Page MenuHomeClusterLabs Projects
Files F3155414

No OneTemporary
Actions

View Options

diff --git a/heartbeat/NodeUtilization b/heartbeat/NodeUtilization
index 61969e6f7..c4708ab43 100755
--- a/heartbeat/NodeUtilization
+++ b/heartbeat/NodeUtilization
@@ -1,226 +1,226 @@
#!/bin/sh
#
#
# NodeUtilization OCF Resource Agent
#
# Copyright (c) 2011 SUSE LINUX, John Shi
# Copyright (c) 2016 SUSE LINUX, Kristoffer Gronlund
# 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.
#
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
#######################################################################
NodeUtilization_meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="NodeUtilization">
<version>1.0</version>
<longdesc lang="en">
The Node Utilization agent detects system parameters like available CPU, host
memory and hypervisor memory availability, and adds them into the CIB for each
node using crm_attribute. Run the agent as a clone resource to have it populate
these parameters on each node.
Note: Setting hv_memory only works with Xen at the moment, using the xl or xm
command line tools.
</longdesc>
<shortdesc lang="en">Node Utilization</shortdesc>
<parameters>
<parameter name="dynamic" unique="0" required="0">
<longdesc lang="en">
If set, parameters will be updated if there are differences between the HA
parameters and the system values when running the monitor action.
If not set, the parameters will be set once when the resource instance starts.
</longdesc>
<shortdesc lang="en">Dynamically update parameters in monitor</shortdesc>
<content type="boolean" default="true" />
</parameter>
<parameter name="utilization_cpu" unique="0" required="0">
<longdesc lang="en">Enable setting node CPU utilization limit.</longdesc>
<shortdesc lang="en">Set node CPU utilization limit.</shortdesc>
<content type="boolean" default="true" />
</parameter>
<parameter name="utilization_cpu_reservation" unique="0" required="0">
<longdesc lang="en">Subtract this value when setting the CPU utilization parameter.</longdesc>
<shortdesc lang="en">CPU reservation.</shortdesc>
<content type="integer" default="1" />
</parameter>
<parameter name="utilization_host_memory" unique="0" required="0">
<longdesc lang="en">Enable setting available host memory.</longdesc>
<shortdesc lang="en">Set available host memory.</shortdesc>
<content type="boolean" default="true" />
</parameter>
<parameter name="utilization_host_memory_reservation" unique="0" required="0">
<longdesc lang="en">Subtract this value when setting host memory utilization, in MB.</longdesc>
<shortdesc lang="en">Host memory reservation, in MB.</shortdesc>
<content type="integer" default="512" />
</parameter>
<parameter name="utilization_hv_memory" unique="0" required="0">
<longdesc lang="en">Enable setting available hypervisor memory.</longdesc>
<shortdesc lang="en">Set available hypervisor memory.</shortdesc>
<content type="boolean" default="true" />
</parameter>
<parameter name="utilization_hv_memory_reservation" unique="0" required="0">
<longdesc lang="en">Subtract this value when setting hypervisor memory utilization, in MB.</longdesc>
<shortdesc lang="en">Hypervisor memory reservation, in MB.</shortdesc>
<content type="integer" default="512" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="90" />
<action name="stop" timeout="100" />
<action name="monitor" timeout="20s" interval="60s"/>
<action name="meta-data" timeout="5" />
<action name="validate-all" timeout="30" />
</actions>
</resource-agent>
END
}
Host_Total_Memory() {
local xentool
xentool=$(which xl 2> /dev/null || which xm 2> /dev/null)
if [ -x $xentool ]; then
$xentool info | awk '/total_memory/{printf("%d\n",$3);exit(0)}'
else
ocf_log warn "Can only set hv_memory for Xen hypervisor"
echo "0"
fi
}
set_utilization() {
host_name="$(ocf_local_nodename)"
if ocf_is_true "$OCF_RESKEY_utilization_cpu"; then
sys_cpu=$(( $(grep -c processor /proc/cpuinfo) - $OCF_RESKEY_utilization_cpu_reservation ))
- uti_cpu=$(crm_attribute -Q -t nodes -U "$host_name" -z -n cpu 2>/dev/null)
+ uti_cpu=$(crm_attribute --quiet -t nodes --node "$host_name" -z -n cpu 2>/dev/null)
if [ "$sys_cpu" != "$uti_cpu" ]; then
- if ! crm_attribute -t nodes -U "$host_name" -z -n cpu -v $sys_cpu; then
+ if ! crm_attribute -t nodes --node "$host_name" -z -n cpu -v $sys_cpu; then
ocf_log err "Failed to set the cpu utilization attribute for $host_name using crm_attribute."
return 1
fi
fi
fi
if ocf_is_true "$OCF_RESKEY_utilization_host_memory"; then
sys_mem=$(( $(awk '/MemTotal/{printf("%d\n",$2/1024);exit(0)}' /proc/meminfo) - $OCF_RESKEY_utilization_host_memory_reservation ))
- uti_mem=$(crm_attribute -Q -t nodes -U "$host_name" -z -n host_memory 2>/dev/null)
+ uti_mem=$(crm_attribute --quiet -t nodes --node "$host_name" -z -n host_memory 2>/dev/null)
if [ "$sys_mem" != "$uti_mem" ]; then
- if ! crm_attribute -t nodes -U "$host_name" -z -n host_memory -v $sys_mem; then
+ if ! crm_attribute -t nodes --node "$host_name" -z -n host_memory -v $sys_mem; then
ocf_log err "Failed to set the host_memory utilization attribute for $host_name using crm_attribute."
return 1
fi
fi
fi
if ocf_is_true "$OCF_RESKEY_utilization_hv_memory"; then
hv_mem=$(( $(Host_Total_Memory) - OCF_RESKEY_utilization_hv_memory_reservation ))
- uti_mem=$(crm_attribute -Q -t nodes -U "$host_name" -z -n hv_memory 2>/dev/null)
+ uti_mem=$(crm_attribute --quiet -t nodes --node "$host_name" -z -n hv_memory 2>/dev/null)
[ $hv_mem -lt 0 ] && hv_mem=0
if [ "$hv_mem" != "$uti_mem" ]; then
- if ! crm_attribute -t nodes -U "$host_name" -z -n hv_memory -v $hv_mem; then
+ if ! crm_attribute -t nodes --node "$host_name" -z -n hv_memory -v $hv_mem; then
ocf_log err "Failed to set the hv_memory utilization attribute for $host_name using crm_attribute."
return 1
fi
fi
fi
}
NodeUtilization_usage() {
cat <<END
usage: $0 {start|stop|monitor|validate-all|meta-data}
Expects to have a fully populated OCF RA-compliant environment set.
END
}
NodeUtilization_start() {
ha_pseudo_resource $statefile start
if ! ocf_is_true "$OCF_RESKEY_dynamic"; then
if ! set_utilization; then
exit $OCF_ERR_GENERIC
fi
fi
exit $OCF_SUCCESS
}
NodeUtilization_stop() {
ha_pseudo_resource $statefile stop
exit $OCF_SUCCESS
}
NodeUtilization_monitor() {
local rc
ha_pseudo_resource $statefile monitor
rc=$?
case $rc in
$OCF_SUCCESS)
if ocf_is_true "$OCF_RESKEY_dynamic"; then
if ! set_utilization; then
exit $OCF_ERR_GENERIC
fi
fi
;;
*) exit $rc;;
esac
}
NodeUtilization_validate() {
exit $OCF_SUCCESS
}
statefile=$OCF_RESOURCE_TYPE.$(echo $OCF_RESOURCE_INSTANCE | sed -e 's/^.*://')
: ${OCF_RESKEY_pidfile:="$HA_VARRUN/NodeUtilization-${OCF_RESOURCE_INSTANCE}"}
: ${OCF_RESKEY_dynamic:="true"}
: ${OCF_RESKEY_utilization_cpu:="true"}
: ${OCF_RESKEY_utilization_cpu_reservation="1"}
: ${OCF_RESKEY_utilization_hv_memory:="true"}
: ${OCF_RESKEY_utilization_hv_memory_reservation="512"}
: ${OCF_RESKEY_utilization_host_memory:="true"}
: ${OCF_RESKEY_utilization_host_memory_reservation="512"}
OCF_REQUIRED_PARAMS=""
OCF_REQUIRED_BINARIES=""
ocf_rarun $*
diff --git a/heartbeat/Xen b/heartbeat/Xen
index 92291ee15..a8e271432 100755
--- a/heartbeat/Xen
+++ b/heartbeat/Xen
@@ -1,572 +1,572 @@
#!/bin/sh
#
#
# Support: users@clusterlabs.org
# License: GNU General Public License (GPL)
#
# Resource Agent for the Xen Hypervisor.
# Manages Xen virtual machine instances by
# mapping cluster resource start and stop,
# to Xen create and shutdown, respectively.
#
# usage: $0 {start|stop|status|monitor|meta-data}
#
# OCF parameters are as below:
# OCF_RESKEY_xmfile
# Absolute path to the Xen control file,
# for this virtual machine.
# OCF_RESKEY_allow_mem_management
# Change memory usage on start/stop/migration
# of virtual machine
# OCF_RESKEY_reserved_Dom0_memory
# minimum memory reserved for domain 0
# OCF_RESKEY_monitor_scripts
# scripts to monitor services within the
# virtual domain
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
#######################################################################
usage() {
cat <<-END
usage: $0 {start|stop|status|monitor|meta-data|validate-all}
END
}
: ${OCF_RESKEY_xmfile=/etc/xen/vm/MyDomU}
: ${OCF_RESKEY_shutdown_acpi=0}
: ${OCF_RESKEY_allow_mem_management=0}
: ${OCF_RESKEY_reserved_Dom0_memory=512}
# prefer xl
xentool=$(which xl 2> /dev/null || which xm)
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="Xen">
<version>1.0</version>
<longdesc lang="en">
Resource Agent for the Xen Hypervisor.
Manages Xen virtual machine instances by mapping cluster resource
start and stop, to Xen create and shutdown, respectively.
A note on names
We will try to extract the name from the config file (the xmfile
attribute). If you use a simple assignment statement, then you
should be fine. Otherwise, if there's some python acrobacy
involved such as dynamically assigning names depending on other
variables, and we will try to detect this, then please set the
name attribute. You should also do that if there is any chance of
a pathological situation where a config file might be missing,
for example if it resides on a shared storage. If all fails, we
finally fall back to the instance id to preserve backward
compatibility.
Para-virtualized guests can also be migrated by enabling the
meta_attribute allow-migrate.
