Page MenuHomeClusterLabs Projects
Files F7609897

No OneTemporary
Actions

View Options

diff --git a/heartbeat/IPsrcaddr b/heartbeat/IPsrcaddr
index 1bd41a930..66e2ad8cd 100755
--- a/heartbeat/IPsrcaddr
+++ b/heartbeat/IPsrcaddr
@@ -1,625 +1,628 @@
#!/bin/sh
#
# Description: IPsrcaddr - Preferred source(/dest) address modification
#
# Author: John Sutton <john@scl.co.uk>
# Support: users@clusterlabs.org
# License: GNU General Public License (GPL)
# Copyright: SCL Internet
#
# Based on the IPaddr script.
#
# This script manages the preferred source address associated with
# packets which originate on the localhost and are routed through the
# matching route. By default, i.e. without the use of this script or
# similar, these packets will carry the IP of the primary i.e. the
# non-aliased interface. This can be a nuisance if you need to ensure
# that such packets carry the same IP irrespective of which host in
# a redundant cluster they actually originate from.
#
# It can add a preferred source address, or remove one.
#
# usage: IPsrcaddr {start|stop|status|monitor|validate-all|meta-data}
#
# The "start" arg adds a preferred source address.
#
# Surprisingly, the "stop" arg removes it. :-)
#
# NOTES:
#
# 1) There must be one and not more than 1 matching route! Mainly because
# I can't see why you should have more than one. And if there is more
# than one, we would have to box clever to find out which one is to be
# modified, or we would have to pass its identity as an argument.
#
# 2) The script depends on Alexey Kuznetsov's ip utility from the
# iproute aka iproute2 package.
#
# 3) No checking is done to see if the passed in IP address can
# reasonably be associated with the interface on which the default
# route exists. So unless you want to deliberately spoof your source IP,
# check it! Normally, I would expect that your haresources looks
# something like:
#
# nodename ip1 ip2 ... ipN IPsrcaddr::ipX
#
# where ipX is one of the ip1 to ipN.
#
# OCF parameters are as below:
# OCF_RESKEY_ipaddress
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
. ${OCF_FUNCTIONS_DIR}/findif.sh
# Defaults
OCF_RESKEY_ipaddress_default=""
OCF_RESKEY_cidr_netmask_default=""
OCF_RESKEY_destination_default="0.0.0.0/0"
OCF_RESKEY_proto_default=""
OCF_RESKEY_metric_default=""
OCF_RESKEY_table_default=""
: ${OCF_RESKEY_ipaddress=${OCF_RESKEY_ipaddress_default}}
: ${OCF_RESKEY_cidr_netmask=${OCF_RESKEY_cidr_netmask_default}}
: ${OCF_RESKEY_destination=${OCF_RESKEY_destination_default}}
: ${OCF_RESKEY_proto=${OCF_RESKEY_proto_default}}
: ${OCF_RESKEY_metric=${OCF_RESKEY_metric_default}}
: ${OCF_RESKEY_table=${OCF_RESKEY_table_default}}
#######################################################################
[ -z "$OCF_RESKEY_proto" ] && PROTO="" || PROTO="proto $OCF_RESKEY_proto"
[ -z "$OCF_RESKEY_table" ] && TABLE="" || TABLE="table $OCF_RESKEY_table"
USAGE="usage: $0 {start|stop|status|monitor|validate-all|meta-data}";
CMDSHOW="$IP2UTIL route show $TABLE to exact $OCF_RESKEY_destination"
CMDCHANGE="$IP2UTIL route change to "
if [ "$OCF_RESKEY_destination" != "0.0.0.0/0" ]; then
CMDSHOW="$CMDSHOW src $OCF_RESKEY_ipaddress"
fi
if [ "$OCF_RESKEY_table" = "local" ]; then
TABLE="$TABLE local"
fi
SYSTYPE="`uname -s`"
usage() {
echo $USAGE >&2
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="IPsrcaddr" version="1.0">
<version>1.0</version>
<longdesc lang="en">
Resource script for IPsrcaddr. It manages the preferred source address
modification.
Note: DHCP should not be enabled for the interface serving the preferred
source address. Enabling DHCP may result in unexpected behavior, such as
the automatic addition of duplicate or conflicting routes. This may
cause the IPsrcaddr resource to fail, or it may produce undesired
behavior while the resource continues to run.
</longdesc>
<shortdesc lang="en">Manages the preferred source address for outgoing IP packets</shortdesc>
<parameters>
<parameter name="ipaddress" unique="0" required="1">
<longdesc lang="en">
The IP address.
</longdesc>
<shortdesc lang="en">IP address</shortdesc>
<content type="string" default="${OCF_RESKEY_ipaddress_default}" />
</parameter>
<parameter name="cidr_netmask">
<longdesc lang="en">
The netmask for the interface in CIDR format. (ie, 24), or in
dotted quad notation 255.255.255.0).
</longdesc>
<shortdesc lang="en">Netmask</shortdesc>
<content type="string" default="${OCF_RESKEY_cidr_netmask_default}"/>
</parameter>
<parameter name="destination">
<longdesc lang="en">
The destination IP/subnet for the route (default: $OCF_RESKEY_destination_default)
</longdesc>
<shortdesc lang="en">Destination IP/subnet</shortdesc>
<content type="string" default="${OCF_RESKEY_destination_default}" />
</parameter>
<parameter name="proto">
<longdesc lang="en">
Proto to match when finding network. E.g. "kernel".
</longdesc>
<shortdesc lang="en">Proto</shortdesc>
<content type="string" default="${OCF_RESKEY_proto_default}" />
</parameter>
<parameter name="metric">
<longdesc lang="en">
Metric. Only needed if incorrect metric value is used.
</longdesc>
<shortdesc lang="en">Metric</shortdesc>
<content type="string" default="${OCF_RESKEY_metric_default}" />
</parameter>
<parameter name="table">
<longdesc lang="en">
Table to modify. E.g. "local".
The table has to have a route matching the "destination" parameter.
</longdesc>
<shortdesc lang="en">Table</shortdesc>
<content type="string" default="${OCF_RESKEY_table_default}" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="20s" />
<action name="stop" timeout="20s" />
<action name="monitor" depth="0" timeout="20s" interval="10s" />
<action name="validate-all" timeout="5s" />
<action name="meta-data" timeout="5s" />
</actions>
</resource-agent>
END
}
errorexit() {
ocf_exit_reason "$*"
exit $OCF_ERR_GENERIC
}
#
# We can distinguish 3 cases: no preferred source address, a
# preferred source address exists which matches that specified, and one
# exists but doesn't match that specified. srca_read() returns 1,0,2
# respectively.
#
# The output of route show is something along the lines of:
#
# default via X.X.X.X dev eth1 src Y.Y.Y.Y
#
# where the src clause "src Y.Y.Y.Y" may or may not be present
WS="[[:blank:]]"
OCTET="[0-9]\{1,3\}"
IPADDR="\($OCTET\.\)\{3\}$OCTET"
SRCCLAUSE="src$WS$WS*\($IPADDR\)"
MATCHROUTE="\(.*${WS}\)\($SRCCLAUSE\)\($WS.*\|$\)"
METRICCLAUSE=".*\(metric$WS[^ ]\+\)"
PROTOCLAUSE=".*\(proto$WS[^ ]\+\).*"
FINDIF=findif
# findif needs that to be set
export OCF_RESKEY_ip=$OCF_RESKEY_ipaddress
srca_read() {
# Capture matching route - doublequotes prevent word splitting...
ROUTE="`$CMDSHOW dev $INTERFACE 2> /dev/null`" || errorexit "command '$CMDSHOW' failed"
# ... so we can make sure there is only 1 matching route
[ 1 -eq `echo "$ROUTE" | wc -l` ] || \
errorexit "more than 1 matching route exists"
# But there might still be no matching route
[ "$OCF_RESKEY_destination" = "0.0.0.0/0" ] && [ -z "$ROUTE" ] && \
! ocf_is_probe && [ "$__OCF_ACTION" != stop ] && errorexit "no matching route exists"
# Sed out the source ip address if it exists
SRCIP=`echo $ROUTE | sed -n "s/$MATCHROUTE/\3/p"`
# and what remains after stripping out the source ip address clause
ROUTE_WO_SRC=`echo $ROUTE | sed "s/$MATCHROUTE/\1\5/"`
# using "src <ip>" only returns output if there's a match
if [ "$OCF_RESKEY_destination" != "0.0.0.0/0" ]; then
[ -z "$ROUTE" ] && return 1 || return 0
fi
[ -z "$SRCIP" ] && return 1
[ $SRCIP = $1 ] && return 0
[ "$__OCF_ACTION" = "monitor" ] || [ "$__OCF_ACTION" = "status" ] && [ "${ROUTE%% *}" = "default" ] && return 1
return 2
}
#
# Add (or change if it already exists) the preferred source address
# The exit code should conform to LSB exit codes.
#
srca_start() {
srca_read $1
rc=$?
if [ $rc = 0 ]; then
rc=$OCF_SUCCESS
ocf_log info "The ip route has been already set.($NETWORK, $INTERFACE, $ROUTE_WO_SRC)"
else
$IP2UTIL route replace $TABLE $NETWORK dev $INTERFACE $PROTO src $1 $METRIC || \
errorexit "command 'ip route replace $TABLE $NETWORK dev $INTERFACE $PROTO src $1 $METRIC' failed"
if [ "$OCF_RESKEY_destination" = "0.0.0.0/0" ] ;then
$CMDCHANGE $ROUTE_WO_SRC src $1 || \
errorexit "command '$CMDCHANGE $ROUTE_WO_SRC src $1' failed"
fi
rc=$?
fi
return $rc
}
#
# Remove (if it exists) the preferred source address.
# If one exists but it's not the same as the one specified, that's
# an error. Maybe that's the wrong behaviour because if this fails
# then when IPaddr releases the associated interface (if there is one)
# your matching route will also get dropped ;-(
# The exit code should conform to LSB exit codes.
#
srca_stop() {
srca_read $1
rc=$?
if [ $rc = 1 ]; then
# We do not have a preferred source address for now
ocf_log info "No preferred source address defined, nothing to stop"
exit $OCF_SUCCESS
fi
[ $rc = 2 ] && errorexit "The address you specified to stop does not match the preferred source address"
if [ -z "$TABLE" ] || [ "${TABLE#table }" = "main" ]; then
SCOPE="link"
else
SCOPE="host"
fi
PRIMARY_IP="$($IP2UTIL -4 -o addr show dev $INTERFACE primary | awk '{split($4,a,"/");print a[1]}')"
OPTS="proto kernel scope $SCOPE src $PRIMARY_IP"
$IP2UTIL route replace $TABLE $NETWORK dev $INTERFACE $OPTS $METRIC || \
errorexit "command 'ip route replace $TABLE $NETWORK dev $INTERFACE $OPTS $METRIC' failed"
if [ "$OCF_RESKEY_destination" = "0.0.0.0/0" ] ;then
$CMDCHANGE $ROUTE_WO_SRC src $PRIMARY_IP || \
errorexit "command '$CMDCHANGE $ROUTE_WO_SRC src $PRIMARY_IP' failed"
fi
return $?
}
srca_status() {
srca_read $1
case $? in
0) echo "OK"
return $OCF_SUCCESS;;
1) echo "No preferred source address defined"
return $OCF_NOT_RUNNING;;
2) echo "Preferred source address has incorrect value"
return $OCF_ERR_GENERIC;;
esac
}
# A not reliable IP address checking function, which only picks up those _obvious_ violations...
#
# It accepts IPv4 address in dotted quad notation, for example "192.168.1.1"
#
# 100% confidence whenever it reports "negative",
# but may get false "positive" answer.
#
CheckIP() {
ip="$1"
case $ip in
*[!0-9.]*) #got invalid char
false;;
.*|*.) #begin or end by ".", which is invalid
false;;
*..*) #consecutive ".", which is invalid
false;;
*.*.*.*.*) #four decimal dots, which is too many
false;;
*.*.*.*) #exactly three decimal dots, candidate, evaluate each field
local IFS=.
