
# if file POWERHA_PRE_721 is available which means to simulate of having
# PowerHA with pre 7.2.1 version
#
if [[ -e "$POWERHA_PRE_721" ]]
then
print_dbgmsg "Simulated of having PowerHA pre-7.2.1, so rejecting commit level operation"
exit 1
fi
PowerHA721_version=17
PowerHA_cluster_version=`/usr/sbin/rsct/install/bin/ct_is_haes_present`
rc=$?
if [[ $rc == 0 ]]
then
if (( $PowerHA_cluster_version < $PowerHA721_version && $CAP_SPLT_MRG_FOUND == 1 ))
then
print_dbgmsg "PowerHA pre-7.2.1 (PowerHA version: $PowerHA_cluster_version) is present, so rejecting commit level operation."
exit 1
fi
fi
print_dbgmsg "Processing commit_lvl_check Done."
}
################################################################################
# #
# commit_lvl_local - The CAA cluster has just been updated to a newer level #
# and now RSCT must in turn update the mapped active version level to match. #
# #
# Arguments: None #
# #
# Return value: None #
# #
################################################################################
commit_lvl_local() {
# Write to CAA Cmd file
echo "START-COMMIT $CLUSTER_NAME $CLUSTER_ID" > $CAACOMMITCTRLFILE
if [[ ! -z $REALM && $REALM != "CAA" ]]; then
# non CAA environment
print_message EMSG110 $basecmd
exit 1
fi
#
# Take migration into account. Do not continue with commit if the
# migration is taking place.
#
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
if [[ "$caa_is_disabled_for_migration" != "true" ]]
then
# Tell ConfigRM to resync
tell_ConfigRM_resync $Op
fi
print_dbgmsg "Processing commit_lvl_local Done."
}
################################################################################
# #
# process_AddNode: process ADD_NODE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST | UNDO_PRE} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Node UUID, canonical hostname, shortid #
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_AddNode() {
# echo "process_AddNode: called"
# if the parameters are passed via options then parsing should have
# been done in the main flow
if [[ -z "$PARSED_PHASE" ]]
then
parse_arguments $*
fi
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID, PARSED_TARGET_SHORT_ID,
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
# ADD_NODE CHECK Processing
print_dbgmsg "ADD_NODE CHECK Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
# check for legacy (non-CAA) clusters. Function below
# should cause the script to exit with 1 if any is found
check_for_legacy_cluster
# check for inconsistent cluster IDs between what local CAA
# reports and what has been passed as parameter
check_for_inconsistent_cluster_ID
else # Not the local node
# That's where checks can be added for the remote node being
# added. For example:
# * Is the new node already in the cluster? (if so, reject)
# * Is the new node's shorthand node ID too large?
# * Do we have too many nodes in the cluster already?
print_dbgmsg "Not the target"
fi
;;
PRE )
# ADD_NODE PRE Processing
print_dbgmsg "ADD_NODE PRE Processing (nothing to do)"
# No real work to do in ADD_NODE PRE (just make sure we exit 0,
# so that CAA will not cancel the operation)
;;
POST )
# ADD_NODE POST Processing
print_dbgmsg "ADD_NODE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node (nothing to do)"
# Making the RPD online is now being done in JOIN_NODE,
# and not in ADD_NODE
# # Create RPD and go online, which includes creating the
# # GS subsystem and starting it
# add_and_start_local
else
print_dbgmsg "Not the target"
fi
;;
UNDO_PRE )
# ADD_NODE UNDO_PRE Processing
print_dbgmsg "ADD_NODE UNDO_PRE Processing (nothing to do)"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# Nothing to do
fi
;;
esac
exit 0
}
################################################################################
# #
# process_JoinNode: process JOIN_NODE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST | UNDO_PRE} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Node UUID #
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_JoinNode() {
# echo "process_JoinNode: called"
# if the parameters are passed via options then parsing should have
# been done in the main flow
if [[ -z "$PARSED_PHASE" ]]
then
parse_arguments $*
fi
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID, PARSED_TARGET_SHORT_ID,
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
# JOIN_NODE CHECK Processing
# CHECK processing should be essentially identical for JOIN_NODE
# and ADD_NODE. We cannot take too much from granted in a
# JOIN_NODE, since (1) lots could have happened to the node
# while it was not active in the cluster and (2) the node might
# have been added to the cluster while it was down.
print_dbgmsg "JOIN_NODE CHECK Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
# check for legacy (non-CAA) clusters. Function below
# should cause the script to exit with 1 if any is found
check_for_legacy_cluster
# check for inconsistent cluster IDs between what local CAA
# reports and what has been passed as parameter
check_for_inconsistent_cluster_ID
else # Not the local node
# That's where checks can be added for the remote node being
# added. For example:
# * Is the new node already in the cluster? (if so, reject)
# * Is the new node's shorthand node ID too large?
# * Do we have too many nodes in the cluster already?
print_dbgmsg "Not the target"
fi
;;
PRE )
# JOIN_NODE PRE Processing
print_dbgmsg "JOIN_NODE PRE Processing (nothing to do)"
# No real work to do in JOIN_NODE PRE (just make sure we exit 0,
# so that CAA will not cancel the operation)
;;
POST )
# JOIN_NODE POST Processing
print_dbgmsg "JOIN_NODE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# Create RPD and go online, which includes creating the
# GS subsystem and starting it
add_and_start_local
else
print_dbgmsg "Not the target"
fi
;;
UNDO_PRE )
# JOIN_NODE UNDO_PRE Processing
print_dbgmsg "JOIN_NODE UNDO_PRE Processing (nothing to do)"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# Nothing to do
else
print_dbgmsg "Not the target"
fi
;;
esac
exit 0
}
################################################################################
# #
# process_RmNode: process RM_NODE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST | UNDO_PRE} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Node UUID #
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_RmNode() {
# echo "process_RmNode: called"
# if the parameters are passed via options then parsing should have
# been done in the main flow
if [[ -z "$PARSED_PHASE" ]]