</longdesc>
<shortdesc lang="en">Manages Xen unprivileged domains (DomUs)</shortdesc>
<parameters>
<parameter name="xmfile" unique="0" required="1">
<longdesc lang="en">
Absolute path to the Xen control file,
for this virtual machine.
</longdesc>
<shortdesc lang="en">Xen control file</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="name" unique="0" required="0">
<longdesc lang="en">
Name of the virtual machine.
</longdesc>
<shortdesc lang="en">Xen DomU name</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="shutdown_timeout">
<longdesc lang="en">
The Xen agent will first try an orderly shutdown using xl shutdown.
Should this not succeed within this timeout, the agent will escalate to
xl destroy, forcibly killing the node.
If this is not set, it will default to two-third of the stop action
timeout.
Setting this value to 0 forces an immediate destroy.
</longdesc>
<shortdesc lang="en">Shutdown escalation timeout</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="shutdown_acpi" unique="0" required="0">
<longdesc lang="en">
Handle shutdown by simulating an ACPI power button event.
Enable this to allow graceful shutdown for HVM domains
without installed PV drivers.
</longdesc>
<shortdesc lang="en">Simulate power button event on shutdown</shortdesc>
<content type="boolean" default="0" />
</parameter>
<parameter name="allow_mem_management" unique="0" required="0">
<longdesc lang="en">
This parameter enables dynamic adjustment of memory for start
and stop actions used for Dom0 and the DomUs. The default is
to not adjust memory dynamically.
</longdesc>
<shortdesc lang="en">Use dynamic memory management</shortdesc>
<content type="boolean" default="0" />
</parameter>
<parameter name="node_ip_attribute">
<longdesc lang="en">
In case of a live migration, the system will default to using the IP
address associated with the hostname via DNS or /etc/hosts.
This parameter allows you to specify a node attribute that will be
queried instead for the target node, overriding the IP address. This
allows you to use a dedicated network for live migration traffic to a
specific node.
Warning: make very sure the IP address does point to the right node. Or
else the live migration will end up somewhere else, greatly confusing
the cluster and causing havoc.
</longdesc>
<shortdesc lang="en">Node attribute containing target IP address</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="reserved_Dom0_memory" unique="0" required="0">
<longdesc lang="en">
In case memory management is used, this parameter
defines the minimum amount of memory to be reserved
for the dom0. The default minimum memory is 512MB.
</longdesc>
<shortdesc lang="en">Minimum Dom0 memory</shortdesc>
<content type="string" default="512" />
</parameter>
<parameter name="monitor_scripts" unique="0" required="0">
<longdesc lang="en">
To additionally monitor services within the unprivileged domain,
add this parameter with a list of scripts to monitor.
</longdesc>
<shortdesc lang="en">list of space separated monitor scripts</shortdesc>
<content type="string" default="" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="60" />
<action name="stop" timeout="40" />
<action name="migrate_from" timeout="120" />
<action name="migrate_to" timeout="120" />
<action name="monitor" depth="0" timeout="30" interval="10" />
<action name="meta-data" timeout="5" />
<action name="validate-all" timeout="5" />
</actions>
</resource-agent>
END
}
Xen_Status() {
if expr "x$xentool" : "x.*xl" >/dev/null; then
$xentool list $1 >/dev/null 2>&1
if [ $? -ne 0 ]; then
return $OCF_NOT_RUNNING
else
return $OCF_SUCCESS
fi
fi
if have_binary xen-list; then
xen-list $1 2>/dev/null | grep -qs "State.*[-r][-b][-p]--" 2>/dev/null
if [ $? -ne 0 ]; then
return $OCF_NOT_RUNNING
else
return $OCF_SUCCESS
fi
fi
STATUS=`$xentool list --long $1 2>/dev/null | grep status 2>/dev/null`
if [ "X${STATUS}" != "X" ]; then
# we have Xen 3.0.4 or higher
STATUS_NOSPACES=`echo "$STATUS" | awk '{ print $1,$2}'`
if [ "$STATUS_NOSPACES" = "(status 2)" -o "$STATUS_NOSPACES" = "(status 1)" ]; then
return $OCF_SUCCESS
else
return $OCF_NOT_RUNNING
fi
else
# we have Xen 3.0.3 or lower
STATUS=`$xentool list --long $1 2>/dev/null | grep state 2>/dev/null`
echo "${STATUS}" | grep -qs "[-r][-b][-p]---"
if [ $? -ne 0 ]; then
return $OCF_NOT_RUNNING
else
return $OCF_SUCCESS
fi
fi
}
# If the guest is rebooting, it may completely disappear from the
# list of defined guests, thus xl/xen-list would return with not
# running; apparently, this period lasts only for a second or
# two
# If a status returns not running, then test status
# again for 5 times (perhaps it'll show up)
Xen_Status_with_Retry() {
local rc cnt=5
Xen_Status $1
rc=$?
while [ $rc -eq $OCF_NOT_RUNNING -a $cnt -gt 0 ]; do
case "$__OCF_ACTION" in
stop)
ocf_log debug "domain $1 reported as not running, waiting $cnt seconds ..."
;;
monitor)
ocf_log warn "domain $1 reported as not running, but it is expected to be running! Retrying for $cnt seconds ..."
;;
*) : not reachable
;;
esac
sleep 1
Xen_Status $1
rc=$?
cnt=$((cnt-1))
done
return $rc
}
Xen_Adjust_Memory() {
if ocf_is_true "${OCF_RESKEY_allow_mem_management}"; then
CNTNEW=$1
RUNNING=`Xen_List_running`
RUNCNT=`Xen_Count_running`
MAXMEM=`Xen_Total_Memory`
if [ ${RUNCNT} -eq 0 -a ${CNTNEW} -eq 0 ]; then
RUNCNT=1
fi
#NEWMEM=`echo "(${MAXMEM}-${OCF_RESKEY_reserved_Dom0_memory})/(${RUNCNT}+${CNTNEW})"|bc`
NEWMEM=$(( (${MAXMEM} - ${OCF_RESKEY_reserved_Dom0_memory}) / (${RUNCNT} + ${CNTNEW} ) ))
# do not rely on ballooning add dom0_mem=512 instead to force memory for dom0
#$xentool mem-set Domain-0 ${OCF_RESKEY_reserved_Dom0_memory}
for DOM in ${RUNNING}; do
$xentool mem-set ${DOM} ${NEWMEM}
done
ocf_log info "Adjusted memory to: $NEWMEM, for the following $RUNCNT domains: $RUNNING"
fi
}
Xen_List_all() {
$xentool list | grep -v -e "Name" -e "Domain-0" | awk '{print $1}'
}
Xen_List_running() {
ALL_DOMS=`Xen_List_all`
for DOM in ${ALL_DOMS}; do
if Xen_Status $DOM; then
echo "${DOM} "
fi
done
}
Xen_Count_running() {
Xen_List_running | wc -w
}
Xen_Monitor() {
if ocf_is_probe; then
Xen_Status ${DOMAIN_NAME}
else
Xen_Status_with_Retry ${DOMAIN_NAME}
fi
if [ $? -eq ${OCF_NOT_RUNNING} ]; then
ocf_is_probe ||
ocf_log err "Xen domain $DOMAIN_NAME stopped"
return ${OCF_NOT_RUNNING}
fi
if [ "X${OCF_RESKEY_monitor_scripts}" = "X" ]; then
return ${OCF_SUCCESS}
fi
for SCRIPT in ${OCF_RESKEY_monitor_scripts}; do
$SCRIPT
if [ $? -ne 0 ]; then
return ${OCF_ERR_GENERIC}
fi
done
return ${OCF_SUCCESS}
}
Xen_Total_Memory() {
$xentool info | grep "^total_memory" | awk '{print $3}'
}
Xen_Start() {
if Xen_Status ${DOMAIN_NAME}; then
ocf_log info "Xen domain $DOMAIN_NAME already running."
return $OCF_SUCCESS
fi
if [ ! -f "${OCF_RESKEY_xmfile}" ]; then
ocf_log err "Config file ${OCF_RESKEY_xmfile} for $DOMAIN_NAME does not exist."
return $OCF_ERR_INSTALLED
fi
if ocf_is_true "${OCF_RESKEY_allow_mem_management}"; then
Xen_Adjust_Memory 1
ocf_log info "New memory for virtual domains: ${NEWMEM}"
sed -i -e "/^memory=/ s/^memory=.*/memory=${NEWMEM}/" ${OCF_RESKEY_xmfile}
$xentool mem-set ${DOMAIN_NAME} ${NEWMEM}
fi
# the latest xl management tool is squeamish about some
# characters in a name (the vm name is xen-f):
# /etc/xen/vm/xen-f:15: config parsing error near `xen':
# syntax error, unexpected IDENT, expecting STRING or NUMBER
# or '['
# /etc/xen/vm/xen-f:15: config parsing error near `-f': lexical error
#
# the older xm management tool cannot digest quotes (see
# https://developerbugs.linuxfoundation.org/show_bug.cgi?id=2671)
#
# hence the following
if expr "x$xentool" : "x.*xl" >/dev/null; then
$xentool create ${OCF_RESKEY_xmfile} name=\"$DOMAIN_NAME\"
else
$xentool create ${OCF_RESKEY_xmfile} name="$DOMAIN_NAME"
fi
rc=$?
if [ $rc -ne 0 ]; then
return $OCF_ERR_GENERIC
else
if ocf_is_true "${OCF_RESKEY_allow_mem_management}"; then
$xentool mem-set ${DOMAIN_NAME} ${NEWMEM}
fi
fi
while sleep 1; do
Xen_Monitor && return $OCF_SUCCESS
done
}
xen_domain_stop() {
local dom=$1
local timeout
if [ -n "$OCF_RESKEY_shutdown_timeout" ]; then
timeout=$OCF_RESKEY_shutdown_timeout
elif [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then
# Allow 2/3 of the action timeout for the orderly shutdown
# (The origin unit is ms, hence the conversion)
timeout=$((OCF_RESKEY_CRM_meta_timeout/1500))
else
timeout=60
fi
if [ "$timeout" -gt 0 ]; then
ocf_log info "Xen domain $dom will be stopped (timeout: ${timeout}s)"
if ocf_is_true "${OCF_RESKEY_shutdown_acpi}"; then
$xentool trigger $dom power
else
$xentool shutdown $dom
fi
while Xen_Status $dom && [ "$timeout" -gt 0 ]; do
ocf_log debug "$dom still not stopped. Waiting..."
timeout=$((timeout-1))
sleep 1
done
fi
if [ "$timeout" -eq 0 ]; then
while Xen_Status $dom; do
ocf_log warn "Xen domain $dom will be destroyed!"