set -- $ip
if
( [ $1 -le 254 ] && [ $2 -le 254 ] && [ $3 -le 254 ] && [ $4 -le 254 ] )
then
if [ $1 -eq 127 ]; then
ocf_exit_reason "IP address [$ip] is a loopback address, thus can not be preferred source address"
exit $OCF_ERR_CONFIGURED
fi
else
true
fi
;;
*) #less than three decimal dots
false;;
esac
return $? # This return is unnecessary, this comment too :)
}
#
# Find out which interface or alias serves the given IP address
# The argument is an IP address, and its output
# is an (aliased) interface name (e.g., "eth0" and "eth0:0").
#
find_interface_solaris() {
$IFCONFIG $IFCONFIG_A_OPT | $AWK '{if ($0 ~ /.*: / && NR > 1) {print "\n"$0} else {print}}' |
while read ifname linkstuff
do
: ifname = $ifname
read inet addr junk
: inet = $inet addr = $addr
while
read line && [ "X$line" != "X" ]
do
: Nothing
done
# This doesn't look right for a box with multiple NICs.
# It looks like it always selects the first interface on
# a machine. Yet, we appear to use the results for this case too...
ifname=`echo "$ifname" | sed s'%:*$%%'`
case $addr in
addr:$BASEIP) echo $ifname; return $OCF_SUCCESS;;
$BASEIP) echo $ifname; return $OCF_SUCCESS;;
esac
done
return $OCF_ERR_GENERIC
}
#
# Find out which interface or alias serves the given IP address
# The argument is an IP address, and its output
# is an (aliased) interface name (e.g., "eth0" and "eth0:0").
#
find_interface_generic() {
local iface=`$IP2UTIL -o -f inet addr show | grep "\ $BASEIP" \
| cut -d ' ' -f2 | grep -v '^ipsec[0-9][0-9]*$'`
if [ -z "$iface" ]; then
return $OCF_ERR_GENERIC
else
echo $iface
return $OCF_SUCCESS
fi
}
#
# Find out which interface or alias serves the given IP address
# The argument is an IP address, and its output
# is an (aliased) interface name (e.g., "eth0" and "eth0:0").
#
find_interface() {
case "$SYSTYPE" in
SunOS)
IF=`find_interface_solaris $BASEIP`
;;
*)
IF=`find_interface_generic $BASEIP`
;;
esac
echo $IF
return $OCF_SUCCESS;
}
ip_status() {
BASEIP="$1"
case "$SYSTYPE" in
Darwin)
# Treat Darwin the same as the other BSD variants (matched as *BSD)
SYSTYPE="${SYSTYPE}BSD"
;;
*)
;;
esac
case "$SYSTYPE" in
*BSD)
$IFCONFIG $IFCONFIG_A_OPT | grep "inet.*[: ]$BASEIP " >/dev/null 2>&1
if [ $? = 0 ]; then
return $OCF_SUCCESS
else
return $OCF_NOT_RUNNING
fi;;
Linux|SunOS)
IF=`find_interface "$BASEIP"`
if [ -z "$IF" ]; then
return $OCF_NOT_RUNNING
fi
case $IF in
lo*)
ocf_exit_reason "IP address [$BASEIP] is served by loopback, thus can not be preferred source address"
exit $OCF_ERR_CONFIGURED
;;
*)return $OCF_SUCCESS;;
esac
;;
*)
if [ -z "$IF" ]; then
return $OCF_NOT_RUNNING
else
return $OCF_SUCCESS
fi;;
esac
}
srca_validate_all() {
if [ -z "$OCF_RESKEY_ipaddress" ]; then
# usage
ocf_exit_reason "Please set OCF_RESKEY_ipaddress to the preferred source IP address!"
return $OCF_ERR_CONFIGURED
fi
if ! echo "$OCF_RESKEY_destination" | grep -q "/"; then
return $OCF_ERR_CONFIGURED
fi
if ! [ "x$SYSTYPE" = "xLinux" ]; then
# checks after this point are only relevant for linux.
return $OCF_SUCCESS
fi
check_binary $AWK
case "$SYSTYPE" in
*BSD|SunOS)
check_binary $IFCONFIG
;;
esac
# The IP address should be in good shape
if CheckIP "$ipaddress"; then
:
else
ocf_exit_reason "Invalid IP address [$ipaddress]"
return $OCF_ERR_CONFIGURED
fi
if ocf_is_probe; then
return $OCF_SUCCESS
fi
# We should serve this IP address of course
- if ip_status "$ipaddress"; then
- :
- else
- ocf_exit_reason "We are not serving [$ipaddress], hence can not make it a preferred source address"
- return $OCF_ERR_INSTALLED
+ if [ "$OCF_CHECK_LEVEL" -eq 10 ]; then
+ if ip_status "$ipaddress"; then
+ :
+ else
+ ocf_exit_reason "We are not serving [$ipaddress], hence can not make it a preferred source address"
+ return $OCF_ERR_INSTALLED
+ fi
fi
return $OCF_SUCCESS
}
if
( [ $# -ne 1 ] )
then
usage
exit $OCF_ERR_ARGS
fi
# These operations do not require the OCF instance parameters to be set
case $1 in
meta-data) meta_data
exit $OCF_SUCCESS
;;
usage) usage
exit $OCF_SUCCESS
;;
*)
;;
esac
ipaddress="$OCF_RESKEY_ipaddress"
+[ "$__OCF_ACTION" != "validate-all" ] && OCF_CHECK_LEVEL=10
srca_validate_all
rc=$?
if [ $rc -ne $OCF_SUCCESS ]; then
case $1 in
# if we can't validate the configuration during a stop, that
# means the resources isn't configured correctly. There's no way
# to actually stop the resource in this situation because there's
# no way it could have even started. Return success here
# to indicate that the resource is not running, otherwise the
# stop action will fail causing the node to be fenced just because
# of a mis configuration.
stop) exit $OCF_SUCCESS;;
*) exit $rc;;
esac
fi
findif_out=`$FINDIF`
rc=$?
[ $rc -ne 0 ] && {
ocf_exit_reason "[$FINDIF] failed"
exit $rc
}
INTERFACE=`echo $findif_out | awk '{print $1}'`
LISTROUTE=`$IP2UTIL route list dev $INTERFACE scope link $PROTO match $ipaddress`
[ -z "$PROTO" ] && PROTO=`echo $LISTROUTE | sed -n "s/$PROTOCLAUSE/\1/p"`
if [ -n "$OCF_RESKEY_metric" ]; then
METRIC="metric $OCF_RESKEY_metric"
elif [ -z "$TABLE" ] || [ "${TABLE#table }" = "main" ]; then
METRIC=`echo $LISTROUTE | sed -n "s/$METRICCLAUSE/\1/p"`
else
METRIC=""
fi
if [ "$OCF_RESKEY_destination" = "0.0.0.0/0" ] ;then
NETWORK=`echo $LISTROUTE | grep -m 1 -o '^[^ ]*'`
if [ -z "$NETWORK" ]; then
err_str="command '$IP2UTIL route list dev $INTERFACE scope link $PROTO"
err_str="$err_str match $ipaddress' failed to find a matching route"
if [ "$__OCF_ACTION" = "start" ]; then
ocf_exit_reason "$err_str"
exit $OCF_ERR_ARGS
elif ! ocf_is_probe; then
ocf_log warn "$err_str"
else
ocf_log debug "$err_str"
fi
fi
else
NETWORK="$OCF_RESKEY_destination"
fi
case $1 in
start) srca_start $ipaddress
;;
stop) srca_stop $ipaddress
;;
status) srca_status $ipaddress
;;
monitor) srca_status $ipaddress
;;
validate-all) srca_validate_all
;;
*) usage
exit $OCF_ERR_UNIMPLEMENTED
;;
esac
exit $?
#
# Version 0.3 2002/11/04 17:00:00 John Sutton <john@scl.co.uk>
# Name changed from IPsrcroute to IPsrcaddr and now reports errors
# using ha_log rather than on stderr.
#
# Version 0.2 2002/11/02 17:00:00 John Sutton <john@scl.co.uk>
# Changed status output to "OK" to satisfy ResourceManager's
# we_own_resource() function.
#
# Version 0.1 2002/11/01 17:00:00 John Sutton <john@scl.co.uk>
# First effort but does the job?
#
diff --git a/heartbeat/galera.in b/heartbeat/galera.in
index cd2fee7c0..6aed3e4b6 100755
--- a/heartbeat/galera.in
+++ b/heartbeat/galera.in
@@ -1,1094 +1,1097 @@
#!@BASH_SHELL@
#
# Copyright (c) 2014 David Vossel <davidvossel@gmail.com>
# All Rights Reserved.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of version 2 of the GNU General Public License as
# published by the Free Software Foundation.
#
# This program is distributed in the hope that it would be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
#
# Further, this software is distributed without any warranty that it is
# free of the rightful claim of any third person regarding infringement
# or the like. Any license provided herein, whether implied or
# otherwise, applies only to this software file. Patent licenses, if
# any, provided herein do not apply to combinations of this program with
# other software, or any other product whatsoever.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write the Free Software Foundation,
# Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA.
#
##
# README.
#
# This agent only supports being configured as a multistate Promoted
# resource.
#
# Unpromoted vs Promoted role:
#
# During the 'Unpromoted' role, galera instances are in read-only mode and
# will not attempt to connect to the cluster. This role exists only as
# a means to determine which galera instance is the most up-to-date. The
# most up-to-date node will be used to bootstrap a galera cluster that
# has no current members.
#
# The galera instances will only begin to be promoted to the Promoted role
# once all the nodes in the 'wsrep_cluster_address' connection address
# have entered read-only mode. At that point the node containing the
# database that is most current will be promoted to Promoted. Once the first
# Promoted instance bootstraps the galera cluster, the other nodes will be
# promoted to Promoted as well.
#
# Example: Create a galera cluster using nodes rhel7-node1 rhel7-node2 rhel7-node3
#
# pcs resource create db galera enable_creation=true \
# wsrep_cluster_address="gcomm://rhel7-auto1,rhel7-auto2,rhel7-auto3" meta promoted-max=3 --promoted
#
# By setting the 'enable_creation' option, the database will be automatically
# generated at startup. The meta attribute 'promoted-max=3' means that all 3
# nodes listed in the wsrep_cluster_address list will be allowed to connect
# to the galera cluster and perform replication.
#
# NOTE: If you have more nodes in the pacemaker cluster then you wish
# to have in the galera cluster, make sure to use location contraints to prevent
# pacemaker from attempting to place a galera instance on a node that is
# not in the 'wsrep_cluster_address" list.
#
##
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
if [ "$__OCF_ACTION" != "meta-data" ]; then
. ${OCF_FUNCTIONS_DIR}/mysql-common.sh
NODENAME=$(ocf_attribute_target)
fi
# It is common for some galera instances to store
# check user that can be used to query status
# in this file
if [ -f "/etc/sysconfig/clustercheck" ]; then
. /etc/sysconfig/clustercheck
elif [ -f "/etc/default/clustercheck" ]; then
. /etc/default/clustercheck
fi
# Parameter defaults
OCF_RESKEY_wsrep_cluster_address_default=""
OCF_RESKEY_cluster_host_map_default=""
OCF_RESKEY_check_user_default=""
OCF_RESKEY_check_passwd_default=""
OCF_RESKEY_two_node_mode_default="false"
: ${OCF_RESKEY_wsrep_cluster_address=${OCF_RESKEY_wsrep_cluster_address_default}}
: ${OCF_RESKEY_cluster_host_map=${OCF_RESKEY_cluster_host_map_default}}
: ${OCF_RESKEY_check_user=${OCF_RESKEY_check_user_default}}
: ${OCF_RESKEY_check_passwd=${OCF_RESKEY_check_passwd_default}}
: ${OCF_RESKEY_two_node_mode=${OCF_RESKEY_two_node_mode_default}}
#######################################################################
# Defaults:
OCF_RESKEY_check_passwd_use_empty_default=0
: ${OCF_RESKEY_check_passwd_use_empty=${OCF_RESKEY_check_passwd_use_empty_default}}
#######################################################################
usage() {
cat <<UEND
usage: $0 (start|stop|validate-all|meta-data|monitor|promote|demote)
$0 manages a galera Database as an HA resource.
The 'start' operation starts the database.
The 'stop' operation stops the database.
The 'status' operation reports whether the database is running
The 'monitor' operation reports whether the database seems to be working
The 'promote' operation makes this mysql server run as promoted
The 'demote' operation makes this mysql server run as unpromoted
The 'validate-all' operation reports whether the parameters are valid
UEND
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="galera" version="1.0">
<version>1.0</version>
<longdesc lang="en">
Resource script for managing galera database.
</longdesc>
<shortdesc lang="en">Manages a galera instance</shortdesc>
<parameters>
<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
Location of the MySQL server binary
</longdesc>
<shortdesc lang="en">MySQL server binary</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}" />
</parameter>
<parameter name="client_binary" unique="0" required="0">
<longdesc lang="en">
Location of the MySQL client binary
</longdesc>
<shortdesc lang="en">MySQL client binary</shortdesc>
<content type="string" default="${OCF_RESKEY_client_binary_default}" />
</parameter>
<parameter name="config" unique="0" required="0">
<longdesc lang="en">
Configuration file
</longdesc>
<shortdesc lang="en">MySQL config</shortdesc>
<content type="string" default="${OCF_RESKEY_config_default}" />
</parameter>
<parameter name="datadir" unique="0" required="0">
<longdesc lang="en">
Directory containing databases
</longdesc>
<shortdesc lang="en">MySQL datadir</shortdesc>
<content type="string" default="${OCF_RESKEY_datadir_default}" />
</parameter>
<parameter name="user" unique="0" required="0">
<longdesc lang="en">
User running MySQL daemon
</longdesc>
<shortdesc lang="en">MySQL user</shortdesc>
<content type="string" default="${OCF_RESKEY_user_default}" />
</parameter>
<parameter name="group" unique="0" required="0">
<longdesc lang="en">
Group running MySQL daemon (for logfile and directory permissions)
</longdesc>
<shortdesc lang="en">MySQL group</shortdesc>
<content type="string" default="${OCF_RESKEY_group_default}"/>
</parameter>
<parameter name="log" unique="0" required="0">
<longdesc lang="en">
The logfile to be used for mysqld.