then
parse_arguments $*
fi
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID, PARSED_TARGET_SHORT_ID,
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
# RM_NODE CHECK Processing
# There is nothing to do for RM_NODE CHECK.
# Exploiters like PowerHA and VIOS may use CHECK to verify that
# the node removal is being orchestrated through the exploiter,
# and reject removals when raw CAA APIs/CLIs are used.
# But RSCT so far has no reason to veto a node/cluster removal.
# Possible reasons to consider rejecting RM_NODE operations:
# * A subsystem is in the middle of a crucial operation
# * Some non-target node in a degraded state (say, GS not
# running)
# * (TBD)
print_dbgmsg "RM_NODE CHECK Processing (nothing to do)"
;;
PRE )
# RM_NODE PRE Processing
print_dbgmsg "RM_NODE PRE Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# This is where the RPD is removed, and the GS daemon is
# stopped. By the time RM_NODE PRE is invoked, the RPD is
# supposed to be offline, since STOP_NODE should have been
# invoked first. However, stop_and_delete_local_pre() does not
# assume that, and will cause GS to stop if it happens to be
# running.
# RPD removal is done at the PRE phase to ensure that the
# removal is processed at the target node *first*, before any
# other node is notified (and starts to take failover/takeover)
# actions.
stop_and_delete_local_pre
else
print_dbgmsg "Not the target node"
fi
;;
POST )
# RM_NODE POST Processing
print_dbgmsg "RM_NODE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# In principle, there is nothing to do since all actions
# were supposed to have been taken during PRE, but this is a
# fallback in case actions in the PRE failed to stop GS and
# remove domain
stop_and_delete_local_post
else
print_dbgmsg "Not the target node"
fi
# For the remote nodes, a check could be added to ensure that
# these nodes no longer see the target node in the cluster
# configuration.
;;
UNDO_PRE )
# RM_NODE UNDO_PRE Processing
print_dbgmsg "RM_NODE UNDO_PRE Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node (nothing to do)"
# In principle, the goal would be to revert the node back to
# the state prior to 'PRE', which was "RPD present but offline",
# while the expected state when UNDO_PRE is invoked is
# "RPD removed and GS is down". However, "doing nothing" in
# UNDO_PRE should be acceptable, since even if JOIN_NODE or
# START_NODE are subsequently run, their processing should
# result in the RPD being (re-) created if not present.
else
print_dbgmsg "Not the target node"
fi
;;
esac
exit 0
}
################################################################################
# #
# process_StartNode: process START_NODE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST | UNDO_PRE} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Node UUID #
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_StartNode() {
# echo "process_StartNode: called"
# if the parameters are passed via options then parsing should have
# been done in the main flow
if [[ -z "$PARSED_PHASE" ]]
then
parse_arguments $*
fi
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID, PARSED_TARGET_SHORT_ID,
# PARSED_AM_I_THE_TARGET
# The processing of START seems identical to that of JOIN. In both cases
# * The CHECK phase should check against non-CAA clusters, given that
# it's not known what happened to the node while it was offline in the
# cluster
# * The RPD domain may have to be created or recreated in case it was
# removed by mistake while the node was offline
# Given the above, we'll just invoke process_JoinNode()
process_JoinNode $*
exit 0
}
################################################################################
# #
# process_StopNode: process STOP_NODE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST | UNDO_PRE} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Node UUID #
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_StopNode() {
# echo "process_StopNode: called"
# if the parameters are passed via options then parsing should have
# been done in the main flow
if [[ -z "$PARSED_PHASE" ]]
then
parse_arguments $*
fi
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID, PARSED_TARGET_SHORT_ID,
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
# STOP_NODE CHECK Processing
# There is nothing to do for STOP_NODE CHECK.
# Exploiters like PowerHA and VIOS may use CHECK to verify that
# the node stop is being orchestrated through the exploiter,
# and reject stops when raw CAA APIs/CLIs are used.
# But RSCT so far has no reason to veto a node stop.
# Possible reasons to consider rejecting STOP_NODE operations:
# * A subsystem is in the middle of a crucial operation
# * (TBD)
print_dbgmsg "STOP_NODE CHECK Processing (nothing to do)"
;;
PRE )
# STOP_NODE PRE Processing
print_dbgmsg "STOP_NODE PRE Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# This is where the GS daemon is stopped, which is done
# with coordination with ConfigRM
stop_local
else
print_dbgmsg "Not the target node"
fi
;;
POST )
# STOP_NODE POST Processing
print_dbgmsg "STOP_NODE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# In principle, there is nothing to do since all actions
# were supposed to have been taken during PRE.
fi
# For the remote nodes, a check could be added to ensure that
# these nodes no longer see the target node as alive in the cluster
;;
UNDO_PRE )
# STOP_NODE UNDO_PRE Processing
print_dbgmsg "STOP_NODE UNDO_PRE Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target node"