$xenkill $dom
sleep 1
done
# Note: This does not give up. stop isn't allowed to to fail.
# If $xentool destroy fails, stop will eventually timeout.
# This is the correct behaviour.
fi
ocf_log info "Xen domain $dom stopped."
}
Xen_Stop() {
local vm
if Xen_Status_with_Retry ${DOMAIN_NAME}; then
vm=${DOMAIN_NAME}
elif Xen_Status migrating-${DOMAIN_NAME}; then
ocf_log info "Xen domain $DOMAIN_NAME is migrating"
vm="migrating-${DOMAIN_NAME}"
else
ocf_log info "Xen domain $DOMAIN_NAME already stopped."
fi
if [ "$vm" ]; then
xen_domain_stop $vm
else
# It is supposed to be gone, but there have been situations where
# $xentool list / xen-list showed it as stopped but it was still
# instantiated. Nuke it once more to make sure:
$xenkill ${DOMAIN_NAME}
fi
Xen_Adjust_Memory 0
return $OCF_SUCCESS
}
Xen_Migrate_To() {
target_node="$OCF_RESKEY_CRM_meta_migrate_target"
target_attr="$OCF_RESKEY_node_ip_attribute"
target_addr="$target_node"
if Xen_Status ${DOMAIN_NAME}; then
ocf_log info "$DOMAIN_NAME: Starting $xentool migrate to $target_node"
if [ -n "$target_attr" ]; then
- nodevalue=`crm_attribute --type nodes --node-uname $target_node --attr-name $target_attr --get-value -q`
+ nodevalue=`crm_attribute --type nodes --node $target_node -n $target_attr -G -q`
if [ -n "${nodevalue}" -a "${nodevalue}" != "(null)" ]; then
target_addr="$nodevalue"
ocf_log info "$DOMAIN_NAME: $target_node is using address $target_addr"
fi
fi
if expr "x$xentool" : "x.*xm" >/dev/null; then
$xentool migrate --live $DOMAIN_NAME $target_addr
else
$xentool migrate $DOMAIN_NAME $target_addr
fi
rc=$?
if [ $rc -ne 0 ]; then
ocf_log err "$DOMAIN_NAME: $xentool migrate to $target_node failed: $rc"
return $OCF_ERR_GENERIC
else
Xen_Adjust_Memory 0
ocf_log info "$DOMAIN_NAME: $xentool migrate to $target_node succeeded."
return $OCF_SUCCESS
fi
else
ocf_log err "$DOMAIN_NAME: migrate_to: Not active locally!"
return $OCF_ERR_GENERIC
fi
}
Xen_Migrate_From() {
if [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then
# Allow 2/3 of the action timeout for status to stabilize
# (The origin unit is ms, hence the conversion)
timeout=$((OCF_RESKEY_CRM_meta_timeout/1500))
else
timeout=10 # should be plenty
fi
while ! Xen_Status ${DOMAIN_NAME} && [ $timeout -gt 0 ]; do
ocf_log debug "$DOMAIN_NAME: Not yet active locally, waiting (timeout: ${timeout}s)"
timeout=$((timeout-1))
sleep 1
done
if Xen_Status ${DOMAIN_NAME}; then
Xen_Adjust_Memory 0
ocf_log info "$DOMAIN_NAME: Active locally, migration successful"
return $OCF_SUCCESS
else
ocf_log err "$DOMAIN_NAME: Not active locally, migration failed!"
return $OCF_ERR_GENERIC
fi
}
Xen_Validate_All() {
return $OCF_SUCCESS
}
if [ $# -ne 1 ]; then
usage
exit $OCF_ERR_ARGS
fi
case $1 in
meta-data)
meta_data
exit $OCF_SUCCESS
;;
usage)
usage
exit $OCF_SUCCESS
;;
esac
# the name business:
#
# 1. use the name attribute, or
# 2. find the name in the config file (if it exists) and use that
# unless it contains funny characters such as '%' or space, or
# 3. use the OCF_RESOURCE_INSTANCE
if [ x"${OCF_RESKEY_name}" != x ]; then
DOMAIN_NAME="${OCF_RESKEY_name}"
else
if [ -f "${OCF_RESKEY_xmfile}" ]; then
DOMAIN_NAME=`awk '$1~/^name(=|$)/{print}' ${OCF_RESKEY_xmfile} | sed 's/.*=[[:space:]]*//' | tr -d "[\"']"`
if echo "$DOMAIN_NAME" | grep -qs '[%[:space:]]'; then
DOMAIN_NAME=""
fi
fi
DOMAIN_NAME=${DOMAIN_NAME:-${OCF_RESOURCE_INSTANCE}}
fi
for binary in sed awk; do
check_binary $binary
done
if have_binary xen-destroy ; then
xenkill="xen-destroy"
else
xenkill="$xentool destroy"
fi
if [ -n "$OCF_RESKEY_shutdown_timeout" ]; then
ocf_is_decimal "$OCF_RESKEY_shutdown_timeout" || {
ocf_log err "shutdown_timeout must be a number"
exit $OCF_ERR_CONFIGURED
}
fi
case $1 in
start)
Xen_Start
;;
stop)
Xen_Stop
;;
migrate_to)
Xen_Migrate_To
;;
migrate_from)
Xen_Migrate_From
;;
monitor)
Xen_Monitor
;;
status)
Xen_Status ${DOMAIN_NAME}
;;
validate-all)
Xen_Validate_All
;;
*)
usage
exit $OCF_ERR_UNIMPLEMENTED
;;
esac
exit $?
diff --git a/heartbeat/db2 b/heartbeat/db2
index f56ec3e71..69d0119aa 100755
--- a/heartbeat/db2
+++ b/heartbeat/db2
@@ -1,902 +1,902 @@
#!/bin/sh
#
# db2
#
# Resource agent that manages a DB2 LUW database in Standard role
# or HADR configuration in master/slave configuration.
# Multi partition is supported as well.
#
# Copyright (c) 2011 Holger Teutsch <holger.teutsch@web.de>
#
# This agent incoporates code of a previous release created by
# Alan Robertson and the community.
#
# 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.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
#
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
#######################################################################
db2_usage() {
echo "db2 start|stop|monitor|promote|demote|notify|validate-all|meta-data"
}
db2_meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="db2">
<version>1.0</version>
<longdesc lang="en">
Resource Agent that manages an IBM DB2 LUW databases in Standard role as primitive or in HADR roles in master/slave configuration. Multiple partitions are supported.
Standard mode:
An instance including all or selected databases is made highly available.
Configure each partition as a separate primitive resource.
HADR mode:
A single database in HADR configuration is made highly available by automating takeover operations.
Configure a master / slave resource with notifications enabled and an
additional monitoring operation with role "Master".
In case of HADR be very deliberate in specifying intervals/timeouts. The detection of a failure including promote must complete within HADR_PEER_WINDOW.
In addition to honoring requirements for crash recovery etc. for your specific database use the following relations as guidance:
"monitor interval" &lt; HADR_PEER_WINDOW - (appr 30 sec)
"promote timeout" &lt; HADR_PEER_WINDOW + (appr 20 sec)
For further information and examples consult http://www.linux-ha.org/wiki/db2_(resource_agent)
</longdesc>
<shortdesc lang="en">Resource Agent that manages an IBM DB2 LUW databases in Standard role as primitive or in HADR roles as master/slave configuration. Multiple partitions are supported.</shortdesc>
<parameters>
<parameter name="instance" unique="1" required="1">
<longdesc lang="en">
The instance of the database(s).
</longdesc>
<shortdesc lang="en">instance</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="dblist" unique="0" required="0">
<longdesc lang="en">
List of databases to be managed, e.g "db1 db2".
Defaults to all databases in the instance. Specify one db for HADR mode.
</longdesc>
<shortdesc lang="en">List of databases to be managed</shortdesc>
<content type="string"/>
</parameter>
<parameter name="admin" unique="0" required="0">
<longdesc lang="en">
DEPRECATED: The admin user of the instance.
</longdesc>
<shortdesc lang="en">DEPRECATED: admin</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="dbpartitionnum" unique="0" required="0">
<longdesc lang="en">
The number of the partion (DBPARTITIONNUM) to be managed.
</longdesc>
<shortdesc lang="en">database partition number (DBPARTITIONNUM)</shortdesc>
<content type="string" default="0" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="120"/>
<action name="stop" timeout="120"/>
<action name="promote" timeout="120"/>
<action name="demote" timeout="120"/>
<action name="notify" timeout="10"/>
<action name="monitor" depth="0" timeout="60" interval="20"/>
<action name="monitor" depth="0" timeout="60" role="Master" interval="22"/>
<action name="validate-all" timeout="5"/>
<action name="meta-data" timeout="5"/>
</actions>
</resource-agent>
END
}
#
# validate
# .. and set global variables
#
# exit on error
#
db2_validate() {
local db2home db2sql db2instance
# db2 uses korn shell
check_binary "ksh"
# check required instance vars
if [ -z "$OCF_RESKEY_instance" ]
then
ocf_log err "DB2 required parameter instance is not set!"
return $OCF_ERR_CONFIGURED
fi
instance=$OCF_RESKEY_instance
if [ -n "$OCF_RESKEY_admin" ]
then
ocf_log warn "DB2 deprecated parameter admin is set, using $OCF_RESKEY_admin as instance."
instance=$OCF_RESKEY_admin
fi
db2node=${OCF_RESKEY_dbpartitionnum:-0}
db2home=$(sh -c "echo ~$instance")
db2sql=$db2home/sqllib
db2profile=$db2sql/db2profile
db2bin=$db2sql/bin
STATE_FILE=${HA_RSCTMP}/db2-${OCF_RESOURCE_INSTANCE}.state
# Let's make sure a few important things are there...
if ! [ -d "$db2sql" -a -d "$db2bin" -a -f "$db2profile" -a \
-x "$db2profile" -a -x "$db2bin/db2" ]
then
ocf_is_probe && exit $OCF_NOT_RUNNING
ocf_log err "DB2 required directories and/or files not found"
exit $OCF_ERR_INSTALLED
fi
db2instance=$(runasdb2 'echo $DB2INSTANCE')
if [ "$db2instance" != "$instance" ]
then
ocf_is_probe && exit $OCF_NOT_RUNNING
ocf_log err "DB2 parameter instance \"$instance\" != DB2INSTANCE \"$db2instance\""
exit $OCF_ERR_CONFIGURED
fi
# enough checking for stop to succeed
[ $__OCF_ACTION = stop ] && return $OCF_SUCCESS
dblist=$OCF_RESKEY_dblist
if [ -n "$dblist" ]
then
# support , as separator as well
dblist=$(echo "$dblist" | sed -e 's/[,]/ /g')
else
if ! dblist=$(db2_dblist)
then
ocf_log err "DB2 $instance($db2node): cannot retrieve db directory"
exit $OCF_ERR_INSTALLED
fi
fi
# check requirements for the HADR case
if ocf_is_ms
then
set -- $dblist
if [ $# != 1 ]
then
ocf_log err "DB2 resource $OCF_RESOURCE_INSTANCE must have exactly one name in dblist"
exit $OCF_ERR_CONFIGURED
fi
if [ $db2node != 0 ]
then
ocf_log err "DB2 resource $OCF_RESOURCE_INSTANCE must have dbpartitionnum=0"
exit $OCF_ERR_CONFIGURED
fi
fi
return $OCF_SUCCESS
}
master_score()
{
if ! have_binary "crm_master"; then
return
fi
crm_master $*
}
#
# Run the given command as db2 instance user
#
runasdb2() {
su $instance -c ". $db2profile; $*"
}
#
# Run a command as the DB2 admin, and log the output
#
logasdb2() {
local output rc
output=$(runasdb2 $*)
rc=$?