</longdesc>
<shortdesc lang="en">MySQL log file</shortdesc>
<content type="string" default="${OCF_RESKEY_log_default}"/>
</parameter>
<parameter name="pid" unique="0" required="0">
<longdesc lang="en">
The pidfile to be used for mysqld.
</longdesc>
<shortdesc lang="en">MySQL pid file</shortdesc>
<content type="string" default="${OCF_RESKEY_pid_default}"/>
</parameter>
<parameter name="socket" unique="0" required="0">
<longdesc lang="en">
The socket to be used for mysqld.
</longdesc>
<shortdesc lang="en">MySQL socket</shortdesc>
<content type="string" default="${OCF_RESKEY_socket_default}"/>
</parameter>
<parameter name="enable_creation" unique="0" required="0">
<longdesc lang="en">
If the MySQL database does not exist, it will be created
</longdesc>
<shortdesc lang="en">Create the database if it does not exist</shortdesc>
<content type="boolean" default="${OCF_RESKEY_enable_creation_default}"/>
</parameter>
<parameter name="additional_parameters" unique="0" required="0">
<longdesc lang="en">
Additional parameters which are passed to the mysqld on startup.
(e.g. --skip-external-locking or --skip-grant-tables)
</longdesc>
<shortdesc lang="en">Additional parameters to pass to mysqld</shortdesc>
<content type="string" default="${OCF_RESKEY_additional_parameters_default}"/>
</parameter>
<parameter name="wsrep_cluster_address" unique="0" required="1">
<longdesc lang="en">
The galera cluster address. This takes the form of:
gcomm://node,node,node
Only nodes present in this node list will be allowed to start a galera instance.
The galera node names listed in this address are expected to match valid
pacemaker node names. If both names need to differ, you must provide a
mapping in option cluster_host_map.
</longdesc>
<shortdesc lang="en">Galera cluster address</shortdesc>
<content type="string" default="${OCF_RESKEY_wsrep_cluster_address_default}"/>
</parameter>
<parameter name="cluster_host_map" unique="0" required="0">
<longdesc lang="en">
A mapping of pacemaker node names to galera node names.
To be used when both pacemaker and galera names need to differ,
(e.g. when galera names map to IP from a specific network interface)
This takes the form of:
pcmk1:node.1.galera;pcmk2:node.2.galera;pcmk3:node.3.galera
where the galera resource started on node pcmk1 would be named
node.1.galera in the wsrep_cluster_address
</longdesc>
<shortdesc lang="en">Pacemaker to Galera name mapping</shortdesc>
<content type="string" default="${OCF_RESKEY_cluster_host_map_default}"/>
</parameter>
<parameter name="check_user" unique="0" required="0">
<longdesc lang="en">
Cluster check user.
</longdesc>
<shortdesc lang="en">MySQL test user</shortdesc>
<content type="string" default="${OCF_RESKEY_check_user_default}" />
</parameter>
<parameter name="check_passwd" unique="0" required="0">
<longdesc lang="en">
Cluster check user password. Empty passwords are ignored unless
the parameter "check_passwd_use_empty" is set to 1.
</longdesc>
<shortdesc lang="en">check password</shortdesc>
<content type="string" default="${OCF_RESKEY_check_passwd_default}" />
</parameter>
<parameter name="check_passwd_use_empty" unique="0" required="0">
<longdesc lang="en">
Use an empty "check_passwd" password. If this parameter is set to 1,
"check_passwd" will be ignored and an empty password is used
when calling the "mysql" client binary.
</longdesc>
<shortdesc lang="en">check password use empty</shortdesc>
<content type="boolean" default="${OCF_RESKEY_check_passwd_use_empty_default}"/>
</parameter>
<parameter name="two_node_mode" unique="0" required="0">
<longdesc lang="en">
If running in a 2-node pacemaker cluster, rely on pacemaker quorum
to allow automatic recovery even when the other node is unreachable.
Use it with caution! (and fencing)
</longdesc>
<shortdesc lang="en">Special recovery when running on a 2-node cluster</shortdesc>
<content type="boolean" default="${OCF_RESKEY_two_node_mode_default}"/>
</parameter>
</parameters>
<actions>
<action name="start" timeout="120s" />
<action name="stop" timeout="120s" />
<action name="status" timeout="60s" />
<action name="monitor" depth="0" timeout="30s" interval="20s" />
<action name="monitor" role="Promoted" depth="0" timeout="30s" interval="10s" />
<action name="monitor" role="Unpromoted" depth="0" timeout="30s" interval="30s" />
<action name="promote" timeout="300s" />
<action name="demote" timeout="120s" />
<action name="validate-all" timeout="5s" />
<action name="meta-data" timeout="5s" />
</actions>
</resource-agent>
END
}
get_option_variable()
{
local key=$1
$MYSQL $MYSQL_OPTIONS_CHECK -e "SHOW VARIABLES like '$key';" | tail -1
}
get_status_variable()
{
local key=$1
$MYSQL $MYSQL_OPTIONS_CHECK -e "show status like '$key';" | tail -1
}
set_bootstrap_node()
{
local node=$(ocf_attribute_target $1)
${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-bootstrap" -v "true"
}
clear_bootstrap_node()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-bootstrap" -D
}
is_bootstrap()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-bootstrap" --quiet 2>/dev/null
}
set_no_grastate()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -v "true"
}
clear_no_grastate()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -D
}
is_no_grastate()
{
local node=$(ocf_attribute_target $1)
${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" --quiet 2>/dev/null
}
clear_last_commit()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" -D
}
set_last_commit()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" -v $1
}
get_last_commit()
{
local node=$(ocf_attribute_target $1)
if [ -z "$node" ]; then
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" --quiet 2>/dev/null
else
${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" --quiet 2>/dev/null
fi
}
clear_safe_to_bootstrap()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" -D
}
set_safe_to_bootstrap()
{
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" -v $1
}
get_safe_to_bootstrap()
{
local node=$(ocf_attribute_target $1)
if [ -z "$node" ]; then
${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" --quiet 2>/dev/null
else
${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" --quiet 2>/dev/null
fi
}
wait_for_sync()
{
local state=$(get_status_variable "wsrep_local_state")
ocf_log info "Waiting for database to sync with the cluster. "
while [ "$state" != "4" ]; do
sleep 1
state=$(get_status_variable "wsrep_local_state")
done
ocf_log info "Database synced."
}
is_primary()
{
cluster_status=$(get_status_variable "wsrep_cluster_status")
if [ "$cluster_status" = "Primary" ]; then
return 0
fi
if [ -z "$cluster_status" ]; then
ocf_exit_reason "Unable to retrieve wsrep_cluster_status, verify check_user '$OCF_RESKEY_check_user' has permissions to view status"
else
ocf_log info "Galera instance wsrep_cluster_status=${cluster_status}"
fi
return 1
}
is_readonly()
{
local res=$(get_option_variable "read_only")
if ! ocf_is_true "$res"; then
return 1
fi
cluster_status=$(get_status_variable "wsrep_cluster_status")
if ! [ "$cluster_status" = "Disconnected" ]; then
return 1
fi
return 0
}
is_two_node_mode_active()
{
# crm_node or corosync-quorumtool cannot access various corosync
# flags when running inside a bundle, so only count the cluster
# members
ocf_is_true "$OCF_RESKEY_two_node_mode" && crm_mon_no_validation -1X | xmllint --xpath "count(//nodes/node[@type='member'])" - | grep -q -w 2
}
is_last_node_in_quorate_partition()
{
# when a network split occurs in a 2-node cluster, pacemaker
# fences the other node and try to retain quorum. So until
# the fencing is resolved (and the status of the peer node
# is clean), we shouldn't consider ourself quorate.
local partition_members=$(${HA_SBIN_DIR}/crm_node -p | wc -w)
local quorate=$(${HA_SBIN_DIR}/crm_node -q)
local clean_members=$(crm_mon_no_validation -1X | xmllint --xpath 'count(//nodes/node[@type="member" and @unclean="false"])' -)
[ "$partition_members" = 1 ] && [ "$quorate" = 1 ] && [ "$clean_members" = 2 ]
}
master_exists()
{
if [ "$__OCF_ACTION" = "demote" ]; then
# We don't want to detect master instances during demote.
# 1. we could be detecting ourselves as being master, which is no longer the case.
# 2. we could be detecting other master instances that are in the process of shutting down.
# by not detecting other master instances in "demote" we are deferring this check
# to the next recurring monitor operation which will be much more accurate
return 1
fi
# determine if a master instance is already up and is healthy
ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.1.0"
res=$?
if [ -z "$OCF_RESKEY_crm_feature_set" ] || [ $res -eq 2 ]; then
XMLOPT="--output-as=xml"
ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.2.0"
if [ $? -eq 1 ]; then
crm_mon_no_validation -1 $XMLOPT >/dev/null 2>&1
if [ $? -ne 0 ]; then
XMLOPT="--as-xml"
fi
fi
else
XMLOPT="--as-xml"
fi
crm_mon_no_validation -1 $XMLOPT | grep -q -i -E "resource.*id=\"${INSTANCE_ATTR_NAME}\".*role=\"(Promoted|Master)\".*active=\"true\".*orphaned=\"false\".*failed=\"false\""
return $?
}
clear_master_score()
{
local node=$(ocf_attribute_target $1)
if [ -z "$node" ]; then
ocf_promotion_score -D
else
ocf_promotion_score -D -N $node
fi
}
set_master_score()
{
local node=$(ocf_attribute_target $1)
if [ -z "$node" ]; then
ocf_promotion_score -v 100
else
ocf_promotion_score -N $node -v 100
fi
}
promote_everyone()
{
for node in $(echo "$OCF_RESKEY_wsrep_cluster_address" | sed 's/gcomm:\/\///g' | tr -d ' ' | tr -s ',' ' '); do
local pcmk_node=$(galera_to_pcmk_name $node)
if [ -z "$pcmk_node" ]; then
ocf_log err "Could not determine pacemaker node from galera name <${node}>."
return
else
node=$pcmk_node
fi
set_master_score $node
done
}
greater_than_equal_long()
{
# there are values we need to compare in this script
# that are too large for shell -gt to process
echo | awk -v n1="$1" -v n2="$2" '{if (n1>=n2) printf ("true"); else printf ("false");}' | grep -q "true"
}
galera_to_pcmk_name()
{
local galera=$1
if [ -z "$OCF_RESKEY_cluster_host_map" ]; then
echo $galera
else
echo "$OCF_RESKEY_cluster_host_map" | tr ';' '\n' | tr -d ' ' | sed 's/:/ /' | awk -F' ' '$2=="'"$galera"'" {print $1;exit}'
fi
}
pcmk_to_galera_name()
{
local pcmk=$1
if [ -z "$OCF_RESKEY_cluster_host_map" ]; then
echo $pcmk
else
echo "$OCF_RESKEY_cluster_host_map" | tr ';' '\n' | tr -d ' ' | sed 's/:/ /' | awk -F' ' '$1=="'"$pcmk"'" {print $2;exit}'
fi
}
detect_first_master()
{
local best_commit=0
local last_commit=0
local missing_nodes=0
local nodes=""
local nodes_recovered=""
local all_nodes
local best_node_gcomm
local best_node
local safe_to_bootstrap
all_nodes=$(echo "$OCF_RESKEY_wsrep_cluster_address" | sed 's/gcomm:\/\///g' | tr -d ' ' | tr -s ',' ' ')
best_node_gcomm=$(echo "$all_nodes" | sed 's/^.* \(.*\)$/\1/')
best_node=$(galera_to_pcmk_name $best_node_gcomm)
if [ -z "$best_node" ]; then
ocf_log err "Could not determine initial best node from galera name <${best_node_gcomm}>."
return
fi
# avoid selecting a recovered node as bootstrap if possible
for node in $all_nodes; do
local pcmk_node=$(galera_to_pcmk_name $node)
if [ -z "$pcmk_node" ]; then
ocf_log err "Could not determine pacemaker node from galera name <${node}>."
return
else
node=$pcmk_node
fi
if is_no_grastate $node; then
nodes_recovered="$nodes_recovered $node"
else
nodes="$nodes $node"
fi
done
for node in $nodes_recovered $nodes; do
# On clean shutdown, galera sets the last stopped node as 'safe to bootstrap',
# so use this hint when we can
safe_to_bootstrap=$(get_safe_to_bootstrap $node)
# Special case for 2-node clusters: during a network split, rely on
# pacemaker's quorum to check whether we can restart galera
if [ "$safe_to_bootstrap" != "1" ] && [ "$node" = "$NODENAME" ] && is_two_node_mode_active; then
is_last_node_in_quorate_partition
if [ $? -eq 0 ]; then
ocf_log warn "Survived a split in a 2-node cluster, considering ourselves safe to bootstrap"
safe_to_bootstrap=1
fi
fi
if [ "$safe_to_bootstrap" = "1" ]; then
# Galera marked the node as safe to boostrap during shutdown. Let's just
# pick it as our bootstrap node.
ocf_log info "Node <${node}> is marked as safe to bootstrap."