# The goal is to revert the node back to
# the state prior to 'PRE', which was "RPD present an online"
# The domain should still be present, since it was not
# supposed to have been removed in the PRE phase.
add_and_start_local
fi
;;
esac
exit 0
}
################################################################################
# #
# process_CommitLvl: process COMMIT_LVL request (all phases) #
# #
# #
# Arguments: <phase>: one of {POST} #
# <cluster id> #
# <other target ids>: #
# flags for new capabilities
# <target indicator>: #
# TARGET_NODE if the local node is the target #
# NON_TARGET_NODE if the local node is NOT the target #
# #
# Return value: None #
# #
################################################################################
process_CommitLvl() {
#echo "process_CommitLvl: called"
# if the parameters are passed via options then parsing should have
# been done in the main flow
if [[ -z "$PARSED_PHASE" ]]
then
parse_arguments $*
fi
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID,
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
#Nothing to do on PRE or UNDO_PRE
CHECK )
print_dbgmsg "COMMIT_LVL CHECK Processing"
commit_lvl_check
;;
PRE )
print_dbgmsg "COMMIT_LVL PRE Processing (nothing to do)"
;;
UNDO_PRE )
print_dbgmsg "COMMIT_LVL UNDO_PRE Processing (nothing to do)"
;;
POST )
print_dbgmsg "COMMIT_LVL POST Processing"
commit_lvl_local
;;
esac
exit 0
}
################################################################################
# #
# check_for_inconsistent_site_info - if CAA is up, check if the #
# local site uuid, name, short id, priority #
# matches the one sent as parameter to the script. #
# #
# The function will output an error message and exit 1 if the above infor #
# don't match. The function will NOT exit if the cluster IDs match. #
# It's up to the caller to exit 0. #
# #
# Depends on the following variables already being set: #
# * SIT_UUID: Local SITE ID (obtained via ct_clusterinfo -U) #
# * SITE_NAME: Local SITE name (obtained via ct_clusterinfo -S) #
# * SITE_NUMBER: Local SITE number (obtained via ct_clusterinfo -N) #
# * SITE_PRIORITY: Local SITE priority (obtained via ct_clusterinfo -p) #
# * PARSED_TARGET_SITE_UUID: Site ID retrieved from the parameters to #
# the script #
# * PARSED_TARGET_SITENAME: Site Name retrieved from the parameters to #
# the script #
# * PARSED_TARGET_SITE_NUMBER: Site number retrieved from the parameters #
# to the script #
# * PARSED_TARGET_SITE_PRIORITY: Site priority retrieved from the #
# parameters to the script #
# #
# Arguments: None #
# #
# Return value: 0: indicates that either CAA is not up or the site #
# information did match #
# #
################################################################################
check_for_inconsistent_site_info() {
if [[ "$REALM" == "CAA" ]]
then
# This indicates that CAA is configured
# just in case: if our site ID is invalid then assume site is not
# really configured
if [[ -z "$SITE_UUID" || "$SITE_UUID" == "00000000-0000-0000-0000-000000000000" ]]
then
print_dbgmsg "Local Site ID ($SITE_UUID) is invalid"
return 0
fi
if [[ -z "$PARSED_TARGET_SITE_UUID" ]]
then
print_dbgmsg "Site ID received as parameter not set"
return 0
fi
if [[ -z "PARSED_TARGET_SITENAME" ]]
then
print_dbgmsg "Site name received as parameter not set"
return 0
fi
if [[ -z "PARSED_TARGET_SITE_NUMBER" ]]
then
print_dbgmsg "Site short id received as parameter not set"
return 0
fi
if [[ -z "PARSED_TARGET_SITE_PRIORITY" ]]
then
print_dbgmsg "Site priority received as parameter not set"
return 0
fi
if [[ "$SITE_UUID" != "$PARSED_TARGET_SITE_UUID" ]]
then
print_dbgmsg "Site ID ($SITE_UUID) does not match ID passed as parameter ($PARSED_TARGET_SITE_UUID)"
exit 1
elif [[ "$SITE_NAME" != "$PARSED_TARGET_SITENAME" ]]
then
print_dbgmsg "Site Name ($SITE_NAME) does not match Name passed as parameter ($PARSED_TARGET_SITENAME)"
exit 1
elif [[ "$SITE_NUMBER" != "$PARSED_TARGET_SITE_NUMBER" ]]
then
print_dbgmsg "Site short id ($SITE_NUMBER) does not match short id passed as parameter ($PARSED_TARGET_SITE_NUMBER)"
exit 1
elif [[ "$SITE_PRIORITY" != "$PARSED_TARGET_SITE_PRIORITY" ]]
then
print_dbgmsg "Site priority ($SITE_PRIORITY) does not match priority passed as parameter ($PARSED_TARGET_SITE_PRIORITY)"
exit 1
else
print_dbgmsg "Site information check succeeded"
fi
fi
return 0
}
################################################################################
# #
# process_AddSite: process ADD_SITE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Site UUID, Site name, Site shortid, Site priority #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_AddSite() {
# echo "process_AddSite: called"
# Assume the following variables have been set:
# PARSED_CLUSTER_UUID, PARSED_TARGET_SITE_NUMBER,
# PARSED_SITE_UUID, PARSED_SITE_NAME, PARSED_SITE_PRIORITY
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
print_dbgmsg "ADD_SITE CHECK Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
# check for inconsistent cluster IDs between what local CAA
# reports and what has been passed as parameter
check_for_inconsistent_cluster_ID
else # Not the local site
print_dbgmsg "Not the target site"
fi
;;
PRE )