if [ $rc -eq 0 ]
then
ocf_log info "$output"
else
ocf_log err "$output"
fi
return $rc
}
#
# maintain the fal (first active log) attribute
# db2_fal_attrib DB {set val|get}
#
db2_fal_attrib() {
local db=$1
local attr val rc id node member me
attr=db2hadr_${instance}_${db}_fal
case "$2" in
set)
me=$(uname -n)
# loop over all member nodes and set attribute
crm_node -l |
while read id node member
do
[ "$member" = member -a "$node" != "$me" ] || continue
crm_attribute -t nodes -l reboot --node=$node -n $attr -v "$3"
rc=$?
ocf_log info "DB2 instance $instance($db2node/$db: setting attrib for FAL to $FIRST_ACTIVE_LOG @ $node"
[ $rc != 0 ] && break
done
;;
get)
- crm_attribute -t nodes -l reboot -n $attr -G -Q 2>&1
+ crm_attribute -t nodes -l reboot -n $attr -G --quiet 2>&1
rc=$?
if [ $rc != 0 ]
then
ocf_log warn "DB2 instance $instance($db2node/$db: can't retrieve attribute $attr, are you sure notifications are enabled ?"
fi
;;
*)
exit $OCF_ERR_CONFIGURED
esac
return $rc
}
#
# unfortunately a first connect after a crash may need several minutes
# for some internal cleanup stuff in DB2.
# We run a connect in background so other connects (i.e. monitoring!) may proceed.
#
db2_run_connect() {
local db=$1
logasdb2 "db2 connect to $db; db2 terminate"
}
#
# get some data from the database config
# sets HADR_ROLE HADR_TIMEOUT HADR_PEER_WINDOW
#
db2_get_cfg() {
local db=$1
local output hadr_vars
output=$(runasdb2 db2 get db cfg for $db)
[ $? != 0 ] && return $OCF_ERR_GENERIC
hadr_vars=$(echo "$output" |
awk '/HADR database role/ {printf "HADR_ROLE='%s'; ", $NF;}
/HADR_TIMEOUT/ {printf "HADR_TIMEOUT='%s'; ", $NF;}
/First active log file/ {printf "FIRST_ACTIVE_LOG='%s'\n", $NF;}
/HADR_PEER_WINDOW/ {printf "HADR_PEER_WINDOW='%s'\n", $NF;}')
# sets HADR_ROLE HADR_TIMEOUT HADR_PEER_WINDOW
eval $hadr_vars
# HADR_PEER_WINDOW comes with V9 and is checked later
if [ -z "$HADR_ROLE" -o -z "$HADR_TIMEOUT" ]
then
ocf_log error "DB2 cfg values invalid for $instance($db2node)/$db: $hadr_vars"
return $OCF_ERR_GENERIC
fi
return $OCF_SUCCESS
}
#
# return the list of databases in the instance
#
db2_dblist() {
local output
output=$(runasdb2 db2 list database directory) || return $OCF_ERR_GENERIC
echo "$output" | grep -i 'Database name.*=' | sed 's%.*= *%%'
}
#
# Delayed check of the compatibility of DB2 instance and pacemaker
# config.
# Logically this belongs to validate but certain parameters can only
# be retrieved once the instance is started.
#
db2_check_config_compatibility() {
local db=$1
local is_ms
ocf_is_ms
is_ms=$?
case "$HADR_ROLE/$is_ms" in
STANDARD/0)
ocf_log err "DB2 database $instance/$db is not in a HADR configuration but I am a M/S resource"
exit $OCF_ERR_INSTALLED
;;
STANDARD/1)
# OK
;;
*/0)
if [ -z "$HADR_PEER_WINDOW" ]
then
ocf_log err "DB2 database $instance: release to old, need HADR_PEER_WINDOW (>=V9)"
exit $OCF_ERR_INSTALLED
fi
;;
*/1)
ocf_log err "DB2 database $instance/$db is in a HADR configuration but I must be a M/S resource"
esac
}
#
# Start instance and DB.
# Standard mode is through "db2 activate" in order to start in previous
# mode (Standy/Primary).
# If the database is a primary AND we can determine that the running master
# has a higher "first active log" we conclude that we come up after a crash
# an the previous Standby is now Primary.
# The db is then started as Standby.
#
# Other cases: danger of split brain, log error and do nothing.
#
db2_start() {
local output start_cmd db
local start_opts="dbpartitionnum $db2node"
# If we detect that db partitions are not in use, and no
# partition is explicitly specified, activate without
# partition information. This allows db2 instances without
# partition support to be managed.
if [ -z "$OCF_RESKEY_dbpartitionnum" ] && ! [ -a "$db2sql/db2nodes.cfg" ]; then
start_opts=""
fi
if output=$(runasdb2 db2start $start_opts)
then
ocf_log info "DB2 instance $instance($db2node) started: $output"
else
case $output in
*SQL1026N*)
ocf_log info "DB2 instance $instance($db2node) already running: $output"
;;
*)
ocf_log err "$output"
return $OCF_ERR_GENERIC
esac
fi
if ! db2_instance_status
then
ocf_log err "DB2 instance $instance($db2node) is not active!"
return $OCF_ERR_GENERIC
fi
[ $db2node = 0 ] || return $OCF_SUCCESS
# activate DB only on node 0
for db in $dblist
do
# sets HADR_ROLE HADR_TIMEOUT HADR_PEER_WINDOW FIRST_ACTIVE_LOG
db2_get_cfg $db || return $?
# Better late than never: can only check this when the instance is already up
db2_check_config_compatibility $db
start_cmd="db2 activate db $db"
if [ $HADR_ROLE = PRIMARY ]
then
local master_fal
# communicate our FAL to other nodes the might start concurrently
db2_fal_attrib $db set $FIRST_ACTIVE_LOG
# ignore false positive:
# error: Can't use > in [ ]. Escape it or use [[..]]. [SC2073]
# see https://github.com/koalaman/shellcheck/issues/691
# shellcheck disable=SC2073
if master_fal=$(db2_fal_attrib $db get) && [ "$master_fal" '>' $FIRST_ACTIVE_LOG ]
then
ocf_log info "DB2 database $instance($db2node)/$db is Primary and outdated, starting as secondary"
start_cmd="db2 start hadr on db $db as standby"
HADR_ROLE=STANDBY
fi
fi
if output=$(runasdb2 $start_cmd)
then
ocf_log info "DB2 database $instance($db2node)/$db started/activated"
[ $HADR_ROLE != STANDBY ] && db2_run_connect $db &
else
case $output in
SQL1490W*|SQL1494W*|SQL1497W*|SQL1777N*)
ocf_log info "DB2 database $instance($db2node)/$db already activated: $output"
;;
SQL1768N*"Reason code = \"7\""*)
ocf_log err "DB2 database $instance($db2node)/$db is a Primary and the Standby is down"
ocf_log err "Possible split brain ! Manual intervention required."
ocf_log err "If this DB is outdated use \"db2 start hadr on db $db as standby\""
ocf_log err "If this DB is the surviving primary use \"db2 start hadr on db $db as primary by force\""
# might be the Standby is not yet there
# might be a timing problem because "First active log" is delayed
# on the next start attempt we might succeed when FAL was advanced
# might be manual intervention is required
# ... so let pacemaker give it another try and we will succeed then
return $OCF_ERR_GENERIC
;;
*)
ocf_log err "DB2 database $instance($db2node)/$db didn't start: $output"
return $OCF_ERR_GENERIC
esac
fi
done
# come here with success
# Even if we are a db2 Primary pacemaker requires start to end up in slave mode
echo SLAVE > $STATE_FILE
return $OCF_SUCCESS
}
#
# helper function to be spawned
# so we can detect a hang of the db2stop command
#
db2_stop_bg() {
local rc output
local stop_opts="dbpartitionnum $db2node"
rc=$OCF_SUCCESS
if [ -z "$OCF_RESKEY_dbpartitionnum" ] && ! [ -a "$db2sql/db2nodes.cfg" ]; then
stop_opts=""
fi
if output=$(runasdb2 db2stop force $stop_opts)
then
ocf_log info "DB2 instance $instance($db2node) stopped: $output"
else
case $output in
*SQL1032N*)
#SQL1032N No start database manager command was issued
ocf_log info "$output"
;;
*)
ocf_log err "DB2 instance $instance($db2node) stop failed: $output"
rc=$OCF_ERR_GENERIC
esac
fi
return $rc
}
#
# Stop the given db2 database instance
#
db2_stop() {
local stop_timeout grace_timeout stop_bg_pid i must_kill
# remove master score
master_score -D -l reboot
# be very early here in order to avoid stale data
rm -f $STATE_FILE
db2_instance_status
if [ $? -eq $OCF_NOT_RUNNING ]; then
ocf_log info "DB2 instance $instance already stopped"
return $OCF_SUCCESS
fi
stop_timeout=${OCF_RESKEY_CRM_meta_timeout:-20000}
# grace_time is 4/5 (unit is ms)
grace_timeout=$((stop_timeout/1250))
# start db2stop in background as this may hang
db2_stop_bg &
stop_bg_pid=$!