best_node=$node
# We don't need to wait for the other nodes to report state in this case
missing_nodes=0
break
fi
last_commit=$(get_last_commit $node)
if [ -z "$last_commit" ]; then
ocf_log info "Waiting on node <${node}> to report database status before Master instances can start."
missing_nodes=1
continue
fi
# this means -1, or that no commit has occured yet.
if [ "$last_commit" = "18446744073709551615" ]; then
last_commit="0"
fi
greater_than_equal_long "$last_commit" "$best_commit"
if [ $? -eq 0 ]; then
best_node=$(ocf_attribute_target $node)
best_commit=$last_commit
fi
done
if [ $missing_nodes -eq 1 ]; then
return
fi
ocf_log info "Promoting $best_node to be our bootstrap node"
set_bootstrap_node $best_node
set_master_score $best_node
}
detect_safe_to_bootstrap()
{
local safe_to_bootstrap=""
local uuid=""
local seqno=""
if [ -f ${OCF_RESKEY_datadir}/grastate.dat ]; then
ocf_log info "attempting to read safe_to_bootstrap flag from ${OCF_RESKEY_datadir}/grastate.dat"
safe_to_bootstrap=$(sed -n 's/^safe_to_bootstrap:\s*\(.*\)$/\1/p' < ${OCF_RESKEY_datadir}/grastate.dat)
uuid=$(sed -n 's/^uuid:\s*\(.*\)$/\1/p' < ${OCF_RESKEY_datadir}/grastate.dat)
seqno=$(sed -n 's/^seqno:\s*\(.*\)$/\1/p' < ${OCF_RESKEY_datadir}/grastate.dat)
fi
if [ -z "$uuid" ] || \
[ "$uuid" = "00000000-0000-0000-0000-000000000000" ]; then
clear_safe_to_bootstrap
return
fi
if [ "$safe_to_bootstrap" = "1" ]; then
if [ -z "$seqno" ] || [ "$seqno" = "-1" ]; then
clear_safe_to_bootstrap
return
fi
fi
if [ "$safe_to_bootstrap" = "1" ] || [ "$safe_to_bootstrap" = "0" ]; then
set_safe_to_bootstrap $safe_to_bootstrap
else
clear_safe_to_bootstrap
fi
}
detect_last_commit()
{
local last_commit
local recover_args="--defaults-file=$OCF_RESKEY_config \
--pid-file=$OCF_RESKEY_pid \
--socket=$OCF_RESKEY_socket \
--datadir=$OCF_RESKEY_datadir"
local recovery_file_regex='s/.*WSREP\:.*position\s*recovery.*--log_error='\''\([^'\'']*\)'\''.*/\1/p'
local recovered_position_regex='s/.*WSREP\:\s*[R|r]ecovered\s*position.*\:\(.*\)\s*$/\1/p'
# codership/galera#354
# Some ungraceful shutdowns can leave an empty gvwstate.dat on
# disk. This will prevent galera to join the cluster if it is
# configured to attempt PC recovery. Removing that file makes the
# node fall back to the normal, unoptimized joining process.
if [ -f ${OCF_RESKEY_datadir}/gvwstate.dat ] && \
[ ! -s ${OCF_RESKEY_datadir}/gvwstate.dat ]; then
ocf_log warn "empty ${OCF_RESKEY_datadir}/gvwstate.dat detected, removing it to prevent PC recovery failure at next restart"
rm -f ${OCF_RESKEY_datadir}/gvwstate.dat
fi
ocf_log info "attempting to detect last commit version by reading ${OCF_RESKEY_datadir}/grastate.dat"
last_commit="$(cat ${OCF_RESKEY_datadir}/grastate.dat | sed -n 's/^seqno.\s*\(.*\)\s*$/\1/p')"
if [ -z "$last_commit" ] || [ "$last_commit" = "-1" ]; then
local tmp=$(mktemp)
chown $OCF_RESKEY_user:$OCF_RESKEY_group $tmp
# if we pass here because grastate.dat doesn't exist,
# try not to bootstrap from this node if possible
if [ ! -f ${OCF_RESKEY_datadir}/grastate.dat ]; then
set_no_grastate
fi
ocf_log info "now attempting to detect last commit version using 'mysqld_safe --wsrep-recover'"
$SU - $OCF_RESKEY_user -s /bin/sh -c \
"${OCF_RESKEY_binary} $recover_args --wsrep-recover --log-error=$tmp 2>/dev/null"
last_commit="$(cat $tmp | sed -n $recovered_position_regex | tail -1)"
if [ -z "$last_commit" ]; then
# Galera uses InnoDB's 2pc transactions internally. If
# server was stopped in the middle of a replication, the
# recovery may find a "prepared" XA transaction in the
# redo log, and mysql won't recover automatically
local recovery_file="$(cat $tmp | sed -n $recovery_file_regex)"
if [ -e $recovery_file ]; then
cat $recovery_file | grep -q -E '\[ERROR\]\s+Found\s+[0-9]+\s+prepared\s+transactions!' 2>/dev/null
if [ $? -eq 0 ]; then
# we can only rollback the transaction, but that's OK
# since the DB will get resynchronized anyway
ocf_log warn "local node <${NODENAME}> was not shutdown properly. Rollback stuck transaction with --tc-heuristic-recover"
$SU - $OCF_RESKEY_user -s /bin/sh -c \
"${OCF_RESKEY_binary} $recover_args --wsrep-recover \
--tc-heuristic-recover=rollback --log-error=$tmp 2>/dev/null"
last_commit="$(cat $tmp | sed -n $recovered_position_regex | tail -1)"
if [ ! -z "$last_commit" ]; then
ocf_log warn "State recovered. force SST at next restart for full resynchronization"
rm -f ${OCF_RESKEY_datadir}/grastate.dat
# try not to bootstrap from this node if possible
set_no_grastate
fi
fi
fi
fi
rm -f $tmp
fi
if [ ! -z "$last_commit" ]; then
ocf_log info "Last commit version found: $last_commit"
set_last_commit $last_commit
return $OCF_SUCCESS
else
ocf_exit_reason "Unable to detect last known write sequence number"
clear_last_commit
return $OCF_ERR_GENERIC
fi
}
# For galera, promote is really start
galera_promote()
{
local rc
local extra_opts
local bootstrap
local safe_to_bootstrap
master_exists
if [ $? -eq 0 ]; then
# join without bootstrapping
extra_opts="--wsrep-cluster-address=${OCF_RESKEY_wsrep_cluster_address}"
else
bootstrap=$(is_bootstrap)
if ocf_is_true $bootstrap; then
# The best node for bootstrapping wasn't cleanly shutdown. Allow
# bootstrapping anyways
if [ "$(get_safe_to_bootstrap)" = "0" ]; then
sed -ie 's/^\(safe_to_bootstrap:\) 0/\1 1/' ${OCF_RESKEY_datadir}/grastate.dat
ocf_log info "safe_to_bootstrap in ${OCF_RESKEY_datadir}/grastate.dat set to 1 on node ${NODENAME}"
fi
ocf_log info "Node <${NODENAME}> is bootstrapping the cluster"
extra_opts="--wsrep-cluster-address=gcomm://"
else
# We are being promoted without having the bootstrap
# attribute in the CIB, which means we are supposed to
# join a cluster; however if we end up here, there is no
# Master remaining right now, which means there is no
# cluster to join anymore. So force a demotion, and and
# let the RA decide later which node should be the next
# bootstrap node.
ocf_log warn "There is no running cluster to join, demoting ourself"
clear_master_score
return $OCF_SUCCESS
fi
fi
galera_monitor
if [ $? -eq $OCF_RUNNING_MASTER ]; then
if ocf_is_true $bootstrap; then
promote_everyone
clear_bootstrap_node
ocf_log info "boostrap node already up, promoting the rest of the galera instances."
fi
clear_safe_to_bootstrap
clear_last_commit
return $OCF_SUCCESS
fi
# last commit/safe_to_bootstrap flag are no longer relevant once promoted
clear_last_commit
clear_safe_to_bootstrap
mysql_common_prepare_dirs
mysql_common_start "$extra_opts"
rc=$?
if [ $rc != $OCF_SUCCESS ]; then
return $rc
fi
galera_monitor
rc=$?
if [ $rc != $OCF_SUCCESS -a $rc != $OCF_RUNNING_MASTER ]; then
ocf_exit_reason "Failed initial monitor action"
return $rc
fi
is_readonly
if [ $? -eq 0 ]; then
ocf_exit_reason "Failure. Master instance started in read-only mode, check configuration."
return $OCF_ERR_GENERIC
fi
is_primary
if [ $? -ne 0 ]; then
ocf_exit_reason "Failure. Master instance started, but is not in Primary mode."
return $OCF_ERR_GENERIC
fi
if ocf_is_true $bootstrap; then
promote_everyone
clear_bootstrap_node
# clear attribute no-grastate. if last shutdown was
# not clean, we cannot be extra-cautious by requesting a SST
# since this is the bootstrap node
clear_no_grastate
ocf_log info "Bootstrap complete, promoting the rest of the galera instances."
else
# if this is not the bootstrap node, make sure this instance
# syncs with the rest of the cluster before promotion returns.
wait_for_sync
# sync is done, clear info about last startup
clear_no_grastate
fi
ocf_log info "Galera started"
return $OCF_SUCCESS
}
galera_demote()
{
mysql_common_stop
rc=$?
if [ $rc -ne $OCF_SUCCESS ] && [ $rc -ne $OCF_NOT_RUNNING ]; then
ocf_exit_reason "Failed to stop Master galera instance during demotion to Master"
return $rc
fi
# if this node was previously a bootstrap node, that is no longer the case.
clear_bootstrap_node
clear_last_commit
clear_no_grastate
clear_safe_to_bootstrap
# Clear master score here rather than letting pacemaker do so once
# demote finishes. This way a promote cannot take place right
# after this demote even if pacemaker is requested to do so. It
# will first have to run a start/monitor op, to reprobe the state
# of the other galera nodes and act accordingly.
clear_master_score
# record last commit for next promotion
detect_safe_to_bootstrap
detect_last_commit
rc=$?
return $rc
}
galera_start()
{
local rc
local galera_node
galera_node=$(pcmk_to_galera_name $NODENAME)
if [ -z "$galera_node" ]; then
ocf_exit_reason "Could not determine galera name from pacemaker node <${NODENAME}>."
return $OCF_ERR_CONFIGURED
fi
echo $OCF_RESKEY_wsrep_cluster_address | grep -q -F $galera_node
if [ $? -ne 0 ]; then
ocf_exit_reason "local node <${NODENAME}> (galera node <${galera_node}>) must be a member of the wsrep_cluster_address <${OCF_RESKEY_wsrep_cluster_address}> to start this galera instance"
return $OCF_ERR_CONFIGURED
fi
galera_monitor
if [ $? -eq $OCF_RUNNING_MASTER ]; then
ocf_exit_reason "master galera instance started outside of the cluster's control"
return $OCF_ERR_GENERIC
fi
mysql_common_prepare_dirs
detect_safe_to_bootstrap
detect_last_commit
rc=$?
if [ $rc -ne $OCF_SUCCESS ]; then
return $rc
fi
master_exists
if [ $? -eq 0 ]; then
ocf_log info "Master instances are already up, setting master score so this instance will join galera cluster."
set_master_score $NODENAME
else
clear_master_score
detect_first_master
fi
return $OCF_SUCCESS
}
galera_monitor()
{
local rc
local galera_node
local status_loglevel="err"
# Set loglevel to info during probe
if ocf_is_probe; then
status_loglevel="info"
fi
mysql_common_status $status_loglevel
rc=$?
if [ $rc -eq $OCF_NOT_RUNNING ]; then
last_commit=$(get_last_commit $node)
if [ -n "$last_commit" ]; then
# if last commit is set, this instance is considered started in slave mode
rc=$OCF_SUCCESS
master_exists
if [ $? -ne 0 ]; then
detect_first_master
else
# a master instance exists and is healthy, promote this
# local read only instance
# so it can join the master galera cluster.
set_master_score
fi
fi
return $rc
elif [ $rc -ne $OCF_SUCCESS ]; then
return $rc
fi
# if we make it here, mysql is running. Check cluster status now.
galera_node=$(pcmk_to_galera_name $NODENAME)
if [ -z "$galera_node" ]; then
ocf_exit_reason "Could not determine galera name from pacemaker node <${NODENAME}>."
return $OCF_ERR_CONFIGURED
fi
echo $OCF_RESKEY_wsrep_cluster_address | grep -q -F $galera_node
if [ $? -ne 0 ]; then
ocf_exit_reason "local node <${NODENAME}> (galera node <${galera_node}>) is started, but is not a member of the wsrep_cluster_address <${OCF_RESKEY_wsrep_cluster_address}>"
return $OCF_ERR_GENERIC
fi
is_primary
if [ $? -eq 0 ]; then
if ocf_is_probe; then
# restore master score during probe
# if we detect this is a master instance
set_master_score
fi
rc=$OCF_RUNNING_MASTER
else
ocf_exit_reason "local node <${NODENAME}> is started, but not in primary mode. Unknown state."
rc=$OCF_ERR_GENERIC
fi
return $rc
}
galera_stop()
{
local rc
# make sure the process is stopped
mysql_common_stop
rc=$1
clear_safe_to_bootstrap
clear_last_commit
clear_master_score
clear_bootstrap_node
clear_no_grastate
return $rc
}
galera_validate()
{
- if ! ocf_is_ms; then
- ocf_exit_reason "Galera must be configured as a multistate Master/Slave resource."