print_dbgmsg "ADD_SITE PRE Processing (nothing to do)"
# No real work to do in ADD_SITE PRE (just make sure we exit 0,
# so that CAA will not cancel the operation)
;;
POST )
print_dbgmsg "ADD_SITE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
tell_ConfigRM_resync
print_dbgmsg "Target site"
else
print_dbgmsg "Not the target site"
fi
;;
esac
exit 0
}
################################################################################
# #
# process_ModSite: process MOD_SITE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Site UUID, Site name, Site shortid, Site priority #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# The final changes would be reflected in POST phase #
# #
# Return value: None #
# #
################################################################################
process_ModSite() {
# echo "process_ModSite: called"
# Assume the following variables have been set:
# PARSED_CLUSTER_UUID, PARSED_TARGET_SITE_NUMBER,
# PARSED_SITE_UUID, PARSED_SITE_NAME, PARSED_SITE_PRIORITY
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
print_dbgmsg "MOD_SITE CHECK Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
# check for inconsistent cluster IDs between what local CAA
# reports and what has been passed as parameter
check_for_inconsistent_cluster_ID
# check for inconsistent site information between what local site
# reports and what has been passed as parameter
check_for_inconsistent_site_info
else # Not the local site
print_dbgmsg "Not the target site"
fi
;;
PRE )
print_dbgmsg "MOD_SITE PRE Processing (nothing to do)"
# No real work to do in MOD_SITE PRE (just make sure we exit 0,
# so that CAA will not cancel the operation)
;;
POST )
print_dbgmsg "MOD_SITE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
tell_ConfigRM_resync
print_dbgmsg "Target site"
else
print_dbgmsg "Not the target site"
fi
;;
esac
exit 0
}
################################################################################
# #
# process_RmSite: process RM_SITE request (all phases) #
# #
# #
# Arguments: <phase>: one of {CHECK| PRE | POST} #
# <cluster id> #
# <other target ids>: #
# NO_OTHER_TARGETS if none #
# <target id>: Site UUID #
# #
# If the script parameters were processed in the main flow (which should be #
# the case) then the environment variables mentioned below are assumed to #
# have been set, and in that case this function is invoked with no arguments. #
# #
# #
# Return value: None #
# #
################################################################################
process_RmSite() {
# echo "process_RmSite: called"
# Assume the following variables have been set:
# PARSED_PHASE, PARSED_CLUSTER_UUID, PARSED_TARGET_SITE_NUMBER,
# PARSED_SITE_UUID, PARSED_SITE_NAME, PARSED_SITE_PRIORITY
# PARSED_AM_I_THE_TARGET
case $PARSED_PHASE in
CHECK )
print_dbgmsg "RM_SITE CHECK Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
print_dbgmsg "Target site"
else # Not the local site
print_dbgmsg "Not the target site"
fi
;;
PRE )
print_dbgmsg "RM_SITE PRE Processing (nothing to do)"
# No real work to do in ADD_SITE PRE (just make sure we exit 0,
# so that CAA will not cancel the operation)
;;
POST )
print_dbgmsg "RM_SITE POST Processing"
if [[ "$PARSED_AM_I_THE_TARGET" == "yes" ]]
then
tell_ConfigRM_resync
print_dbgmsg "Target site"
else
print_dbgmsg "Not the target site"
fi
;;
esac
exit 0
}
################################################################################
# #
# #
# MAIN FUNCTION #
# #
# #
# #
# #
################################################################################
#
get_current_locale
# parse the arguments
gs_rc=0 # cthagsctrl return code
Argcount=0
#
# Expected syntax:
# ctcaactrl -O <operation>
# -P <phase>
# -T <trans id>
# -c <cluster id>
# -o <other targets> (not for ADD_NODE)
# -t <target>
# [-m]
# [-g]
# [-i]
#
BAD_ARG=0
UNKNOWN_OPERATION=0
UNKNOWN_PHASE=0
SITE_OPERATION=0
OPERATION=""
PARSED_PHASE=""
PARSED_TRANS_ID=""
PARSED_CLUSTER_UUID=""
PARSED_TARGET_UUID=""
PARSED_TARGET_HOSTNAME=""
PARSED_TARGET_SHORT_ID=""
PARSED_SITE_UUID=""
PARSED_SITENAME=""
PARSED_SITE_NUMBER=""
PARSED_SITE_PRIORITY=""
PARSED_TARGET_SITE_UUID=""
PARSED_TARGET_SITENAME=""
PARSED_TARGET_SITE_NUMBER=""
PARSED_TARGET_SITE_PRIORITY=""
PARSED_AM_I_THE_TARGET="no" # assume not the target unless -m specified
PARSED_AM_I_THE_INITIATOR="no" # assume not the initiator unless -i specified
PARSED_TARGET_IS_GATEWAY_SERVER="no"
LKU_PD_MARKER_FILE="/var/ct/lvupd/lku_PD_mark_file"
LKU_DISABLE_START=0
OPERATION_CALLER=""
CT_CAA_STATE_DEFINED_FILE="/var/ct/cfg/caa-state-defined"
CT_CAA_STATE_STOPPED_FILE="/var/ct/cfg/caa-state-stopped"
CAP_SPLT_MRG_FOUND=0
# Create logfile if defined
create_logfile_with_permission "$HA_CCAL_LOG" "0640"
#
print_dbgmsg "--------------------------------------"
print_dbgmsg "Invoked with parameters: $*"
print_process_tree # show process tree
#
# LKU (live kernel update) changes
#
# If marker file (/var/ct/lvupd/lku_mark_file - This file is controlled by "ctlvupdate" script) is
# exists don't allow any operation as presence of the marker file means currently LKU is going on.
#
if [[ -e "$LKU_PD_MARKER_FILE" ]]
then
print_dbgmsg "The LKU is currently going on, start operations cannot be executed."