# wait for grace_timeout
i=0
while [ $i -lt $grace_timeout ]
do
kill -0 $stop_bg_pid 2>/dev/null || break;
sleep 1
i=$((i+1))
done
# collect exit status but don't hang
if kill -0 $stop_bg_pid 2>/dev/null
then
stoprc=1
kill -9 $stop_bg_pid 2>/dev/null
else
wait $stop_bg_pid
stoprc=$?
fi
must_kill=0
if [ $stoprc -ne 0 ]
then
ocf_log warn "DB2 instance $instance($db2node): db2stop failed, using db2nkill"
must_kill=1
elif ! db2_instance_dead
then
ocf_log warn "DB2 instance $instance($db2node): db2stop indicated success but there a still processes, using db2nkill"
must_kill=1
fi
if [ $must_kill -eq 1 ]
then
# db2nkill kills *all* partions on the node
if [ -x $db2bin/db2nkill ]
then
logasdb2 $db2bin/db2nkill $db2node
elif [ -x $db2bin/db2_kill ]
then
logasdb2 $db2bin/db2_kill
fi
# loop forever (or lrmd kills us due to timeout) until the
# instance is dead
while ! db2_instance_dead
do
ocf_log info "DB2 instance $instance($db2node): waiting for processes to exit"
sleep 1
done
ocf_log info "DB2 instance $instance($db2node) is now dead"
fi
return $OCF_SUCCESS
}
#
# check whether `enough´ processes for a healthy instance are up
#
db2_instance_status() {
local pscount
pscount=$(runasdb2 $db2bin/db2nps $db2node | cut -c9- | grep ' db2[^ ]' | wc -l)
if [ $pscount -ge 4 ]; then
return $OCF_SUCCESS;
elif [ $pscount -ge 1 ]; then
return $OCF_ERR_GENERIC
fi
return $OCF_NOT_RUNNING
}
#
# is the given db2 instance dead?
#
db2_instance_dead() {
local pscount
pscount=$(runasdb2 $db2bin/db2nps $db2node | cut -c9- | grep ' db2[^ ]' | wc -l)
test $pscount -eq 0
}
#
# return the status of the db as "Role/Status"
# e.g. Primary/Peer, Standby/RemoteCatchupPending
#
# If not in HADR configuration return "Standard/Standalone"
#
db2_hadr_status() {
local db=$1
local output
output=$(runasdb2 db2pd -hadr -db $db)
if [ $? != 0 ]
then
echo "Down/Off"
return 1
fi
echo "$output" |
awk '/^\s+HADR_(ROLE|STATE) =/ {printf $3"/"}
/^\s+HADR_CONNECT_STATUS =/ {print $3; exit; }
/^HADR is not active/ {print "Standard/Standalone"; exit; }
/^Role *State */ {getline; printf "%s/%s\n", $1, $2; exit; }'
}
#
# Monitor the db
# And as side effect set crm_master / FAL attribute
#
db2_monitor() {
local CMD output hadr db
local rc
db2_instance_status
rc=$?
if [ $rc -ne $OCF_SUCCESS ]; then
# instance is dead remove master score
master_score -D -l reboot
exit $rc
fi
[ $db2node = 0 ] || return 0
# monitoring only for partition 0
for db in $dblist
do
hadr=$(db2_hadr_status $db) || return $OCF_ERR_GENERIC
ocf_log debug "Monitor: DB2 database $instance($db2node)/$db has HADR status $hadr"
# set master preference accordingly
case "$hadr" in
PRIMARY/*|Primary/*|Standard/*)
# perform a basic health check
CMD="if db2 connect to $db;
then
db2 select \* from sysibm.sysversions ; rc=\$?;
db2 terminate;
else
rc=\$?;
fi;
exit \$rc"
if ! output=$(runasdb2 $CMD)
then
case "$output" in
SQL1776N*)
# can't connect/select on standby, may be spurious turing takeover
;;
*)
ocf_log err "DB2 database $instance($db2node)/$db is not working"
ocf_log err "DB2 message: $output"
# dead primary, remove master score
master_score -D -l reboot
return $OCF_ERR_GENERIC
esac
fi
ocf_log debug "DB2 database $instance($db2node)/$db appears to be working"
ocf_is_ms && master_score -v 10000 -l reboot
;;
STANDBY/*PEER/*|Standby/*Peer)
master_score -v 8000 -l reboot
;;
STANDBY/*|Standby/*)
ocf_log warn "DB2 database $instance($db2node)/$db in status $hadr can never be promoted"
master_score -D -l reboot
;;
*)
return $OCF_ERR_GENERIC
esac
done
# everything OK, return if running as slave
grep MASTER $STATE_FILE >/dev/null 2>&1 || return $OCF_SUCCESS
return $OCF_RUNNING_MASTER
}
#
# Promote db to Primary
#
db2_promote() {
# validate ensured that dblist contains only one entry
local db=$dblist
local i hadr output force
# we run this twice as after a crash of the other node
# within HADR_TIMEOUT the status may be still reported as Peer
# although a connection no longer exists
for i in 1 2
do
hadr=$(db2_hadr_status $db) || return $OCF_ERR_GENERIC
ocf_log info "DB2 database $instance($db2node)/$db has HADR status $hadr and will be promoted"
case "$hadr" in
Standard/Standalone)
# this case only to keep ocf-tester happy
return $OCF_SUCCESS
;;
PRIMARY/PEER/*|PRIMARY/REMOTE_CATCHUP/*|Primary/Peer)
# nothing to do, only update pacemaker's view
echo MASTER > $STATE_FILE
return $OCF_SUCCESS
;;
STANDBY/PEER/CONNECTED|Standby/Peer)
# must take over
;;
STANDBY/*PEER/DISCONNECTED|Standby/DisconnectedPeer)
# must take over forced
force="by force peer window only"
;;
*)
return $OCF_ERR_GENERIC
esac
if output=$(runasdb2 db2 takeover hadr on db $db $force)
then
# update pacemaker's view
echo MASTER > $STATE_FILE
# turn the log so we rapidly get a new FAL
logasdb2 "db2 archive log for db $db"
return $OCF_SUCCESS
fi
case "$output" in
SQL1770N*"Reason code = \"7\""*)
# expected, HADR_TIMEOUT is now expired
# go for the second try
continue
;;
*)
ocf_log err "DB2 database $instance($db2node)/$db promote failed: $output"
return $OCF_ERR_GENERIC
esac
done
return $OCF_ERR_GENERIC
}
#
# Demote db to standby
#
db2_demote() {
# validate ensured that dblist contains only one entry
local db=$dblist
local hadr
# house keeping, set pacemaker's view to slave
echo SLAVE > $STATE_FILE
hadr=$(db2_hadr_status $dblist) || return $OCF_ERR_GENERIC
ocf_log info "DB2 database $instance($db2node)/$db has HADR status $hadr and will be demoted"
db2_monitor
return $?
}
#
# handle pre start notification
# We record our first active log on the other nodes.
# If two primaries come up after a crash they can safely determine who is
# the outdated one.
#
db2_notify() {
local node
# only interested in pre-start
[ $OCF_RESKEY_CRM_meta_notify_type = pre \
-a $OCF_RESKEY_CRM_meta_notify_operation = start ] || return $OCF_SUCESS
# gets FIRST_ACTIVE_LOG
db2_get_cfg $dblist || return $?
db2_fal_attrib $dblist set $FIRST_ACTIVE_LOG || return $OCF_ERR_GENERIC
exit $OCF_SUCCESS
}
########
# Main #
########
case "$__OCF_ACTION" in
meta-data)
db2_meta_data
exit $OCF_SUCCESS
;;
usage)
db2_usage
exit $OCF_SUCCESS
;;
start)
db2_validate
db2_start || exit $?
db2_monitor
exit $?
;;
stop)
db2_validate
db2_stop
exit $?
;;
promote)
db2_validate
db2_promote
exit $?
;;
demote)
db2_validate
db2_demote
exit $?
;;
notify)
db2_validate
db2_notify
exit $?
;;
monitor)
db2_validate
db2_monitor
exit $?
;;
validate-all)
db2_validate
exit $?
;;
*)
db2_usage
exit $OCF_ERR_UNIMPLEMENTED
esac
diff --git a/heartbeat/galera b/heartbeat/galera
index ee8451427..5c3c80d9d 100755
--- a/heartbeat/galera
+++ b/heartbeat/galera
@@ -1,975 +1,975 @@
#!/bin/sh
#
# 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 Master
# resource.
#
# Slave vs Master role:
#
# During the 'Slave' 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 Master 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 Master. Once the first
# Master instance bootstraps the galera cluster, the other nodes will be
# promoted to Master 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 master-max=3 --master
#
# By setting the 'enable_creation' option, the database will be automatically
# generated at startup. The meta attribute 'master-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
. ${OCF_FUNCTIONS_DIR}/mysql-common.sh
NODENAME=$(ocf_attribute_target)
# 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
#######################################################################
usage() {
cat <<UEND
usage: $0 (start|stop|validate-all|meta-data|monitor|promote|demote)
$0 manages a galera Database as an HA resource.
The 'start' operation starts the database.
The 'stop' operation stops the database.
The 'status' operation reports whether the database is running
The 'monitor' operation reports whether the database seems to be working
The 'promote' operation makes this mysql server run as master
The 'demote' operation makes this mysql server run as slave
The 'validate-all' operation reports whether the parameters are valid
UEND
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="galera">
<version>1.0</version>
<longdesc lang="en">
Resource script for managing galara database.
</longdesc>
<shortdesc lang="en">Manages a galara instance</shortdesc>
<parameters>
<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
Location of the MySQL server binary
</longdesc>
<shortdesc lang="en">MySQL server binary</shortdesc>
<content type="string" default="${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=""/>
</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=""/>
</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="root" />
</parameter>
<parameter name="check_passwd" unique="0" required="0">
<longdesc lang="en">
Cluster check user password
</longdesc>
<shortdesc lang="en">check password</shortdesc>
<content type="string" default="" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="120" />
<action name="stop" timeout="120" />
<action name="status" timeout="60" />
<action name="monitor" depth="0" timeout="30" interval="20" />
<action name="monitor" role="Master" depth="0" timeout="30" interval="10" />
<action name="monitor" role="Slave" depth="0" timeout="30" interval="30" />
<action name="promote" timeout="300" />
<action name="demote" timeout="120" />
<action name="validate-all" timeout="5" />
<action name="meta-data" timeout="5" />
</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" -Q 2>/dev/null
+ ${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" -Q 2>/dev/null
+ ${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" -Q 2>/dev/null
+ ${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" -Q 2>/dev/null
+ ${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" -Q 2>/dev/null
+ ${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" -Q 2>/dev/null
+ ${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
}
master_exists()
{
if [ "$__OCF_ACTION" = "demote" ]; then
# We don't want to detect master instances during demote.
# 1. we could be detecting ourselves as being master, which is no longer the case.
# 2. we could be detecting other master instances that are in the process of shutting down.
# by not detecting other master instances in "demote" we are deferring this check
# to the next recurring monitor operation which will be much more accurate
return 1
fi
# determine if a master instance is already up and is healthy
crm_mon --as-xml | grep "resource.*id=\"${INSTANCE_ATTR_NAME}\".*role=\"Master\".*active=\"true\".*orphaned=\"false\".*failed=\"false\"" > /dev/null 2>&1
return $?