- return $OCF_ERR_CONFIGURED
+ if [ "$OCF_CHECK_LEVEL" -eq 10 ]; then
+ if ! ocf_is_ms; then
+ ocf_exit_reason "Galera must be configured as a multistate Master/Slave resource."
+ return $OCF_ERR_CONFIGURED
+ fi
fi
if [ -z "$OCF_RESKEY_wsrep_cluster_address" ]; then
ocf_exit_reason "Galera must be configured with a wsrep_cluster_address value."
return $OCF_ERR_CONFIGURED
fi
mysql_common_validate
}
case "$1" in
meta-data) meta_data
exit $OCF_SUCCESS;;
usage|help) usage
exit $OCF_SUCCESS;;
esac
+[ "$__OCF_ACTION" = "start" ] && OCF_CHECK_LEVEL=10
galera_validate
rc=$?
LSB_STATUS_STOPPED=3
if [ $rc -ne 0 ]; then
case "$1" in
stop) exit $OCF_SUCCESS;;
monitor) exit $OCF_NOT_RUNNING;;
status) exit $LSB_STATUS_STOPPED;;
*) exit $rc;;
esac
fi
if [ -z "${OCF_RESKEY_check_passwd}" ]; then
# This value is automatically sourced from /etc/sysconfig/checkcluster if available
OCF_RESKEY_check_passwd=${MYSQL_PASSWORD}
fi
if [ -z "${OCF_RESKEY_check_user}" ]; then
# This value is automatically sourced from /etc/sysconfig/checkcluster if available
OCF_RESKEY_check_user=${MYSQL_USERNAME}
fi
: ${OCF_RESKEY_check_user="root"}
MYSQL_OPTIONS_CHECK="-nNE --user=${OCF_RESKEY_check_user}"
if ocf_is_true "${OCF_RESKEY_check_passwd_use_empty}"; then
MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK --password="
elif [ -n "${OCF_RESKEY_check_passwd}" ]; then
MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK --password=${OCF_RESKEY_check_passwd}"
fi
# This value is automatically sourced from /etc/sysconfig/checkcluster if available
if [ -n "${MYSQL_HOST}" ]; then
MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK -h ${MYSQL_HOST}"
fi
# This value is automatically sourced from /etc/sysconfig/checkcluster if available
if [ -n "${MYSQL_PORT}" ]; then
MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK -P ${MYSQL_PORT}"
fi
# What kind of method was invoked?
case "$1" in
start) galera_start;;
stop) galera_stop;;
status) mysql_common_status err;;
monitor) galera_monitor;;
promote) galera_promote;;
demote) galera_demote;;
validate-all) exit $OCF_SUCCESS;;
*) usage
exit $OCF_ERR_UNIMPLEMENTED;;
esac
# vi:sw=4:ts=4:et:
diff --git a/heartbeat/mpathpersist.in b/heartbeat/mpathpersist.in
index 0e2c2a4a0..8a46b9930 100644
--- a/heartbeat/mpathpersist.in
+++ b/heartbeat/mpathpersist.in
@@ -1,681 +1,686 @@
#!@BASH_SHELL@
#
#
# OCF Resource Agent compliant PERSISTENT SCSI RESERVATION on multipath devices resource script.
# Testversion for a mpathpersist implementation for demo purposes by Andreas Thomas
#
# Copyright (c) 2017 Evgeny Nifontov, lwang@suse.com,
# Andreas Tomas<Andreas.Tomas@suse.com>,
# Zhu Lingshan<lszhu@suse.com>
# All Rights Reserved.
#
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of version 2 of the GNU General Public License as
# published by the Free Software Foundation.
#
# This program is distributed in the hope that it would be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
#
# Further, this software is distributed without any warranty that it is
# free of the rightful claim of any third person regarding infringement
# or the like. Any license provided herein, whether implied or
# otherwise, applies only to this software file. Patent licenses, if
# any, provided herein do not apply to combinations of this program with
# other software, or any other product whatsoever.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write the Free Software Foundation,
# Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA.
#
#
# OCF instance parameters
# OCF_RESKEY_binary
# OCF_RESKEY_devs
# OCF_RESKEY_required_devs_no
# OCF_RESKEY_reservation_type
# OCF_RESKEY_master_score_base
# OCF_RESKEY_master_score_dev_factor
# OCF_RESKEY_master_score_delay
#
# TODO
#
# 1) PROBLEM: devices which were not accessible during 'start' action, will be never registered/reserved
# TODO: 'Master' and 'Slave' registers new devs in 'monitor' action
# TODO: 'Master' reserves new devs in 'monitor' action
#Defaults
OCF_RESKEY_mpathpersist_binary_default="mpathpersist"
OCF_RESKEY_required_devs_no_default=1
OCF_RESKEY_reservation_type_default=1
OCF_RESKEY_master_score_base_default=0
OCF_RESKEY_master_score_dev_factor_default=100
OCF_RESKEY_master_score_delay_default=30
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
# set default values
: ${OCF_RESKEY_mpathpersist_binary=${OCF_RESKEY_mpathpersist_binary_default}} # binary name for the resource
: ${OCF_RESKEY_required_devs_no=${OCF_RESKEY_required_devs_no_default}} # number of required devices
: ${OCF_RESKEY_reservation_type=${OCF_RESKEY_reservation_type_default}} # reservation type
: ${OCF_RESKEY_master_score_base=${OCF_RESKEY_master_score_base_default}} # master score base
: ${OCF_RESKEY_master_score_dev_factor=${OCF_RESKEY_master_score_dev_factor_default}} # device factor for master score
: ${OCF_RESKEY_master_score_delay=${OCF_RESKEY_master_score_delay_default}} # delay for master score
#######################################################################
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="mpathpersist" version="1.1">
<version>1.0</version>
<longdesc lang="en">
This resource agent manages SCSI persistent reservations on multipath devices.
"mpathpersist" from multipath-tools is used, please see its documentation.
Should be used as multistate (Promotable) resource
Unpromoted registers its node id ("crm_node -i") as reservation key ( --param-sark ) on each device in the params "devs" list.
Promoted reserves all devices from params "devs" list with reservation "--prout-type" value from "reservation_type" parameter.
Please see man sg_persist(8) and mpathpersist(8) for reservation_type details.
</longdesc>
<shortdesc lang="en">Manages SCSI persistent reservations on multipath devices</shortdesc>
<parameters>
<parameter name="binary" unique="0">
<longdesc lang="en">
The name of the binary that manages the resource.
</longdesc>
<shortdesc lang="en">the binary name of the resource</shortdesc>
<content type="string" default="${OCF_RESKEY_mpathpersist_binary_default}"/>
</parameter>
<parameter name="devs" unique="0" required="1">
<longdesc lang="en">
Device list. Multiple devices can be listed with blank space as separator.
Shell wildcards are allowed.
</longdesc>
<shortdesc lang="en">device list</shortdesc>
<content type="string"/>
</parameter>
<parameter name="required_devs_no" unique="0" required="0">
<longdesc lang="en">
Minimum number of "working" devices from device list
1) existing
2) "mpathpersist --in --read-keys &lt;device&gt;" works (Return code 0)
resource actions "start","monitor","promote" and "validate-all" return "OCF_ERR_INSTALLED"
if the actual number of "working" devices is less than "required_devs_no".
resource actions "stop" and "demote" tries to remove reservations and registration keys from
all working devices, but always return "OCF_SUCCESS"
</longdesc>
<shortdesc lang="en">minimum number of working devices</shortdesc>
<content type="string" default="${OCF_RESKEY_required_devs_no_default}"/>
</parameter>
<parameter name="reservation_type" unique="0" required="0">
<longdesc lang="en">
reservation type
</longdesc>
<shortdesc lang="en">reservation type</shortdesc>
<content type="string" default="${OCF_RESKEY_reservation_type_default}" />
</parameter>
<parameter name="master_score_base" unique="0" required="0">
<longdesc lang="en">
master_score_base value
"master_score_base" value is used in "master_score" calculation:
master_score = master_score_base + master_score_dev_factor * working_devs
if set to bigger value in mpathpersist resource configuration on some node, this node will be "preferred" for promoted role.
</longdesc>
<shortdesc lang="en">base master_score value</shortdesc>
<content type="string" default="${OCF_RESKEY_master_score_base_default}" />
</parameter>
<parameter name="master_score_dev_factor" unique="0" required="0">
<longdesc lang="en">
Working device factor in promoted calculation
each "working" device provides additional value to "master_score",
so the node that sees more devices will be preferred for the "Promoted"-role
Setting it to 0 will disable this behavior.
</longdesc>
<shortdesc lang="en">working device factor in master_score calculation</shortdesc>
<content type="string" default="${OCF_RESKEY_master_score_dev_factor_default}" />
</parameter>
<parameter name="master_score_delay" unique="0" required="0">
<longdesc lang="en">
promoted/unpromoted decreases/increases its master_score after delay of "master_score_delay" seconds
so if some device gets inaccessible, the unpromoted decreases its promoted first and the resource will no be watched
and after this device reappears again the promoted increases its master_score first
this can work only if the master_score_delay is bigger then monitor interval on both promoted and unpromoted
Setting it to 0 will disable this behavior.
</longdesc>
<shortdesc lang="en">master_score decrease/increase delay time</shortdesc>
<content type="string" default="${OCF_RESKEY_master_score_delay_default}" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="30s" />
<action name="promote" timeout="30s" />
<action name="demote" timeout="30s" />
<action name="notify" timeout="30s" />
<action name="stop" timeout="30s" />
<action name="monitor" depth="0" timeout="20s" interval="29s" role="Unpromoted" />
<action name="monitor" depth="0" timeout="20s" interval="60s" role="Promoted" />
<action name="meta-data" timeout="5s" />
<action name="validate-all" timeout="30s" />
</actions>
</resource-agent>
END
exit $OCF_SUCCESS
}
mpathpersist_init() {
if ! ocf_is_root ; then
ocf_log err "You must be root to perform this operation."
exit $OCF_ERR_PERM
fi
MPATHPERSIST="${OCF_RESKEY_mpathpersist_binary}"
check_binary $MPATHPERSIST
ROLE=$OCF_RESKEY_CRM_meta_role
NOW=$(date +%s)
RESOURCE="${OCF_RESOURCE_INSTANCE}"
MASTER_SCORE_VAR_NAME="master-${OCF_RESOURCE_INSTANCE}"
PENDING_VAR_NAME="pending-$MASTER_SCORE_VAR_NAME"
#only works with corocync
CRM_NODE="${HA_SBIN_DIR}/crm_node"
NODE_ID_DEC=$($CRM_NODE -i)
NODE=$($CRM_NODE -l | $GREP -w ^$NODE_ID_DEC)
NODE=${NODE#$NODE_ID_DEC }
NODE=${NODE% *}
MASTER_SCORE_ATTRIBUTE="${HA_SBIN_DIR}/crm_attribute --lifetime=reboot --name=$MASTER_SCORE_VAR_NAME --node=$NODE"
PENDING_ATTRIBUTE="${HA_SBIN_DIR}/crm_attribute --lifetime=reboot --name=$PENDING_VAR_NAME --node=$NODE"
NODE_ID_HEX=$(printf '0x%x' $NODE_ID_DEC)
if [ -z "$NODE_ID_HEX" ]; then
ocf_log err "Couldn't get node id with \"$CRM_NODE\""
exit $OCF_ERR_INSTALLED
fi
ocf_log debug "$RESOURCE: NODE:$NODE, ROLE:$ROLE, NODE_ID DEC:$NODE_ID_DEC HEX:$NODE_ID_HEX"
DEVS="${OCF_RESKEY_devs}"
REQUIRED_DEVS_NO="${OCF_RESKEY_required_devs_no}"
RESERVATION_TYPE="${OCF_RESKEY_reservation_type}"
MASTER_SCORE_BASE="${OCF_RESKEY_master_score_base}"
MASTER_SCORE_DEV_FACTOR="${OCF_RESKEY_master_score_dev_factor}"
MASTER_SCORE_DELAY="${OCF_RESKEY_master_score_delay}"
ocf_log debug "$RESOURCE: DEVS=$DEVS"
ocf_log debug "$RESOURCE: REQUIRED_DEVS_NO=$REQUIRED_DEVS_NO"
ocf_log debug "$RESOURCE: RESERVATION_TYPE=$RESERVATION_TYPE"
ocf_log debug "$RESOURCE: MASTER_SCORE_BASE=$MASTER_SCORE_BASE"
ocf_log debug "$RESOURCE: MASTER_SCORE_DEV_FACTOR=$MASTER_SCORE_DEV_FACTOR"
ocf_log debug "$RESOURCE: MASTER_SCORE_DELAY=$MASTER_SCORE_DELAY"
#expand path wildcards
DEVS=$(echo $DEVS)
if [ -z "$DEVS" ]; then
ocf_log err "\"devs\" not defined"
exit $OCF_ERR_INSTALLED
fi
mpathpersist_check_devs
mpathpersist_get_status
}
mpathpersist_action_usage() {
cat <<END
usage: $0 {start|stop|monitor|validate-all|promote|demote|notify|meta-data}
Expects to have a fully populated OCF RA-compliant environment set.