LKU_DISABLE_START=1
fi
# Note the ':' as first argument in getopts: this changes how getopts
# deals with unknown options. The goal is to be able to ignore functions
# we don't understand.
while getopts ":O:P:T:c:o:t:s:l:mgiv:" opt
do
case $opt in
O ) # operation: ADD_NODE, etc
OPERATION=$OPTARG
;;
P )
PARSED_PHASE=$OPTARG
case $PARSED_PHASE in
CHECK | PRE | POST | UNDO_PRE)
;;
*) # unknown operation/phase
UNKNOWN_PHASE=1
;;
esac
Argcount=$((Argcount + 1))
;;
T )
PARSED_TRANS_ID=$OPTARG
;;
c )
PARSED_CLUSTER_UUID=$OPTARG
;;
o ) # other targets
# (not used for now)
;;
s ) # Site UUID
typeset site_info=$OPTARG
if [[ "$OPERATION" == "ADD_SITE" || "$OPERATION" == "MOD_SITE" || "$OPERATION" == "RM_SITE" ]]
then
SITE_OPERATION=1
if echo $site_info | grep "," > /dev/null
then
# Format assumed to be: Site UUID, Site name, site short id, site priority
# Site UUID
PARSED_SITE_UUID=${site_info%%,*}
rest=${site_info#*,}
# Site Name
PARSED_SITENAME=${rest%%,*}
rest=${rest#*,}
# Site short id
PARSED_SITE_NUMBER=${rest%%,*}
rest=${rest#*,}
# Site priority
PARSED_SITE_PRIORITY=${rest%%,*}
fi
else
# format assumed to Site UUID
PARSED_SITE_UUID=$site_info
fi
;;
t ) # target
typeset target_id_info=$OPTARG
if [[ "$OPERATION" == "ADD_SITE" || "$OPERATION" == "MOD_SITE" || "$OPERATION" == "RM_SITE" ]]
then
SITE_OPERATION=1
if echo $target_id_info | grep "," > /dev/null
then
# Format assumed to be: Site UUID, Site name, site short id, site priority
# Site UUID
PARSED_TARGET_SITE_UUID=${target_id_info%%,*}
rest=${target_id_info#*,}
# Site Name
PARSED_TARGET_SITENAME=${rest%%,*}
rest=${rest#*,}
# Site short id
PARSED_TARGET_SITE_NUMBER=${rest%%,*}
rest=${rest#*,}
# Site priority
PARSED_TARGET_SITE_PRIORITY=${rest%%,*}
else
# Format assumed to be: Site UUID
PARSED_TARGET_SITE_UUID=$target_id_info
fi
else
if echo $target_id_info | grep "," > /dev/null
then
# Format assumed to be: Node UUID, canonical hostname, shortid
PARSED_TARGET_UUID=${target_id_info%%,*}
rest=${target_id_info#*,}
# echo "PARSED_TARGET_UUID=$PARSED_TARGET_UUID"
PARSED_TARGET_HOSTNAME=${rest%%,*}
rest=${rest#*,}
# echo "PARSED_TARGET_HOSTNAME=$PARSED_TARGET_HOSTNAME"
PARSED_TARGET_SHORT_ID=${rest%%,*}
# echo "PARSED_TARGET_SHORT_ID=$PARSED_TARGET_SHORT_ID"
else
# Format assumed to be: Node UUID
PARSED_TARGET_UUID=$target_id_info
fi
fi
;;
m ) # local node is the target
PARSED_AM_I_THE_TARGET="yes"
;;
g ) # target is gateway server
PARSED_TARGET_IS_GATEWAY_SERVER="yes"
;;
i ) # local node is the initiator
PARSED_AM_I_THE_INITIATOR="yes"
;;
v )
if [[ "$OPERATION" == "MOD_TUNE" ]]
then
# Tunable details
typeset tunable_info=$OPTARG
# Format assumed to be: Parameter,Value
print_dbgmsg "Received tunable setting: $tunable_info"
PARSED_TUNABLE_PARAMETER=${tunable_info%%,*}
PARSED_TUNABLE_VALUE=${tunable_info#*,}
print_dbgmsg "Parameter: $PARSED_TUNABLE_PARAMETER"
print_dbgmsg "Value: $PARSED_TUNABLE_VALUE"
elif [[ "$OPERATION" == "COMMIT_LVL" ]]
then
typeset capabilities=$OPTARG
# check if SPLT_MRG capability found from the input list
if [[ "$capabilities" == *"SPLT_MRG"* ]]
then
CAP_SPLT_MRG_FOUND=1
fi
fi
;;
a ) Op=AddAndStart; Argcount=$((Argcount + 1));;
s ) Op=AddAndStart; Argcount=$((Argcount + 1));;
d ) Op=StopAndDelete; Argcount=$((Argcount + 1));;
k ) Op=stop; Argcount=$((Argcount + 1));;
r ) Op=refresh; Argcount=$((Argcount + 1));;
h ) Op=help
print_message IMSG_ctcaactrl_Usage $basecmd
exit 0;;
l ) # flag to include the caller
OPERATION_CALLER=$OPTARG
;;
?|* )
print_message EMSG100 $basecmd $OPTARG
# since ':' was used as first argument in getopts, the unknown option
# is stored in $OPTARG
# No longer quit if an unknown option is received: ignore the option.
# This allows CAA to add new arguments that can be ignored without
# causing a failure
# print_message IMSG_ctcaactrl_Usage $basecmd
# exit 1;;
;;
esac
done
#
# Log the caller that initiated the operation
#
if [[ -n "$OPERATION_CALLER" ]]
then
print_dbgmsg "The operation has been initiated by $OPERATION_CALLER"
fi
#
# Some options are mandatory
#
# if OPERATION is set then must have PARSED_PHASE
if [[ -n "$OPERATION" ]]
then
if [[ -z "$PARSED_PHASE" ]]
then
BAD_ARG=1
fi
fi
# if OPERATION/PARSED_PHASE are set then must have cluster ID and target
# (removed requirement for target when op is COMMIT_LVL - does not require one)
if [[ -n "$PARSED_PHASE" ]]
then
if [[ -z "$PARSED_CLUSTER_UUID" ]]
then
BAD_ARG=1
fi
#
# both PARSED_TARGET_UUID & PARSED_TARGET_SITE_UUID
# can't be NULL
# or
# can have values
# at same time when OP is not COMMIT_LVL
#
if [[ "$OPERATION" != "COMMIT_LVL" ]]
then
if [[ -z "$PARSED_TARGET_UUID" && -z "$PARSED_TARGET_SITE_UUID" || \
-n "$PARSED_TARGET_UUID" && -n "$PARSED_TARGET_SITE_UUID" ]]
then
BAD_ARG=1
fi
fi
fi
if [[ "$BAD_ARG" == 1 ]]
then
print_message IMSG_ctcaactrl_Usage $basecmd
exit 1
fi
#remove options
shift $(($OPTIND - 1))
if [[ -n "$OPERATION" ]]
then
# if -o option specified then use that as 'operation'
CAACmdOpt=$OPERATION
# [ADD_NODE | RM_NODE | STOP_NODE | START_NODE |
# JOIN_NODE | ADD_DISK | RM_DISK | COMMIT_LVL |
# ADD_SITE | MOD_SITE | RM_SITE]
else
CAACmdOpt=$1 # [START | STOP | PRE_START | POST_START | PRE_STOP]
fi
if [[ "$LKU_DISABLE_START" == 1 ]]
then
echo "ADD_NODE START_NODE JOIN_NODE START PRE_START POST_START" | grep -w $CAACmdOpt > /dev/null 2>&1
rc=$?