}
clear_master_score()
{
local node=$(ocf_attribute_target $1)
if [ -z "$node" ]; then
$CRM_MASTER -D
else
$CRM_MASTER -D -N $node
fi
}
set_master_score()
{
local node=$(ocf_attribute_target $1)
if [ -z "$node" ]; then
$CRM_MASTER -v 100
else
$CRM_MASTER -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 $nodes; do
safe_to_bootstrap=$(get_safe_to_bootstrap $node)
if [ "$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_master_score $best_node
set_bootstrap_node $best_node
}
detect_safe_to_bootstrap()
{
local safe_to_bootstrap=""
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' < ${OCF_RESKEY_datadir}/grastate.dat)
fi
if [ "$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=$OCF_RESKEY_datadir \
--user=$OCF_RESKEY_user"
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'"
${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"
${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/' ${OCF_RESKEY_datadir}/grastate.dat
fi
ocf_log info "Node <${NODENAME}> is bootstrapping the cluster"
extra_opts="--wsrep-cluster-address=gcomm://"
else
ocf_exit_reason "Failure, Attempted to promote Master instance of $OCF_RESOURCE_INSTANCE before bootstrap node has been detected."
clear_last_commit
return $OCF_ERR_GENERIC
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 $OCF_ERR_CONFIGURED
fi
if [ -z "$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 $OCF_SUCCESS;;
esac
galera_validate
rc=$?
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 [ -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/redis b/heartbeat/redis
index 9aae69cda..fcd8c2345 100755
--- a/heartbeat/redis
+++ b/heartbeat/redis
@@ -1,626 +1,626 @@
#!/bin/bash
#
# Resource agent script for redis server.
#
# Copyright (c) 2013 Patrick Hemmer <patrick.hemmer@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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
#
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
: ${OCF_RESKEY_bin:=/usr/bin/redis-server}
: ${OCF_RESKEY_client_bin:=/usr/bin/redis-cli}
: ${OCF_RESKEY_user:=redis}
: ${OCF_RESKEY_rundir:=/var/run/redis}
: ${OCF_RESKEY_pidfile_name:=redis-server.pid}
: ${OCF_RESKEY_socket_name:=redis.sock}
: ${OCF_RESKEY_port:=6379}
if [ -z "$OCF_RESKEY_config" ]; then
if [ -f "/etc/redis.conf" ]; then
OCF_RESKEY_config="/etc/redis.conf"
else
OCF_RESKEY_config="/etc/redis/redis.conf"
fi
fi
CHECK_SLAVE_STATE=0
REDIS_CHECK_DUMP="/usr/bin/redis-check-dump"
REDIS_SERVER="$OCF_RESKEY_bin"
REDIS_CLIENT="$OCF_RESKEY_client_bin"
REDIS_CONFIG="$OCF_RESKEY_config"
REDIS_USER="$OCF_RESKEY_user"
REDIS_RUNDIR="$OCF_RESKEY_rundir"
REDIS_PIDFILE="$OCF_RESKEY_rundir/$OCF_RESKEY_pidfile_name"
REDIS_SOCKET="$OCF_RESKEY_rundir/$OCF_RESKEY_socket_name"
REDIS_REPLICATION_PORT="$OCF_RESKEY_port"
if ! [ -f $REDIS_CHECK_DUMP ]; then
REDIS_CHECK_DUMP="$(which redis-check-dump 2>/dev/null)"
fi
if [ -z "$REDIS_CHECK_DUMP" ]; then
REDIS_CHECK_DUMP="$(which redis-check-rdb 2>/dev/null)"
fi
if [ -r "$REDIS_CONFIG" ]; then
REDIS_DUMP_DIR="$(grep "^\s*dir\s" < "$REDIS_CONFIG" | awk '{ print $2 }' 2>/dev/null)"
REDIS_DUMP_FILE="$(grep "^\s*dbfilename\s" < "$REDIS_CONFIG" | awk '{ print $2 }' 2>/dev/null)"
fi
: ${REDIS_DUMP_DIR:=/var/lib/redis/}
: ${REDIS_DUMP_FILE:=dump.rdb}
redis_meta_data() {
cat <<EOI
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="redis">
<version>1.0</version>
<longdesc lang="en">
Resource agent script for redis server.
This resource fully supports master/slave replication. The master preference of a node is determined by the 'slave_priority' parameter of the redis config.
When taking the resource from 'unmanaged' to 'managed', the currently active master will be given a priority of 1000 (plus 1 for each active connection). The default 'slave_priority' is 100, so the master will stay master. For a slave to become master after converting the resource to managed, set a slave_priority greater than 1000.
</longdesc>
<shortdesc lang="en">Redis server</shortdesc>
<parameters>
<parameter name="bin" unique="0" required="0">
<longdesc lang="en">
Path to \`redis-server\`
</longdesc>
<shortdesc lang="en">Path to \`redis-server\`</shortdesc>
<content type="string" default="${OCF_RESKEY_bin}" />
</parameter>
<parameter name="client_bin" unique="0" required="0">
<longdesc lang="en">
Path to \`redis-cli\`
</longdesc>
<shortdesc lang="en">Path to \`redis-cli\`</shortdesc>
<content type="string" default="${OCF_RESKEY_client_bin}" />
</parameter>
<parameter name="config" unique="1" required="0">
<longdesc lang="en">
Path to 'redis.conf'
</longdesc>
<shortdesc lang="en">Path to 'redis.conf'</shortdesc>
<content type="string" default="${OCF_RESKEY_config}" />
</parameter>
<parameter name="user" unique="0" required="0">
<longdesc lang="en">
User to run redis as
</longdesc>
<shortdesc lang="en">Redis user</shortdesc>
<content type="string" default="${OCF_RESKEY_user}" />
</parameter>
<parameter name="rundir" unique="1" required="0">
<longdesc lang="en">
Directory to store socket and pid file in
</longdesc>
<shortdesc lang="en">Redis var/run dir</shortdesc>
<content type="string" default="${OCF_RESKEY_rundir}"/>
</parameter>
<parameter name="pidfile_name" unique="0" required="0">
<longdesc lang="en">
The filename to use for the pidfile. Will be created in the rundir.
Should only be a basename, not a full path.
</longdesc>
<shortdesc lang="en">Redis pidfile name</shortdesc>
<content type="string" default="${OCF_RESKEY_pidfile_name}"/>
</parameter>
<parameter name="socket_name" unique="0" required="0">
<longdesc lang="en">
The filename to use for the socket. Will be crated in the rundir.
Should only be a basename, not a full path.
</longdesc>
<shortdesc lang="en">Redis socket name</shortdesc>
<content type="string" default="${OCF_RESKEY_socket_name}"/>
</parameter>
<parameter name="port" unique="0" required="0">
<longdesc lang="en">
Port for replication client to connect to on remote server
</longdesc>
<shortdesc lang="en">Replication port</shortdesc>
<content type="string" default="${OCF_RESKEY_port}"/>
</parameter>
<parameter name="wait_last_known_master" unique="0" required="0">
<longdesc lang="en">
During redis cluster bootstrap, wait for the last known master to be
promoted before allowing any other instances in the cluster to be
promoted. This lessens the risk of data loss when persistent data
is in use.
</longdesc>
<shortdesc lang="en">Wait for last known master</shortdesc>
<content type="boolean" default="false"/>
</parameter>
</parameters>
<actions>
<action name="start" timeout="120" />
<action name="stop" timeout="120" />
<action name="status" timeout="60" />
<action name="monitor" depth="0" timeout="60" interval="45" />
<action name="monitor" role="Master" depth="0" timeout="60" interval="20" />
<action name="monitor" role="Slave" depth="0" timeout="60" interval="60" />
<action name="promote" timeout="120" />
<action name="demote" timeout="120" />
<action name="notify" timeout="90" />
<action name="validate-all" timeout="5" />
<action name="meta-data" timeout="5" />
</actions>
</resource-agent>
EOI
}
INSTANCE_ATTR_NAME=$(echo "${OCF_RESOURCE_INSTANCE}" | awk -F : '{print $1}')
CRM_ATTR_REPL_INFO="${HA_SBIN_DIR}/crm_attribute --type crm_config --name ${INSTANCE_ATTR_NAME}_REPL_INFO -s redis_replication"
MASTER_HOST=""
MASTER_ACTIVE_CACHED=""
MASTER_ACTIVE=""
master_is_active()
{
if [ -z "$MASTER_ACTIVE_CACHED" ]; then
# determine if a master instance is already up and is healthy
crm_mon --as-xml | grep "resource.*id=\"${OCF_RESOURCE_INSTANCE}\".*role=\"Master\".*active=\"true\".*orphaned=\"false\".*failed=\"false\"" > /dev/null 2>&1
MASTER_ACTIVE=$?
MASTER_ACTIVE_CACHED="true"
fi
return $MASTER_ACTIVE
}
set_master()
{
MASTER_HOST="$1"
${CRM_ATTR_REPL_INFO} -v "$1" -q
}
last_known_master()
{
if [ -z "$MASTER_HOST" ]; then
MASTER_HOST="$(${CRM_ATTR_REPL_INFO} --query -q 2>/dev/null)"
fi
echo "$MASTER_HOST"
}
crm_master_reboot() {
local node
node=$(ocf_attribute_target)
"${HA_SBIN_DIR}/crm_master" -N "$node" -l reboot "$@"
}
calculate_score()
{
perf_score="$1"
connected_clients="$2"
if ocf_is_true "$OCF_RESKEY_wait_last_known_master"; then
# only set perferred score by slave_priority if
# we are not waiting for the last known master. Otherwise
# we want the agent to have complete control over the scoring.
perf_score=""
connected_clients="0"
fi
if [[ -z "$perf_score" ]]; then
if [[ "$(last_known_master)" == "$NODENAME" ]]; then
perf_score=1000
else
perf_score=1
fi
fi
perf_score=$(( perf_score + connected_clients ))
echo "$perf_score"
}
set_score()
{
local score
local last_master
score="$1"
if ocf_is_true "$OCF_RESKEY_wait_last_known_master" && ! master_is_active; then
last_master="$(last_known_master)"
if [ -n "$last_master" ] && [[ "$last_master" != "$NODENAME" ]]; then
ocf_log info "Postponing setting master score for ${NODENAME} until last known master instance [${last_master}] is promoted"
return
fi
fi
ocf_log debug "monitor: Setting master score to '$score'"
crm_master_reboot -v "$score"
}
redis_client() {
ocf_log debug "redis_client: '$REDIS_CLIENT' -s '$REDIS_SOCKET' $*"
if [ -n "$clientpasswd" ]; then
"$REDIS_CLIENT" -s "$REDIS_SOCKET" -a "$clientpasswd" "$@" | sed 's/\r//'
else
"$REDIS_CLIENT" -s "$REDIS_SOCKET" "$@" | sed 's/\r//'
fi
}
simple_status() {
local pid
if ! [ -f "$REDIS_PIDFILE" ]; then
return $OCF_NOT_RUNNING
fi
pid="$(<"$REDIS_PIDFILE")"
pidof "$REDIS_SERVER" | grep -q "\<$pid\>" || return $OCF_NOT_RUNNING
ocf_log debug "monitor: redis-server running under pid $pid"
return $OCF_SUCCESS
}
redis_monitor() {
local res
simple_status
res=$?