END
}
mpathpersist_get_status() {
unset WORKING_DEVS[*]
for dev in ${EXISTING_DEVS[*]}
do
READ_KEYS=`$MPATHPERSIST --in --read-keys $dev 2>&1`
if [ $? -eq 0 ]; then
WORKING_DEVS+=($dev)
echo "$READ_KEYS" | $GREP -w $NODE_ID_HEX\$ >/dev/null
if [ $? -eq 0 ]; then
REGISTERED_DEVS+=($dev)
READ_RESERVATION=`$MPATHPERSIST --in --read-reservation $dev 2>&1`
if [ $? -eq 0 ]; then
echo "$READ_RESERVATION" | $GREP -w $NODE_ID_HEX\$ >/dev/null
if [ $? -eq 0 ]; then
RESERVED_DEVS+=($dev)
fi
reservation_key=`echo $READ_RESERVATION | $GREP -o 'Key = 0x[0-9a-f]*' | $GREP -o '0x[0-9a-f]*'`
if [ -n "$reservation_key" ]; then
DEVS_WITH_RESERVATION+=($dev)
RESERVATION_KEYS+=($reservation_key)
fi
fi
fi
fi
done
WORKING_DEVS_NO=${#WORKING_DEVS[*]}
ocf_log debug "$RESOURCE: working devices: `mpathpersist_echo_array ${WORKING_DEVS[*]}`"
ocf_log debug "$RESOURCE: number of working devices: $WORKING_DEVS_NO"
ocf_log debug "$RESOURCE: registered devices: `mpathpersist_echo_array ${REGISTERED_DEVS[*]}`"
ocf_log debug "$RESOURCE: reserved devices: `mpathpersist_echo_array ${RESERVED_DEVS[*]}`"
ocf_log debug "$RESOURCE: devices with reservation: `mpathpersist_echo_array ${DEVS_WITH_RESERVATION[*]}`"
ocf_log debug "$RESOURCE: reservation keys: `mpathpersist_echo_array ${RESERVATION_KEYS[*]}`"
MASTER_SCORE=$(($MASTER_SCORE_BASE + $MASTER_SCORE_DEV_FACTOR*$WORKING_DEVS_NO))
ocf_log debug "$RESOURCE: master_score: $MASTER_SCORE_BASE + $MASTER_SCORE_DEV_FACTOR*$WORKING_DEVS_NO = $MASTER_SCORE"
}
mpathpersist_check_devs() {
for dev in $DEVS
do
if [ -e "$dev" ]; then
EXISTING_DEVS+=($dev)
fi
done
EXISTING_DEVS_NO=${#EXISTING_DEVS[*]}
if [ $EXISTING_DEVS_NO -lt $REQUIRED_DEVS_NO ]; then
ocf_log err "Number of existing devices=$EXISTING_DEVS_NO less then required_devs_no=$REQUIRED_DEVS_NO"
exit $OCF_ERR_INSTALLED
fi
}
mpathpersist_is_registered() {
for registered_dev in ${REGISTERED_DEVS[*]}
do
if [ "$registered_dev" == "$1" ]; then
return 0
fi
done
return 1
}
mpathpersist_get_reservation_key() {
for array_index in ${!DEVS_WITH_RESERVATION[*]}
do
if [ "${DEVS_WITH_RESERVATION[$array_index]}" == "$1" ]; then
echo ${RESERVATION_KEYS[$array_index]}
return 0
fi
done
echo ""
}
mpathpersist_echo_array() {
str_count=0
arr_str=""
for str in "$@"
do
arr_str="$arr_str[$str_count]:$str "
str_count=$(($str_count+1))
done
echo $arr_str
}
mpathpersist_parse_act_pending() {
ACT_PENDING_TS=0
ACT_PENDING_SCORE=0
if [ -n "$ACT_PENDING" ]; then
ACT_PENDING_TS=${ACT_PENDING%%_*}
ACT_PENDING_SCORE=${ACT_PENDING##*_}
fi
}
mpathpersist_clear_pending() {
if [ -n "$ACT_PENDING" ]; then
DO_PENDING_UPDATE="YES"
NEW_PENDING=""
fi
}
mpathpersist_new_master_score() {
DO_MASTER_SCORE_UPDATE="YES"
NEW_MASTER_SCORE=$1
}
mpathpersist_new_pending() {
DO_PENDING_UPDATE="YES"
NEW_PENDING=$1
}
# Functions invoked by resource manager actions
mpathpersist_action_start() {
ocf_run $MASTER_SCORE_ATTRIBUTE --update=$MASTER_SCORE
ocf_run $PENDING_ATTRIBUTE --update=""
if [ $WORKING_DEVS_NO -lt $REQUIRED_DEVS_NO ]; then
ocf_log err "$RESOURCE: Number of working devices=$WORKING_DEVS_NO less then required_devs_no=$REQUIRED_DEVS_NO"
exit $OCF_ERR_GENERIC
fi
for dev in ${WORKING_DEVS[*]}
do
if mpathpersist_is_registered $dev ; then
: OK
else
ocf_run $MPATHPERSIST --out --register --param-sark=$NODE_ID_HEX $dev
if [ $? -ne $OCF_SUCCESS ]
then
return $OCF_ERR_GENERIC
fi
fi
done
return $OCF_SUCCESS
}
mpathpersist_action_stop() {
if [ ${#REGISTERED_DEVS[*]} -eq 0 ]; then
ocf_log debug "$RESOURCE stop: already no registrations"
else
# Clear preference for becoming master
ocf_run $MASTER_SCORE_ATTRIBUTE --delete
ocf_run $PENDING_ATTRIBUTE --delete
for dev in ${REGISTERED_DEVS[*]}
do
ocf_run $MPATHPERSIST --out --register --param-rk=$NODE_ID_HEX $dev
done
fi
return $OCF_SUCCESS
}
mpathpersist_action_monitor() {
ACT_MASTER_SCORE=`$MASTER_SCORE_ATTRIBUTE --query --quiet 2>&1`
ocf_log debug "$RESOURCE monitor: ACT_MASTER_SCORE=$ACT_MASTER_SCORE"
ACT_PENDING=`$PENDING_ATTRIBUTE --query --quiet 2>&1`
ocf_log debug "$RESOURCE monitor: ACT_PENDING=$ACT_PENDING"
mpathpersist_parse_act_pending
ocf_log debug "$RESOURCE monitor: ACT_PENDING_TS=$ACT_PENDING_TS"
ocf_log debug "$RESOURCE monitor: ACT_PENDING_VAL=$ACT_PENDING_SCORE"
ocf_log debug "$MASTER_SCORE, $ACT_MASTER_SCORE, $ROLE"
DO_MASTER_SCORE_UPDATE="NO"
DO_PENDING_UPDATE="NO"
if [ -n "$ACT_MASTER_SCORE" ]
then
if [ $ACT_MASTER_SCORE -eq $MASTER_SCORE ]; then
mpathpersist_clear_pending
else
case $ROLE in
Master)
if [ $MASTER_SCORE -lt $ACT_MASTER_SCORE ]; then
if [ -n "$ACT_PENDING" ]
then
if [ $(($NOW-$ACT_PENDING_TS-$MASTER_SCORE_DELAY)) -ge 0 ]; then
mpathpersist_new_master_score $MASTER_SCORE
mpathpersist_clear_pending
fi
else
if [ $MASTER_SCORE_DELAY -eq 0 ]; then
mpathpersist_new_master_score $MASTER_SCORE
mpathpersist_clear_pending
else
mpathpersist_new_pending "${NOW}_${MASTER_SCORE}"
fi
fi
else
mpathpersist_new_master_score $MASTER_SCORE
mpathpersist_clear_pending
fi
;;
Slave)
if [ $MASTER_SCORE -gt $ACT_MASTER_SCORE ]; then
if [ -n "$ACT_PENDING" ]; then
if [ $(($NOW-$ACT_PENDING_TS-$MASTER_SCORE_DELAY)) -ge 0 ]; then
mpathpersist_new_master_score $MASTER_SCORE
mpathpersist_clear_pending
fi
else
if [ $MASTER_SCORE_DELAY -eq 0 ]; then
mpathpersist_new_master_score $MASTER_SCORE
mpathpersist_clear_pending
else
mpathpersist_new_pending "${NOW}_${MASTER_SCORE}"
fi
fi
else
mpathpersist_new_master_score $MASTER_SCORE
mpathpersist_clear_pending
fi
;;
*)
;;
esac
fi
fi
if [ $DO_MASTER_SCORE_UPDATE == "YES" ]; then
ocf_run $MASTER_SCORE_ATTRIBUTE --update=$NEW_MASTER_SCORE
fi
if [ $DO_PENDING_UPDATE == "YES" ]; then
ocf_run $PENDING_ATTRIBUTE --update=$NEW_PENDING
fi
if [ ${#REGISTERED_DEVS[*]} -eq 0 ]; then
ocf_log debug "$RESOURCE monitor: no registrations"
return $OCF_NOT_RUNNING
fi
if [ ${#RESERVED_DEVS[*]} -eq ${#WORKING_DEVS[*]} ]; then
return $OCF_RUNNING_MASTER
fi
if [ ${#REGISTERED_DEVS[*]} -eq ${#WORKING_DEVS[*]} ]; then
if [ $RESERVATION_TYPE -eq 7 ] || [ $RESERVATION_TYPE -eq 8 ]; then
if [ ${#DEVS_WITH_RESERVATION[*]} -gt 0 ]; then
return $OCF_RUNNING_MASTER
else
return $OCF_SUCCESS
fi
else
return $OCF_SUCCESS
fi
fi
ocf_log err "$RESOURCE monitor: unexpected state"
return $OCF_ERR_GENERIC
}
mpathpersist_action_promote() {
if [ ${#RESERVED_DEVS[*]} -gt 0 ]; then
ocf_log info "$RESOURCE promote: already master"
return $OCF_SUCCESS
fi
for dev in ${WORKING_DEVS[*]}
do
reservation_key=`mpathpersist_get_reservation_key $dev`
case $RESERVATION_TYPE in
1|3|5|6)
if [ -z "$reservation_key" ]; then
ocf_run $MPATHPERSIST --out --reserve --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
else
ocf_run $MPATHPERSIST --out --preempt --param-sark=$reservation_key --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
fi
;;
7|8)
if [ -z "$reservation_key" ]; then
ocf_run $MPATHPERSIST --out --reserve --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]
then
return $OCF_ERR_GENERIC
fi
else
ocf_log info "$RESOURCE promote: there already exist an reservation holder, all registrants become reservation holders"
return $OCF_SUCCESS
fi
;;
*)
return $OCF_ERR_ARGS
;;
esac
done
return $OCF_SUCCESS
}
mpathpersist_action_demote() {
case $RESERVATION_TYPE in
1|3|5|6)
if [ ${#RESERVED_DEVS[*]} -eq 0 ]; then
ocf_log info "$RESOURCE demote: already slave"
return $OCF_SUCCESS
fi
for dev in ${RESERVED_DEVS[*]}
do
ocf_run $MPATHPERSIST --out --release --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
done
;;
7|8) #in case of 7/8, --release won't release the reservation unless unregister the key.
if [ ${#REGISTERED_DEVS[*]} -eq 0 ]; then
ocf_log info "$RESOURCE demote: already slave"
return $OCF_SUCCESS
fi
for dev in ${REGISTERED_DEVS[*]}
do
ocf_run $MPATHPERSIST --out --register --param-rk=$NODE_ID_HEX --param-sark=0 $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
done
;;
*)
return $OCF_ERR_ARGS
;;
esac
return $OCF_SUCCESS
}
mpathpersist_action_notify() {
local n_type="$OCF_RESKEY_CRM_meta_notify_type"
local n_op="$OCF_RESKEY_CRM_meta_notify_operation"
set -- $OCF_RESKEY_CRM_meta_notify_active_resource
local n_active="$#"
set -- $OCF_RESKEY_CRM_meta_notify_stop_resource
local n_stop="$#"
set -- $OCF_RESKEY_CRM_meta_notify_start_resource
local n_start="$#"
ocf_log debug "$RESOURCE notify: $n_type for $n_op - counts: active $n_active - starting $n_start - stopping $n_stop"
return $OCF_SUCCESS
}
mpathpersist_action_validate_all () {
-
- if [ "$OCF_RESKEY_CRM_meta_master_max" != "1" ] && [ "$RESERVATION_TYPE" != "7" ] && [ "$RESERVATION_TYPE" != "8" ]; then
- ocf_log err "Master options misconfigured."
- exit $OCF_ERR_CONFIGURED
+ if [ "$OCF_CHECK_LEVEL" -eq 10 ]; then
+ if [ "$OCF_RESKEY_CRM_meta_master_max" != "1" ] && [ "$RESERVATION_TYPE" != "7" ] && [ "$RESERVATION_TYPE" != "8" ]; then
+ ocf_log err "Master options misconfigured."