if [[ 0 == $rc ]]
then
print_dbgmsg "The ${CAACmdOpt} operation cannot be executed."
exit 0
fi
fi
CAAForceOpt=$2 # [FORCE] for stop
print_dbgmsg "Command: $CAACmdOpt Phase: $PARSED_PHASE TransID: $PARSED_TRANS_ID"
print_dbgmsg "Cluster ID: $PARSED_CLUSTER_UUID"
if [[ "$SITE_OPERATION" == 1 ]]
then
print_dbgmsg "Target Site UUID: $PARSED_TARGET_SITE_UUID"
else
print_dbgmsg "Target Node UUID: $PARSED_TARGET_UUID"
fi
if [[ "$SITE_OPERATION" == 0 && -n "$PARSED_SITE_UUID" ]]
then
print_dbgmsg "Target Site UUID: $PARSED_SITE_UUID"
fi
if [[ "$SITE_OPERATION" == 1 ]]
then
print_dbgmsg "Is local site the target: $PARSED_AM_I_THE_TARGET"
print_dbgmsg "Is local site the initiator: $PARSED_AM_I_THE_INITIATOR"
else
print_dbgmsg "Is local node the target: $PARSED_AM_I_THE_TARGET"
print_dbgmsg "Is local node the initiator: $PARSED_AM_I_THE_INITIATOR"
fi
if [[ "$UNKNOWN_PHASE" == 1 ]]
then
print_dbgmsg "Unknown phase ($PARSED_PHASE) ignored"
exit 0
fi
if [[ ! -z $CAACmdOpt ]]
then
case $CAACmdOpt in
ADD_NODE ) Op=AddNode
Argcount=$((Argcount + 1))
;;
RM_NODE ) Op=RmNode
Argcount=$((Argcount + 1))
;;
STOP_NODE ) Op=StopNode
Argcount=$((Argcount + 1))
;;
START_NODE ) Op=StartNode
Argcount=$((Argcount + 1))
;;
JOIN_NODE ) Op=JoinNode
Argcount=$((Argcount + 1))
;;
ADD_DISK ) Op=AddDisk
Argcount=$((Argcount + 1))
;;
RM_DISK ) Op=RmDisk
Argcount=$((Argcount + 1))
;;
COMMIT_LVL ) Op=CommitLvl
Argcount=$((Argcount + 1))
;;
START ) Op=AddAndStart
Argcount=$((Argcount + 1))
;;
STOP ) Op=StopAndDelete
Argcount=$((Argcount + 1))
;;
PRE_START ) Op=PreStart
Argcount=$((Argcount + 1))
;;
POST_START ) Op=AddAndStart # same as START
Argcount=$((Argcount + 1))
;;
PRE_STOP ) Op=StopAndDelete # same as STOP
Argcount=$((Argcount + 1))
;;
POST_STOP ) Op=PostStop # leftover cleanup steps
Argcount=$((Argcount + 1))
;;
PRE_SYNC ) Op=PreSync #NoOp
Argcount=$((Argcount + 1))
;;
POST_SYNC ) Op=PostSync #NoOp
Argcount=$((Argcount + 1))
;;
ADD_SITE ) Op=AddSite
Argcount=$((Argcount + 1))
;;
MOD_SITE ) Op=ModSite
Argcount=$((Argcount + 1))
;;
RM_SITE ) Op=RmSite
Argcount=$((Argcount + 1))
;;
MOD_TUNE ) Op=ModTune
Argcount=$((Argcount + 1))
;;
* ) # Unknown operation: ignore
UNKNOWN_OPERATION=1
;;
esac
fi
if [[ "$UNKNOWN_OPERATION" == 1 ]]
then
print_dbgmsg "Unknown operation ignored"
exit 0
fi
if [[ -z $CAAForceOpt ]]; then
CAAForceOpt=NONE
fi
# To shortcut....
if [[ "PreSync" == "$Op" || "PostSync" == "$Op" || "AddDisk" == "$Op" || "RmDisk" == "$Op" ]]
then
# Note: PreSync/PostSync seem happenning too often.
# RSCT does not use those events.
# Ignore those events even from the logging.
exit 0;
fi
# Check CCAL log file size and backup if necessary
backup_ccal_log_file
#
if [[ -n "$*" ]]
then
print_dbgmsg ""
print_dbgmsg "$* called"
fi
#
if ((Argcount == 0))
then
# no argument specified
print_message EMSG101 $basecmd
print_message IMSG_ctcaactrl_Usage $basecmd
exit 1
fi
## if ((Argcount > 1))
## then
## # More than one argument specified
## print_message EMSG102 $basecmd
## print_message IMSG_ctcaactrl_Usage $basecmd
## exit 1
## fi
print_dbgmsg "Op=$Op"
# get the cluster/caa info
set -A clinfo $(${RSCTBIN}/ct_clusterinfo -r -c -n -i -u -M)
REALM=${clinfo[0]} # realm (CAA or CLUSTER)
CLUSTER_NAME=${clinfo[1]} # cluster-name
NODE_NUMBER=${clinfo[2]} # node-number
CLUSTER_ID=${clinfo[3]} # cluster-id
CAA_UUID=${clinfo[4]} # caa-cluster-uuid
if [[ "$SITE_OPERATION" == 1 ]]
then
set -A clinfo $(${RSCTBIN}/ct_clusterinfo -U -S -N -p)
SITE_UUID=${clinfo[1]} # site uuid
SITE_NAME=${clinfo[2]} # site name
SITE_NUMBER=${clinfo[3]} # site number
SITE_PRIORITY=${clinfo[4]} # site priority
fi
print_dbgmsg "REALM=$REALM CLUSTER_NAME=$CLUSTER_NAME NODE_NUMBER=$NODE_NUMBER CLUSTER_ID=$CLUSTER_ID CAA_UUID=$CAA_UUID"
# No need to shift if the parameters were passed via options
if [[ -z "$OPERATION" ]]
then
shift # shift arguments since operand (1st argument) has been processed
fi
# Start 'switch' for the arguments
case $Op in
AddNode )
process_AddNode $*
;;
RmNode )
process_RmNode $*
;;
StopNode )
process_StopNode $*
;;
StartNode )
process_StartNode $*
;;
JoinNode )
process_JoinNode $*
;;
CommitLvl )
process_CommitLvl $*
;;
AddSite )
process_AddSite $*
;;
ModSite )
process_ModSite $*
;;
RmSite )
process_RmSite $*
;;
esac
#end of 'switch' for the arguments
if [[ "PreStart" == "$Op" ]]
then
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
if [[ "$caa_is_disabled_for_migration" == "true" ]]