if (( res != OCF_SUCCESS )); then
return $res
fi
typeset -A info
while read line; do
[[ "$line" == "#"* ]] && continue
[[ "$line" != *":"* ]] && continue
IFS=':' read -r key value <<< "$line"
info[$key]="$value"
done < <(redis_client info)
if [[ -z "${info[role]}" ]]; then
ocf_log err "monitor: Could not get role from \`$REDIS_CLIENT -s $REDIS_SOCKET info\`"
return $OCF_ERR_GENERIC
fi
if ocf_is_ms; then
# Here we see if a score has already been set.
# If score isn't set we the redis setting 'slave_priority'.
# If that isn't set, we default to 1000 for a master, and 1 for slave.
# We then add 1 for each connected client
- score="$(crm_master_reboot --get-value --quiet 2>/dev/null)"
+ score="$(crm_master_reboot -G --quiet 2>/dev/null)"
if [[ -z "$score" ]]; then
score=$(calculate_score "${info[slave_priority]}" "${info[connected_clients]}")
set_score "$score"
fi
if [[ "${info[role]}" == "master" ]]; then
if ocf_is_probe; then
set_master "$NODENAME"
fi
return $OCF_RUNNING_MASTER
fi
if [ "$CHECK_SLAVE_STATE" -eq 1 ]; then
if [[ "${info[master_link_status]}" != "up" ]]; then
ocf_log info "monitor: Slave mode link has not yet been established (link=${info[master_link_status]})"
return $OCF_ERR_GENERIC
fi
if [[ "${info[master_host]}" != "$(last_known_master)" ]]; then
ocf_log err "monitor: Slave mode current master does not match running master. current=${info[master_host]}, running=$(last_known_master)"
return $OCF_ERR_GENERIC
fi
fi
fi
return $OCF_SUCCESS
}
check_dump_file()
{
if ! have_binary "$REDIS_CHECK_DUMP"; then
return 0
fi
$REDIS_CHECK_DUMP ${REDIS_DUMP_DIR}/${REDIS_DUMP_FILE} 2>&1
}
redis_start() {
local size
redis_monitor
status=$?
if (( status == OCF_SUCCESS )) || (( status == OCF_RUNNING_MASTER )); then
ocf_log info "start: redis is already running"
return $OCF_SUCCESS
fi
[[ ! -d "$REDIS_RUNDIR" ]] && mkdir -p "$REDIS_RUNDIR"
chown -R "$REDIS_USER" "$REDIS_RUNDIR"
if have_binary "restorecon"; then
restorecon -Rv "$REDIS_RUNDIR"
fi
# check for 0 byte database dump file. This is an unrecoverable start
# condition that we can avoid by deleting the 0 byte database file.
if [ -f "${REDIS_DUMP_DIR}/${REDIS_DUMP_FILE}" ]; then
size="$(stat --format "%s" ${REDIS_DUMP_DIR}/${REDIS_DUMP_FILE})"
if [ "$?" -eq "0" ] && [ "$size" -eq "0" ]; then
ocf_log notice "Detected 0 byte ${REDIS_DUMP_FILE}, deleting zero length file to avoid start failure."
rm -f "${REDIS_DUMP_DIR}/${REDIS_DUMP_FILE}"
fi
fi
ocf_log info "start: $REDIS_SERVER --daemonize yes --unixsocket '$REDIS_SOCKET' --pidfile '$REDIS_PIDFILE'"
output="$(su "$REDIS_USER" -s /bin/sh -c "cd '$REDIS_RUNDIR'; exec '$REDIS_SERVER' '$REDIS_CONFIG' --daemonize yes --unixsocket '$REDIS_SOCKET' --pidfile '$REDIS_PIDFILE'" 2>&1)"
while true; do
# wait for redis to start
typeset -A info
while read line; do
[[ "$line" == "#"* ]] && continue
[[ "$line" != *":"* ]] && continue
IFS=':' read -r key value <<< "$line"
info[$key]="$value"
done < <(redis_client info)
if (( info[loading] == 0 )); then
break
elif (( info[loading] == 1 )); then
sleep "${info[loading_eta_seconds]}"
elif pidof "$REDIS_SERVER" >/dev/null; then
# unknown error, but the process still exists.
# This check is mainly because redis daemonizes before it starts listening, causing `redis-cli` to fail
# See https://github.com/antirez/redis/issues/2368
# It's possible that the `pidof` will pick up a different redis, but in that case, the start operation will just time out
sleep 1
else
check_output="$(check_dump_file)"
ocf_log err "start: Unknown error waiting for redis to start. redis-check-dump output=${check_output//$'\n'/; }"
return $OCF_ERR_GENERIC
fi
done
while ! [ -s "$REDIS_PIDFILE" ]; do
ocf_log debug "start: Waiting for pid file '$REDIS_PIDFILE' to appear"
sleep 1
done
ocf_is_ms && redis_demote # pacemaker expects resources to start in slave mode
redis_monitor
status=$?
if (( status == OCF_SUCCESS )) || (( status == OCF_RUNNING_MASTER )); then
return $OCF_SUCCESS
fi
check_output="$(check_dump_file)"
ocf_log err "start: Unknown error starting redis. redis-server output=${output//$'\n'/; } redis-check-dump output=${check_output//$'\n'/; }"
return $status
}
redis_stop() {
redis_monitor
status=$?
if (( status == OCF_NOT_RUNNING )); then
ocf_log info "stop: redis is already stopped"
crm_master_reboot -D
return $OCF_SUCCESS
fi
pid="$(<"$REDIS_PIDFILE")"
kill -TERM "$pid"
while true; do
simple_status
status=$?
if (( status == OCF_NOT_RUNNING )); then
crm_master_reboot -D
return $OCF_SUCCESS
fi
sleep 1
done
}
redis_promote() {
redis_monitor
status=$?
if (( status == OCF_RUNNING_MASTER )); then
ocf_log info "promote: Already running as master"
set_master "$NODENAME"
return $OCF_SUCCESS
elif (( status != OCF_SUCCESS )); then
ocf_log err "promote: Node is not running as a slave"
return $OCF_ERR_GENERIC
fi
redis_client slaveof no one
redis_monitor
status=$?
if (( status == OCF_RUNNING_MASTER )); then
set_master "$NODENAME"
return $OCF_SUCCESS
fi
ocf_log err "promote: Unknown error while promoting to master (status=$status)"
return $OCF_ERR_GENERIC
}
redis_demote() {
local master_host
local master_port
# client kill is only supported in Redis 2.8.12 or greater
version=$(redis_client -v | awk '{print $NF}')
ocf_version_cmp "$version" "2.8.11"
client_kill=$?
CHECK_SLAVE_STATE=1
redis_monitor
status=$?
if (( status == OCF_SUCCESS )); then
ocf_log info "demote: Already running as slave"
return $OCF_SUCCESS
elif (( status == OCF_NOT_RUNNING )); then
ocf_log err "demote: Failed to demote, redis not running."
return $OCF_NOT_RUNNING
fi
master_host="$(last_known_master)"
master_port="${REDIS_REPLICATION_PORT}"
# The elected master has to remain a slave during startup.
# During this period a placeholder master host is assigned.
if [ -z "$master_host" ] || [[ "$master_host" == "$NODENAME" ]]; then
CHECK_SLAVE_STATE=0
master_host="no-such-master"
elif ! master_is_active; then
# no master has been promoted yet. we'll be notified when the
# master starts.
CHECK_SLAVE_STATE=0
master_host="no-such-master"
fi
ocf_log info "demote: Setting master to '$master_host'"
redis_client slaveof "$master_host" "$master_port"
# Wait forever for the slave to connect to the master and finish the
# sync. Timeout is controlled by Pacemaker "op start timeout=XX".
#
# hint: redis master_link_status will only come "up" when
# the SYNC with the master has completed.
# This can take an arbitraty time (data) and should
# only be parametrized by the start operation timeout
# by the administrator, not by this resource agent code
while true; do
# Wait infinite if replication is syncing
# Then start/demote operation timeout determines timeout
if [ "$client_kill" -eq 2 ]; then
redis_client CLIENT PAUSE 2000
fi
redis_monitor
status=$?
if (( status == OCF_SUCCESS )); then
if [ "$client_kill" -eq 2 ]; then
redis_client CLIENT KILL type normal
fi
return $OCF_SUCCESS
fi
sleep 1
done
ocf_log err "demote: Unexpected error setting slave mode (status=$status)"
return $OCF_ERR_GENERIC
}
redis_notify() {
mode="${OCF_RESKEY_CRM_meta_notify_type}-${OCF_RESKEY_CRM_meta_notify_operation}"
case "$mode" in
post-demote|post-promote) # change the master
redis_monitor
status=$?