+ exit $OCF_ERR_CONFIGURED
+ fi
fi
return $OCF_SUCCESS
}
if [ $# -ne 1 ]; then
echo "Incorrect parameter count."
mpathpersist_action_usage
exit $OCF_ERR_ARGS
fi
ACTION=$1
case $ACTION in
meta-data)
meta_data
;;
validate-all)
mpathpersist_init
mpathpersist_action_validate_all
;;
start|promote|monitor|stop|demote)
ocf_log debug "$RESOURCE: starting action \"$ACTION\""
mpathpersist_init
+ if [ "$__OCF_ACTION" = "start" ]; then
+ OCF_CHECK_LEVEL=10
+ mpathpersist_action_validate_all
+ fi
mpathpersist_action_$ACTION
exit $?
;;
notify)
mpathpersist_action_notify
exit $?
;;
usage|help)
mpathpersist_action_usage
exit $OCF_SUCCESS
;;
*)
mpathpersist_action_usage
exit $OCF_ERR_ARGS
;;
esac
diff --git a/heartbeat/sg_persist.in b/heartbeat/sg_persist.in
index 16048ea6f..620c02f4a 100644
--- a/heartbeat/sg_persist.in
+++ b/heartbeat/sg_persist.in
@@ -1,694 +1,699 @@
#!@BASH_SHELL@
#
#
# OCF Resource Agent compliant PERSISTENT SCSI RESERVATION resource script.
#
#
# Copyright (c) 2011 Evgeny Nifontov and lwang@suse.com All Rights Reserved.
#
# "Heartbeat drbd OCF Resource Agent: 2007, Lars Marowsky-Bree" was used
# as example of multistate OCF Resource Agent.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of version 2 of the GNU General Public License as
# published by the Free Software Foundation.
#
# This program is distributed in the hope that it would be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
#
# Further, this software is distributed without any warranty that it is
# free of the rightful claim of any third person regarding infringement
# or the like. Any license provided herein, whether implied or
# otherwise, applies only to this software file. Patent licenses, if
# any, provided herein do not apply to combinations of this program with
# other software, or any other product whatsoever.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write the Free Software Foundation,
# Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA.
#
#
# OCF instance parameters
# OCF_RESKEY_binary
# OCF_RESKEY_devs
# OCF_RESKEY_required_devs_nof
# OCF_RESKEY_reservation_type
# OCF_RESKEY_master_score_base
# OCF_RESKEY_master_score_dev_factor
# OCF_RESKEY_master_score_delay
#
# TODO
#
# 1) PROBLEM: devices which were not accessible during 'start' action, will be never registered/reserved
# TODO: 'Master' and 'Salve' registers new devs in 'monitor' action
# TODO: 'Master' reserves new devs in 'monitor' action
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
# Parameter defaults
OCF_RESKEY_binary_default="sg_persist" # binary name for the resource
OCF_RESKEY_devs_default="" # device list
OCF_RESKEY_required_devs_nof_default="1" # number of required devices
OCF_RESKEY_reservation_type_default="1" # reservation type
OCF_RESKEY_master_score_base_default="0" # master score base
OCF_RESKEY_master_score_dev_factor_default="100" # device factor for master score
OCF_RESKEY_master_score_delay_default="30" # delay for master score
: ${OCF_RESKEY_binary=${OCF_RESKEY_binary_default}}
: ${OCF_RESKEY_devs=${OCF_RESKEY_devs_default}}
: ${OCF_RESKEY_required_devs_nof=${OCF_RESKEY_required_devs_nof_default}}
: ${OCF_RESKEY_reservation_type=${OCF_RESKEY_reservation_type_default}}
: ${OCF_RESKEY_master_score_base=${OCF_RESKEY_master_score_base_default}}
: ${OCF_RESKEY_master_score_dev_factor=${OCF_RESKEY_master_score_dev_factor_default}}
: ${OCF_RESKEY_master_score_delay=${OCF_RESKEY_master_score_delay_default}}
#######################################################################
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="sg_persist" version="1.1">
<version>1.0</version>
<longdesc lang="en">
This resource agent manages SCSI PERSISTENT RESERVATIONS.
"sg_persist" from sg3_utils is used, please see its documentation.
Should be used as multistate (Master/Slave) resource
Slave registers its node id ("crm_node -i") as reservation key ( --param-rk ) on each device in the "devs" list.
Master reserves all devices from "devs" list with reservation "--prout-type" value from "reservation_type" parameter.
</longdesc>
<shortdesc lang="en">Manages SCSI PERSISTENT RESERVATIONS</shortdesc>
<parameters>
<parameter name="binary" unique="0">
<longdesc lang="en">
The name of the binary that manages the resource.
</longdesc>
<shortdesc lang="en">the binary name of the resource</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}"/>
</parameter>
<parameter name="devs" unique="0" required="1">
<longdesc lang="en">
Device list. Multiple devices can be listed with blank space as separator.
Shell wildcards are allowed.
</longdesc>
<shortdesc lang="en">device list</shortdesc>
<content type="string"/>
</parameter>
<parameter name="required_devs_nof" unique="0" required="0">
<longdesc lang="en">
Minimum number of "working" devices from device list
1) existing
2) "sg_persist --read-keys \$device" works (Return code 0)
resource actions "start","monitor","promote" and "validate-all" return "\$OCF_ERR_INSTALLED"
if the actual number of "working" devices is less then "required_devs_nof".
resource actions "stop" and "demote" tries to remove reservations and registration keys from
all working devices, but always return "\$OCF_SUCCESS"
</longdesc>
<shortdesc lang="en">minimum number of working devices</shortdesc>
<content type="string" default="${OCF_RESKEY_required_devs_nof_default}"/>
</parameter>
<parameter name="reservation_type" unique="0" required="0">
<longdesc lang="en">
reservation type
</longdesc>
<shortdesc lang="en">reservation type</shortdesc>
<content type="string" default="${OCF_RESKEY_reservation_type_default}" />
</parameter>
<parameter name="master_score_base" unique="0" required="0">
<longdesc lang="en">
master_score_base value
"master_score_base" value is used in "master_score" calculation:
master_score = \$master_score_base + \$master_score_dev_factor * \$working_devs
if set to bigger value in sg_persist resource configuration on some node, this node will be "preferred" for master role.
</longdesc>
<shortdesc lang="en">base master_score value</shortdesc>
<content type="string" default="${OCF_RESKEY_master_score_base_default}" />
</parameter>
<parameter name="master_score_dev_factor" unique="0" required="0">
<longdesc lang="en">
Working device factor in master_score calculation
each "working" device provides additional value to "master_score",
so the node that sees more devices will be preferred for the "Master"-role
Setting it to 0 will disable this behavior.
</longdesc>
<shortdesc lang="en">working device factor in master_score calculation</shortdesc>
<content type="string" default="${OCF_RESKEY_master_score_dev_factor_default}" />
</parameter>
<parameter name="master_score_delay" unique="0" required="0">
<longdesc lang="en">
master/slave decreases/increases its master_score after delay of \$master_score_delay seconds
so if some device gets inaccessible, the slave decreases its master_score first and the resource will no be watched
and after this device reappears again the master increases its master_score first
this can work only if the master_score_delay is bigger then monitor interval on both master and slave
Setting it to 0 will disable this behavior.
</longdesc>
<shortdesc lang="en">master_score decrease/increase delay time</shortdesc>
<content type="string" default="${OCF_RESKEY_master_score_delay_default}" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="30s" />
<action name="promote" timeout="30s" />
<action name="demote" timeout="30s" />
<action name="notify" timeout="30s" />
<action name="stop" timeout="30s" />
<action name="monitor" depth="0" timeout="20s" interval="29s" role="Unpromoted" />
<action name="monitor" depth="0" timeout="20s" interval="60s" role="Promoted" />
<action name="meta-data" timeout="5s" />
<action name="validate-all" timeout="30s" />
</actions>
</resource-agent>
END
exit $OCF_SUCCESS
}
sg_persist_init() {
if ! ocf_is_root ; then
ocf_log err "You must be root to perform this operation."
exit $OCF_ERR_PERM
fi
SG_PERSIST=${OCF_RESKEY_binary}
check_binary $SG_PERSIST
ROLE=$OCF_RESKEY_CRM_meta_role
NOW=$(date +%s)
RESOURCE="${OCF_RESOURCE_INSTANCE}"
MASTER_SCORE_VAR_NAME="master-${OCF_RESOURCE_INSTANCE//:/-}"
PENDING_VAR_NAME="pending-$MASTER_SCORE_VAR_NAME"
#only works with corocync
CRM_NODE="${HA_SBIN_DIR}/crm_node"
NODE_ID_DEC=$($CRM_NODE -i)
NODE=$($CRM_NODE -l | $GREP -w ^$NODE_ID_DEC)
NODE=${NODE#$NODE_ID_DEC }
NODE=${NODE% *}
MASTER_SCORE_ATTRIBUTE="${HA_SBIN_DIR}/crm_attribute --lifetime=reboot --name=$MASTER_SCORE_VAR_NAME --node=$NODE"
PENDING_ATTRIBUTE="${HA_SBIN_DIR}/crm_attribute --lifetime=reboot --name=$PENDING_VAR_NAME --node=$NODE"
NODE_ID_HEX=$(printf '0x%x' $NODE_ID_DEC)
if [ -z "$NODE_ID_HEX" ]; then
ocf_log err "Couldn't get node id with \"$CRM_NODE\""
exit $OCF_ERR_INSTALLED
fi
ocf_log debug "$RESOURCE: NODE:$NODE, ROLE:$ROLE, NODE_ID DEC:$NODE_ID_DEC HEX:$NODE_ID_HEX"
DEVS=${OCF_RESKEY_devs}
REQUIRED_DEVS_NOF=${OCF_RESKEY_required_devs_nof}
RESERVATION_TYPE=${OCF_RESKEY_reservation_type}
MASTER_SCORE_BASE=${OCF_RESKEY_master_score_base}
MASTER_SCORE_DEV_FACTOR=${OCF_RESKEY_master_score_dev_factor}
MASTER_SCORE_DELAY=${OCF_RESKEY_master_score_delay}
ocf_log debug "$RESOURCE: DEVS=$DEVS"
ocf_log debug "$RESOURCE: REQUIRED_DEVS_NOF=$REQUIRED_DEVS_NOF"
ocf_log debug "$RESOURCE: RESERVATION_TYPE=$RESERVATION_TYPE"
ocf_log debug "$RESOURCE: MASTER_SCORE_BASE=$MASTER_SCORE_BASE"
ocf_log debug "$RESOURCE: MASTER_SCORE_DEV_FACTOR=$MASTER_SCORE_DEV_FACTOR"
ocf_log debug "$RESOURCE: MASTER_SCORE_DELAY=$MASTER_SCORE_DELAY"
#expand path wildcards
DEVS=$(echo $DEVS)
if [ -z "$DEVS" ]; then
ocf_log err "\"devs\" not defined"
exit $OCF_ERR_INSTALLED
fi
sg_persist_check_devs
sg_persist_get_status
}
sg_persist_action_usage() {
cat <<END
usage: $0 {start|stop|monitor|validate-all|promote|demote|notify|meta-data}
Expects to have a fully populated OCF RA-compliant environment set.