then
#
# If "disabled for migration", the PowerHA/RPD domain is being
# migrated in coordination with RSCT, so having a previously
# defined domain is not an error.
#
print_dbgmsg "SUCESS: RSCT-to-CAA is in progress"
exit 0
fi
clusters=$(Get_CAA_RPD_Clusters)
if [[ -n "$clusters" ]]
then
#
# The existence of a CAA RPD cluster means we have a previously
# defined domain, which is not an error.
#
print_dbgmsg "SUCCESS: CAA RPD cluster already exists: $clusters"
exit 0
fi
#
# See if a PowerHA cluster is configured which includes this node.
# If so, the odm entry for the HACMPcluster object class will have
# a non-zero id attribute value, and a non-empty name attribute value.
#
power_ha=`/usr/bin/odmget HACMPcluster 2>/dev/null`
rc=$?
if [[ 0 == $rc ]]
then
#
# grep for the id.
#
power_ha_id=`echo "$power_ha" | /usr/bin/grep -w "id"`
rc=$?
if [[ 0 == $rc ]]
then
#
# grep for the id.
#
power_ha_name=`echo "$power_ha" | /usr/bin/grep -w "name"`
rc=$?
if [[ 0 == $rc ]]
then
#
# The id is an integer. Remove the "large left" pattern up
# through a space to leave only the integer itself.
#
power_ha_id=${power_ha_id##* }
#
# The name is a string. Remove the "small left" pattern up
# through the first double quote, then the "small right"
# pattern from the second double quote on, leaving only the
# string itself.
#
power_ha_name=${power_ha_name#*\"}
power_ha_name=${power_ha_name%\"*}
if [[ "0" == "$power_ha_id" ]]
then
print_dbgmsg "SUCCESS: power_ha_id is 0"
exit 0
fi
if [ -z "$power_ha_name" ]
then
print_dbgmsg "SUCCESS: power_ha is not defined"
exit 0
fi
#
# The id and the name are legitimate. Consider PowerHA
# configured.
#
# If PowerHA cluster version is at least 7.1 then
# this cluster creation is being made on behalf of a
# CAA-aware instance of PowerHA. Allow cluster.
PowerHA7_version=12
cluster_version=`/usr/es/sbin/cluster/utilities/clmixver`
rc=$?
if [[ ($rc == 0) || ($rc == 1) ]]
then
if (( $cluster_version >= $PowerHA7_version ))
then
print_dbgmsg "PowerHA 7.1 or later present"
exit 0
else
print_dbgmsg "PowerHA pre-7.1 present: cluster version $cluster_version"
fi
else
# A problem with clmixver: should not happen, but it does,
# we assume that a propoer PowerHA cluster is not present
print_dbgmsg "clmixver error: $rc. Assume cluster not present"
exit 0
fi
print_message EMSG111 $basecmd $power_ha_id $power_ha_name
exit 1
fi
fi
fi
#
# See if an RPD is configured which includes this node. If so,
# there will be more than just an IW entry in /var/ct/cfg/clusters.
#
clusters=$(GetRPDClusters)
if [[ "" != "$clusters" ]]
then
#
# An RPD is configured on this node.
#
migration_realm=`ct_caa_get_migration_realm`
print_dbgmsg "migration_realm=$migration_realm"
if [[ "$caa_is_disabled_for_migration" == "false" || "$migration_realm" != "RPD" ]]
then
# RPD is not in migration.
print_dbgmsg "Fail: RPDs: $clusters are already defined"
print_message EMSG112 $basecmd
exit 1
fi
fi
exit 0
fi
if [[ "PreSync" == "$Op" || "PostSync" == "$Op" ]]
then
# do nothing
exit 0;
fi
if [[ -z $REALM ]]; then
REALM=CLUSTER
fi
if [[ -z ${CLUSTER_ID} ]]; then
# No CAA cluster info found
print_message EMSG110 $basecmd
exit 1
fi
## Now CAA is active
#
#
case $Op in
AddAndStart )
#
# Start with assumptions about what this block should accomplish.
#
print_dbgmsg "Processing AddAndStart ... "
# Important: Write CAA cmd file so that ConfigRM can understand the operation
echo "START $CLUSTER_NAME $CLUSTER_ID" > $CAACMDCTRLFILE
if [[ ! -z $REALM && $REALM != "CAA" ]]; then
# non CAA environment
print_message EMSG110 $basecmd
exit 1
fi
#
create_group_services_subsystem=true
start_group_services_subsystem=true
refresh_ConfigRM=true
#
# Take migration into account. Start by discovering whether operating in a
# CAA environment should be disabled because migration is taking place.
#
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
if [[ "$caa_is_disabled_for_migration" == "true" ]]
then
start_group_services_subsystem=false
refresh_ConfigRM=false
migration_realm=`ct_caa_get_migration_realm`
print_dbgmsg "migration_realm=$migration_realm"
if [[ "$migration_realm" == "RPD" ]]
then
#
# Don't need to create the Group Services subsystem in an RPD
# environment, because it should already have been created in
# that environment.
#
create_group_services_subsystem=false
fi
else # not disabled for migration (but migration may still be under way)