END
}
sg_persist_get_status() {
unset WORKING_DEVS[*]
for dev in ${EXISTING_DEVS[*]}
do
READ_KEYS=`$SG_PERSIST --in --read-keys $dev 2>&1`
[ $? -eq 0 ] || continue
WORKING_DEVS+=($dev)
echo "$READ_KEYS" | $GREP -qw $NODE_ID_HEX\$
[ $? -eq 0 ] || continue
REGISTERED_DEVS+=($dev)
READ_RESERVATION=`$SG_PERSIST --in --read-reservation $dev 2>&1`
[ $? -eq 0 ] || continue
echo "$READ_RESERVATION" | $GREP -qw $NODE_ID_HEX\$
if [ $? -eq 0 ]; then
RESERVED_DEVS+=($dev)
fi
reservation_key=`echo $READ_RESERVATION | $GREP -o 'Key=0x[0-9a-f]*' | $GREP -o '0x[0-9a-f]*'`
if [ -n "$reservation_key" ]; then
DEVS_WITH_RESERVATION+=($dev)
RESERVATION_KEYS+=($reservation_key)
fi
done
WORKING_DEVS_NOF=${#WORKING_DEVS[*]}
ocf_log debug "$RESOURCE: working devices: `sg_persist_echo_array ${WORKING_DEVS[*]}`"
ocf_log debug "$RESOURCE: number of working devices: $WORKING_DEVS_NOF"
ocf_log debug "$RESOURCE: registered devices: `sg_persist_echo_array ${REGISTERED_DEVS[*]}`"
ocf_log debug "$RESOURCE: reserved devices: `sg_persist_echo_array ${RESERVED_DEVS[*]}`"
ocf_log debug "$RESOURCE: devices with reservation: `sg_persist_echo_array ${DEVS_WITH_RESERVATION[*]}`"
ocf_log debug "$RESOURCE: reservation keys: `sg_persist_echo_array ${RESERVATION_KEYS[*]}`"
MASTER_SCORE=$(($MASTER_SCORE_BASE + $MASTER_SCORE_DEV_FACTOR*$WORKING_DEVS_NOF))
ocf_log debug "$RESOURCE: master_score: $MASTER_SCORE_BASE + $MASTER_SCORE_DEV_FACTOR*$WORKING_DEVS_NOF = $MASTER_SCORE"
}
sg_persist_check_devs() {
for dev in $DEVS
do
if [ -e "$dev" ]; then
EXISTING_DEVS+=($dev)
fi
done
EXISTING_DEVS_NOF=${#EXISTING_DEVS[*]}
if [ $EXISTING_DEVS_NOF -lt $REQUIRED_DEVS_NOF ]; then
ocf_log err "Number of existing devices=$EXISTING_DEVS_NOF less then required_devs_nof=$REQUIRED_DEVS_NOF"
exit $OCF_ERR_INSTALLED
fi
}
sg_persist_is_registered() {
for registered_dev in ${REGISTERED_DEVS[*]}
do
if [ "$registered_dev" == "$1" ]; then
return 0
fi
done
return 1
}
sg_persist_get_reservation_key() {
for array_index in ${!DEVS_WITH_RESERVATION[*]}
do
if [ "${DEVS_WITH_RESERVATION[$array_index]}" == "$1" ]; then
echo ${RESERVATION_KEYS[$array_index]}
return 0
fi
done
echo ""
}
sg_persist_echo_array() {
str_count=0
arr_str=""
for str in "$@"
do
arr_str="$arr_str[$str_count]:$str "
str_count=$(($str_count+1))
done
echo $arr_str
}
sg_persist_parse_act_pending() {
ACT_PENDING_TS=0
ACT_PENDING_SCORE=0
if [ -n "$ACT_PENDING" ]; then
ACT_PENDING_TS=${ACT_PENDING%%_*}
ACT_PENDING_SCORE=${ACT_PENDING##*_}
fi
}
sg_persist_clear_pending() {
if [ -n "$ACT_PENDING" ]; then
DO_PENDING_UPDATE="YES"
NEW_PENDING=""
fi
}
sg_persist_new_master_score() {
DO_MASTER_SCORE_UPDATE="YES"
NEW_MASTER_SCORE=$1
}
sg_persist_new_pending() {
DO_PENDING_UPDATE="YES"
NEW_PENDING=$1
}
# Functions invoked by resource manager actions
sg_persist_action_start() {
ocf_run $MASTER_SCORE_ATTRIBUTE --update=$MASTER_SCORE
ocf_run $PENDING_ATTRIBUTE --update=""
if [ $WORKING_DEVS_NOF -lt $REQUIRED_DEVS_NOF ]; then
ocf_log err "$RESOURCE: Number of working devices=$WORKING_DEVS_NOF less then required_devs_nof=$REQUIRED_DEVS_NOF"
exit $OCF_ERR_GENERIC
fi
for dev in ${WORKING_DEVS[*]}
do
if sg_persist_is_registered $dev ; then
: OK
else
ocf_run $SG_PERSIST --out --no-inquiry --register --param-rk=0 --param-sark=$NODE_ID_HEX $dev
if [ $? -ne $OCF_SUCCESS ]
then
return $OCF_ERR_GENERIC
fi
fi
done
return $OCF_SUCCESS
}
sg_persist_action_stop() {
if [ ${#REGISTERED_DEVS[*]} -eq 0 ]; then
ocf_log debug "$RESOURCE stop: already no registrations"
else
# Clear preference for becoming master
ocf_run $MASTER_SCORE_ATTRIBUTE --delete
ocf_run $PENDING_ATTRIBUTE --delete
for dev in ${REGISTERED_DEVS[*]}
do
ocf_run $SG_PERSIST --out --no-inquiry --register --param-rk=$NODE_ID_HEX --param-sark=0 $dev
done
fi
return $OCF_SUCCESS
}
sg_persist_action_monitor() {
ACT_MASTER_SCORE=`$MASTER_SCORE_ATTRIBUTE --query --quiet 2>/dev/null`
ocf_log debug "$RESOURCE monitor: ACT_MASTER_SCORE=$ACT_MASTER_SCORE"
ACT_PENDING=`$PENDING_ATTRIBUTE --query --quiet 2>/dev/null`
ocf_log debug "$RESOURCE monitor: ACT_PENDING=$ACT_PENDING"
sg_persist_parse_act_pending
ocf_log debug "$RESOURCE monitor: ACT_PENDING_TS=$ACT_PENDING_TS"
ocf_log debug "$RESOURCE monitor: ACT_PENDING_VAL=$ACT_PENDING_SCORE"
ocf_log debug "$MASTER_SCORE, $ACT_MASTER_SCORE, $ROLE"
DO_MASTER_SCORE_UPDATE="NO"
DO_PENDING_UPDATE="NO"
if [ -n "$ACT_MASTER_SCORE" ]
then
if [ $ACT_MASTER_SCORE -eq $MASTER_SCORE ]; then
sg_persist_clear_pending
else
case $ROLE in
Master)
if [ $MASTER_SCORE -lt $ACT_MASTER_SCORE ]; then
if [ -n "$ACT_PENDING" ]
then
if [ $(($NOW-$ACT_PENDING_TS-$MASTER_SCORE_DELAY)) -ge 0 ]; then
sg_persist_new_master_score $MASTER_SCORE
sg_persist_clear_pending
fi
else
if [ $MASTER_SCORE_DELAY -eq 0 ]; then
sg_persist_new_master_score $MASTER_SCORE
sg_persist_clear_pending
else
sg_persist_new_pending "${NOW}_${MASTER_SCORE}"
fi
fi
else
sg_persist_new_master_score $MASTER_SCORE
sg_persist_clear_pending
fi
;;
Slave)
if [ $MASTER_SCORE -gt $ACT_MASTER_SCORE ]; then
if [ -n "$ACT_PENDING" ]; then
if [ $(($NOW-$ACT_PENDING_TS-$MASTER_SCORE_DELAY)) -ge 0 ]; then
sg_persist_new_master_score $MASTER_SCORE
sg_persist_clear_pending
fi
else
if [ $MASTER_SCORE_DELAY -eq 0 ]; then
sg_persist_new_master_score $MASTER_SCORE
sg_persist_clear_pending
else
sg_persist_new_pending "${NOW}_${MASTER_SCORE}"
fi
fi
else
sg_persist_new_master_score $MASTER_SCORE
sg_persist_clear_pending
fi
;;
*)
;;
esac
fi
fi
if [ $DO_MASTER_SCORE_UPDATE == "YES" ]; then
ocf_run $MASTER_SCORE_ATTRIBUTE --update=$NEW_MASTER_SCORE
fi
if [ $DO_PENDING_UPDATE == "YES" ]; then
ocf_run $PENDING_ATTRIBUTE --update=$NEW_PENDING
fi
if [ ${#REGISTERED_DEVS[*]} -eq 0 ]; then
ocf_log debug "$RESOURCE monitor: no registrations"
if [ -n "$ACT_MASTER_SCORE" ]; then
ocf_run $MASTER_SCORE_ATTRIBUTE --delete
ocf_run $PENDING_ATTRIBUTE --delete
fi
return $OCF_NOT_RUNNING
fi
if [ ${#RESERVED_DEVS[*]} -eq ${#WORKING_DEVS[*]} ]; then
if [ -z "$ACT_MASTER_SCORE" ]; then
ocf_run $MASTER_SCORE_ATTRIBUTE --update=$MASTER_SCORE
ocf_run $PENDING_ATTRIBUTE --update=""
fi
return $OCF_RUNNING_MASTER
fi
if [ ${#REGISTERED_DEVS[*]} -eq ${#WORKING_DEVS[*]} ]; then
if [ -z "$ACT_MASTER_SCORE" ]; then
ocf_run $MASTER_SCORE_ATTRIBUTE --update=$MASTER_SCORE
ocf_run $PENDING_ATTRIBUTE --update=""
fi
if [ $RESERVATION_TYPE -eq 7 ] || [ $RESERVATION_TYPE -eq 8 ]; then
if [ ${#DEVS_WITH_RESERVATION[*]} -gt 0 ]; then
return $OCF_RUNNING_MASTER
else
return $OCF_SUCCESS
fi
else
return $OCF_SUCCESS
fi
fi
ocf_log err "$RESOURCE monitor: unexpected state"
return $OCF_ERR_GENERIC
}
sg_persist_action_promote() {
if [ ${#RESERVED_DEVS[*]} -gt 0 ]; then
ocf_log info "$RESOURCE promote: already master"
return $OCF_SUCCESS
fi
for dev in ${WORKING_DEVS[*]}
do
reservation_key=`sg_persist_get_reservation_key $dev`
case $RESERVATION_TYPE in
1|3|5|6)
if [ -z "$reservation_key" ]; then
ocf_run $SG_PERSIST --out --no-inquiry --reserve --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
else
ocf_run $SG_PERSIST --out --no-inquiry --preempt --param-sark=$reservation_key --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
fi
;;
7|8)
if [ -z "$reservation_key" ]; then
ocf_run $SG_PERSIST --out --no-inquiry --reserve --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]
then
return $OCF_ERR_GENERIC
fi
else
ocf_log info "$RESOURCE promote: there already exist an reservation holder, all registrants become reservation holders"
return $OCF_SUCCESS
fi
;;
*)
return $OCF_ERR_ARGS
;;
esac
done
return $OCF_SUCCESS
}
sg_persist_action_demote() {
case $RESERVATION_TYPE in
1|3|5|6)
if [ ${#RESERVED_DEVS[*]} -eq 0 ]; then
ocf_log info "$RESOURCE demote: already slave"
return $OCF_SUCCESS
fi
for dev in ${RESERVED_DEVS[*]}
do
ocf_run $SG_PERSIST --out --no-inquiry --release --param-rk=$NODE_ID_HEX --prout-type=$RESERVATION_TYPE $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
done
;;
7|8) #in case of 7/8, --release won't release the reservation unless unregister the key.
if [ ${#REGISTERED_DEVS[*]} -eq 0 ]; then
ocf_log info "$RESOURCE demote: already slave"
return $OCF_SUCCESS
fi
for dev in ${REGISTERED_DEVS[*]}
do
ocf_run $SG_PERSIST --out --no-inquiry --register --param-rk=$NODE_ID_HEX --param-sark=0 $dev
if [ $? -ne $OCF_SUCCESS ]; then
return $OCF_ERR_GENERIC
fi
done
;;
*)
return $OCF_ERR_ARGS
;;
esac
return $OCF_SUCCESS
}
sg_persist_action_notify() {
local n_type="$OCF_RESKEY_CRM_meta_notify_type"
local n_op="$OCF_RESKEY_CRM_meta_notify_operation"
set -- $OCF_RESKEY_CRM_meta_notify_active_resource
local n_active="$#"
set -- $OCF_RESKEY_CRM_meta_notify_stop_resource
local n_stop="$#"
set -- $OCF_RESKEY_CRM_meta_notify_start_resource
local n_start="$#"
ocf_log debug "$RESOURCE notify: $n_type for $n_op - counts: active $n_active - starting $n_start - stopping $n_stop"
return $OCF_SUCCESS
}
sg_persist_action_validate_all () {
-
- if [ "$OCF_RESKEY_CRM_meta_master_max" != "1" ] && [ "$RESERVATION_TYPE" != "7" ] && [ "$RESERVATION_TYPE" != "8" ]; then
- ocf_log err "Master options misconfigured."
- exit $OCF_ERR_CONFIGURED
+ if [ "$OCF_CHECK_LEVEL" -eq 10 ]; then
+ if [ "$OCF_RESKEY_CRM_meta_master_max" != "1" ] && [ "$RESERVATION_TYPE" != "7" ] && [ "$RESERVATION_TYPE" != "8" ]; then
+ ocf_log err "Master options misconfigured."
+ exit $OCF_ERR_CONFIGURED
+ fi
fi
return $OCF_SUCCESS
}
if [ $# -ne 1 ]; then
echo "Incorrect parameter count."
sg_persist_action_usage
exit $OCF_ERR_ARGS
fi
ACTION=$1
case $ACTION in
meta-data)
meta_data
;;
validate-all)
sg_persist_init
sg_persist_action_validate_all
;;
start|promote|monitor|stop|demote)
ocf_log debug "$RESOURCE: starting action \"$ACTION\""
sg_persist_init
+ if [ "$__OCF_ACTION" = "start" ]; then
+ OCF_CHECK_LEVEL=10
+ sg_persist_action_validate_all
+ fi
sg_persist_action_$ACTION
exit $?
;;
notify)
sg_persist_action_notify
exit $?
;;
usage|help)
sg_persist_action_usage
exit $OCF_SUCCESS
;;
*)
sg_persist_action_usage
exit $OCF_ERR_ARGS
;;
esac

File Metadata

Mime Type
text/x-diff
Expires
Thu, Oct 16, 12:21 AM (22 h, 53 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
2530817
Default Alt Text
(101 KB)

Event Timeline