# ctcaactrl may have been invoked via ct_caa_gs_migrate_ctrl.
# If the GS daemon undegoing migration is still running then
# it should not be started. Especially critical in PowerHA migration,
# where the subsystem being running is grpsvcs, not cthags.
# In addition, if the migrated GS daemon is still running, the
# cthags subsystem will have been already created, at the time
# ctcaactrl POST_START was invoked upon CAA cluster creation. So no
# need to create GS subsystems again
migrated_GS_daemon_still_running=`ct_caa_migrated_GS_still_running`
if [[ $migrated_GS_daemon_still_running == "true" ]]
then
print_dbgmsg "GS daemon which underwent migration is still running: do not start cthags"
create_group_services_subsystem=false
start_group_services_subsystem=false
fi
fi
print_dbgmsg "refresh_ConfigRM=$refresh_ConfigRM"
print_dbgmsg "start_group_services_subsystem=$start_group_services_subsystem"
print_dbgmsg "create_group_services_subsystem=$create_group_services_subsystem"
#
# Setup the cluster info & dir.
#
if [[ ! -d /var/ct/${CLUSTER_ID} || ! -d /var/ct/${CLUSTER_NAME} ]]
then
# cluster does not exist...then create it
print_dbgmsg "Configure CAA ${CLUSTER_NAME} ${CLUSTER_ID} ${NODE_NUMBER}"
cfghact -c ${CLUSTER_NAME} -n ${NODE_NUMBER} -y ${CLUSTER_ID} -t CAA -f
else
print_dbgmsg "/var/ct/${CLUSTER_ID} is already found"
fi
if [[ "$create_group_services_subsystem" == "true" ]]
then
print_dbgmsg "Calling $RSCTBIN/cthagsctrl -a ..."
$RSCTBIN/cthagsctrl -a
gs_rc=$?
fi
if [[ "$start_group_services_subsystem" == "true" ]]
then
print_dbgmsg "Calling $RSCTBIN/cthagsctrl -s ..."
$RSCTBIN/cthagsctrl -s
gs_rc=$?
fi
if [[ "$refresh_ConfigRM" == "true" ]]
then
tell_ConfigRM_resync $Op
else
# Make ConfigRM auto-startable
touch /var/ct/cfg/ConfigRM.autostart
fi
print_dbgmsg "Initialization done rc=$gs_rc"
exit_script # exit the script
;;
StopAndDelete )
#StopAndDelete subsystems
# Note: since this script is being invoked with PRE_STOP, a non-zero
# exit code will indicate to CAA that this code cannot be deleted from
# the cluster.
print_dbgmsg "Processing StopAndDelete... "
# Write CMD file (STOP)
echo "STOP $CLUSTER_NAME $CLUSTER_ID" > $CAACMDCTRLFILE
echo "STOP-COMMIT $CLUSTER_NAME $CLUSTER_ID" > $CAACOMMITCTRLFILE
if [[ $REALM != "CAA" && $CAAForceOpt != "FORCE" ]]; then
# non CAA environment and Not force
print_message EMSG110 $basecmd
print_dbgmsg "Assume StopAndDelete as successful even if REALM is not CAA"
# exiting 0, since we don't the PRE_STOP invocation to return
# non-zero, except in cases where cluster cannot be allowed to be
# removed
exit 0
fi
# If ConfigRM is running
state=`LC_ALL=C /usr/bin/lssrc -s IBM.ConfigRM 2> /dev/null | LC_ALL=C /bin/sed -e"1d;s/.* \([a-zA-Z0-9]*\)$/\1/"`;
if [[ -z $state || $state != "active" ]]
then
# ConfigRM is not running, remove CAA cluster structure
print_dbgmsg "Unconfig CAA cluster..."
# put "IW" to default_cluster so that ConfigRM won't react to online
IWClusterID=$(grep -w "IW" /var/ct/cfg/clusters | awk '{print $1}')
if [[ -n ${IWClusterID} ]]; then
print "${IWClusterID}" > /var/ct/cfg/default_cluster
fi
# Stop HAGS if ConfigRM is not running
cthagsctrl -k
fi
# Tell ConfigRM to resync
tell_ConfigRM_resync $Op
# Wait a few seconds so that ConfigRM can process the active RMC requests (e.g. rmrpdomain)
# This will wait upto 60secs
waitfor_hags_termination_or_stopsrc
print_dbgmsg "Processing StopAndDelete Done. "
exit 0 # see comment above on the exit value
;;
PostStop )
# show the current states
caa_is_disabled_for_migration=`ct_caa_is_disabled_for_migration`
print_dbgmsg "caa_is_disabled_for_migration=$caa_is_disabled_for_migration"
migration_realm=`ct_caa_get_migration_realm`
print_dbgmsg "migration_realm=$migration_realm"
#
# See if an RPD is configured which includes this node. If so,
# there will be more than just an IW entry in /var/ct/cfg/clusters.
#
clusters=$(GetRPDClusters)
if [[ "" != "$clusters" ]]
then
# An RPD is configured on this node.
print_dbgmsg "RPDs: $clusters are already defined. HAGS subsystem will not be stopped."
else
print_dbgmsg "Stopping cthags if it is still up"
# Final cleanup steps in case StopAndDelete did not get to complete them
cthagsctrl -k
fi
# Tell ConfigRM to resync (so that it can delete tree if it is removed)
tell_ConfigRM_resync $Op
# Make sure HAGS to be stopped
waitfor_hags_termination_or_stopsrc
#
# Defect 166865
# Migration cleanup.
#
ct_caa_set_migration_cleanup_complete
print_dbgmsg "Processing PostStop Done."
;;
refresh )
# refresh HAGS
$RSCTBIN/cthagsctrl -r
# refresh ConfigRM
refresh -s IBM.ConfigRM
gs_rc=$?
;;
stop )
$RSCTBIN/cthagsctrl -k
gs_rc=$?
;;
ModTune )
if [[ "POST" != $PARSED_PHASE ]]
then
print_dbgmsg "Tunables only handled in POST phase"
else
print_dbgmsg "No action for $PARSED_TUNABLE_PARAMETER"
fi
;;
esac
exit_script # exit this script
exit 1 # just in case
#-----------------------------------------------------