if (( status == OCF_SUCCESS )); then # were a slave
# calling demote updates the slave's connection
# to the newly appointed Master instance.
redis_demote
fi
;;
esac
return $OCF_SUCCESS
}
redis_validate() {
if [[ -x "$REDIS_SERVER" ]]; then
ocf_log err "validate: $REDIS_SERVER does not exist or is not executable"
return $OCF_ERR_INSTALLED
fi
if [[ -x "$REDIS_CLIENT" ]]; then
ocf_log err "validate: $REDIS_CLIENT does not exist or is not executable"
return $OCF_ERR_INSTALLED
fi
if [[ -f "$REDIS_CONFIG" ]]; then
ocf_log err "validate: $REDIS_CONFIG does not exist"
return $OCF_ERR_CONFIGURED
fi
if ! getent passwd "$REDIS_USER" &>/dev/null; then
ocf_log err "validate: $REDIS_USER is not a valid user"
return $OCF_ERR_CONFIGURED
fi
}
NODENAME=$(ocf_attribute_target)
if [ -r "$REDIS_CONFIG" ]; then
clientpasswd="$(sed -n -e 's/^\s*requirepass\s*\(.*\)\s*$/\1/p' < $REDIS_CONFIG | tail -n 1)"
fi
ocf_log debug "action=${1:-$__OCF_ACTION} notify_type=${OCF_RESKEY_CRM_meta_notify_type} notify_operation=${OCF_RESKEY_CRM_meta_notify_operation} master_host=${OCF_RESKEY_CRM_meta_notify_master_uname} slave_host=${OCF_RESKEY_CRM_meta_notify_slave_uname} promote_host=${OCF_RESKEY_CRM_meta_notify_promote_uname} demote_host=${OCF_RESKEY_CRM_meta_notify_demote_uname}; params: bin=${OCF_RESKEY_bin} client_bin=${OCF_RESKEY_client_bin} config=${OCF_RESKEY_config} user=${OCF_RESKEY_user} rundir=${OCF_RESKEY_rundir} port=${OCF_RESKEY_port}"
case "${1:-$__OCF_ACTION}" in
status|monitor)
redis_monitor
;;
start)
redis_start
;;
stop)
redis_stop
;;
restart)
redis_stop && redis_start
;;
promote)
redis_promote
;;
demote)
redis_demote
;;
notify)
redis_notify
;;
meta-data)
redis_meta_data
;;
validate-all)
redis_validate
;;
*)
echo "Usage: $0 {monitor|start|stop|restart|promote|demote|notify|validate-all|meta-data}"
exit $OCF_ERR_UNIMPLEMENTED
;;
esac
status=$?
ocf_log debug "exit_status=$status"
exit $status
diff --git a/heartbeat/vmware b/heartbeat/vmware
index 3c728911c..e25c73d96 100755
--- a/heartbeat/vmware
+++ b/heartbeat/vmware
@@ -1,393 +1,393 @@
#!/bin/sh
#
# VMware OCF resource agent
#
# Copyright (c) 2010 Apra Sistemi s.r.l.
# All Rights Reserved.
#
# Description: Manages VMware server 2.0 virtual machines
# as High-Availability resources
#
#
# Author: Cristian Mammoli <c.mammoli AT apra DOT it>
# License: GNU General Public License (GPL)
# Copyright: (C) 2010 Apra Sistemi s.r.l.
#
# See usage() function below for more details...
#
# OCF instance parameters:
# * OCF_RESKEY_vmxpath (mandatory: full path to the virtual machine vmx file)
# * OCF_RESKEY_vimshbin (optional: full path to the vmware-vim-cmd executable,
# fallback to default location if not declared)
#
# Requirements/caveats:
# * vmware-server 2.0 installed and autostarted on all nodes
# * vmdk files must be in the same directory of the vmx file
# * vmx filenames must be unique, even if stored in different directories
-# * Default_Action_Timeout stock value (20 sec) isn't enough if you are
+# * The default value of operation timeout (20 sec) isn't enough if you are
# dealing with many virtual machines: raise it to something around 600 secs
# or use operation attributes with the proposed values
# * Moving a vm among nodes will cause its mac address to change: if you need
# to preserve the mac address set it manually in the nic options
# * The script should be able to deal with paths and filenames with spaces,
# anyway try to avoid it
# Initialization
#################################################################
# Source ocf shell functions
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
# Basic variables configuration
OCF_RESKEY_vimshbin_default="/usr/bin/vmware-vim-cmd"
: ${OCF_RESKEY_vimshbin=${OCF_RESKEY_vimshbin_default}}
#################################################################
# Path to the virtual machine configuration file
VMXPATH="$OCF_RESKEY_vmxpath"
# Path to the vmware-vim-cmd executable
VIMSHBIN="$OCF_RESKEY_vimshbin"
# Global variables
VMXDIR=
RELVMXPATH=
VMID=
VM=
VMAUTOMSG=
# vmware-vim-cmd functions
#################################################################
# Get virtual machine vid
vmware_get_vid() {
$VIMSHBIN vmsvc/getallvms \
| awk '/\/'"$1"'/ {print $1}'
}
# Is the vm waiting for input after a migration?
vmware_uuid_alt() {
$VIMSHBIN vmsvc/message $1 \
| awk /^msg.uuid.altered/
}
# Get message id
vmware_get_msgid() {
$VIMSHBIN vmsvc/message $1 \
| awk '/^Virtual machine message/ {print $4}' \
| awk -F : '{print $1}'
}
# Answers message
vmware_answer_msg() {
$VIMSHBIN vmsvc/message $1 $2 $3 >/dev/null
}
# Register a virtual machine
vmware_register_vm() {
$VIMSHBIN solo/registervm '"'$1'"' >/dev/null
}
# Unregister a virtual machine
vmware_unregister_vm() {
$VIMSHBIN vmsvc/unregister $1 >/dev/null
}
# Start a virtual machine
vmware_poweron_vm() {
$VIMSHBIN vmsvc/power.on $1 >/dev/null
}
# Suspend a virtual machine
vmware_suspend_vm() {
$VIMSHBIN vmsvc/power.suspend $1 >/dev/null
}
# Get virtual machine power state
vmware_get_status() {
$VIMSHBIN vmsvc/power.getstate $1 \
| awk '/^Powered on/ || /^Powered off/ || /^Suspended/'
}
# Get vid of missing virtual machines
vmware_get_broken() {
$VIMSHBIN vmsvc/getallvm 2>&1 \
| awk -F \' '/^Skipping/ {print $2}'
}
# Variables depending on the above functions
#################################################################
vmware_set_env() {
# Directory containing the virtual machine
VMXDIR="`dirname "$VMXPATH"`"
# Basename of the configuration file
RELVMXPATH="`basename "$VMXPATH"`"
# Vid of the virtual machine (can be empty if the vm is not registered)
VMID=`vmware_get_vid "$RELVMXPATH"`
# Virtual machine name
VM="`awk -F '"' '/^displayName/ {print $2}' "$VMXPATH"`"
# msg.autoAnswer value in config file
VMAUTOMSG="`awk -F '"' '/^msg.autoAnswer/ {print toupper($2)}' "$VMXPATH"`"
}
# Main functions
#################################################################
# Print usage summary
vmware_usage() {
cat <<END
usage: $0 {start|stop|status|monitor|meta-data|validate-all}
Expects to have a fully populated OCF RA-compliant environment set.
END
}
# Check for mandatory files presence and consistency
vmware_validate() {
if [ -z "`pidof vmware-hostd`" ]; then
ocf_log err "vmware-hostd is not running"
exit $OCF_ERR_GENERIC
fi
if [ ! -x "$VIMSHBIN" ]; then
ocf_log err "vmware-vim-cmd executable missing or not in path ($VIMSHBIN)"
exit $OCF_ERR_ARGS
fi
if [ ! -f "$VMXPATH" ]; then
ocf_log err "Specified vmx file ($VMXPATH) does not exist"
exit $OCF_ERR_ARGS
fi
# Now we can safely setup variables...
vmware_set_env
# ... and verify them
if [ -z "$VM" ]; then
ocf_log err "Could not find out virtual machine name"
exit $OCF_ERR_ARGS
fi
if [ "$VMAUTOMSG" != "TRUE" ]; then
ocf_log warn "Please set msg.autoAnswer = \"TRUE\" in your config file"
fi
# $VMID is allowed to be empty in case we are validating a
# virtual machine which is not registered
return $OCF_SUCCESS
}
# More relaxed checking in case of probes
vmware_validate_probe() {
if [ ! -x "$VIMSHBIN" ]; then
ocf_log warn "vmware-vim-cmd executable missing or not in path ($VIMSHBIN)"
exit $OCF_NOT_RUNNING
fi
if [ ! -f "$VMXPATH" ]; then
ocf_log warn "Specified vmx file ($VMXPATH) does not exist"
exit $OCF_NOT_RUNNING
fi
# Now we can safely setup variables...
vmware_set_env
}
# Start a virtual machine
vmware_start() {
# Don't start a VM if it's already running
if vmware_monitor; then
ocf_log info "Virtual machine $VM is already running"
return $OCF_SUCCESS
else
# Removes stale lockfiles and missing virtual machines
# in case of a crash.
# Do not use with a clustered filesystem or you could
# end up starting the same VM in more than one node
ocf_log info "Removing stale lockfiles"
find "$VMXDIR" -name \*.lck -type f -exec rm "{}" \;
for BVM in `vmware_get_broken`; do
ocf_log info "Unregistering missing virtual machine $BVM"
vmware_unregister_vm $BVM
done
if [ -z "$VMID" ]; then
# VM is not registered, need to register
ocf_log info "Virtual machine $VM is not registered"
ocf_log info "Registering Virtual machine $VM"
vmware_register_vm "$VMXPATH"
VMID=`vmware_get_vid "$RELVMXPATH"`
if [ -z "$VMID" ]; then
ocf_log err "Could not register virtual machine $VM"
exit $OCF_ERR_GENERIC
fi
ocf_log info "Virtual machine $VM registered with ID $VMID"
fi
ocf_log info "Powering on virtual machine $VM"
vmware_poweron_vm $VMID
# Give the VM some time to initialize
sleep 10
if [ "$VMAUTOMSG" != "TRUE" ]; then
# msg.autoAnswer is not set: we try to deal with the
# most common question: msg.uuid.altered
ocf_log info "Checking msg.uuid.altered on VM $VM"
if [ -n "`vmware_uuid_alt $VMID`" ]; then
MSGID=`vmware_get_msgid $VMID`
vmware_answer_msg $VMID $MSGID 2
fi
fi
# Check if the VM is running. We don't bother
# with timeouts: we rely on the CRM for that.
while :; do
vmware_monitor && break
ocf_log info "Virtual machine $VM is still stopped: delaying 10 seconds"
sleep 10
done
ocf_log info "Virtual machine $VM is running"
return $OCF_SUCCESS
fi
}
# Stop a virtual machine
vmware_stop() {
# Don't stop a VM if it's not registered
if [ -z "$VMID" ]; then
ocf_log info "Virtual machine $VM is not registered"
return $OCF_SUCCESS
else
# Don't stop a VM if it's already stopped
if vmware_monitor; then
# If the VM is running send a suspend signal and wait
# until it is off. We don't bother with timeouts: we
# rely on the CRM for that.
ocf_log info "Virtual machine $VM is running: suspending it"
vmware_suspend_vm $VMID
sleep 5
while vmware_monitor; do
ocf_log info "Virtual machine $VM is still running: delaying 10 seconds"
sleep 10
done
else
ocf_log info "Virtual machine $VM is already stopped"
fi
# VMware randomly fails to unregister VMs,
# so we loop until we have success or timeout
ocf_log info "Unregistering virtual machine $VM"
vmware_unregister_vm $VMID
VMID=`vmware_get_vid "$RELVMXPATH"`
while [ -n "$VMID" ]; do
ocf_log warn "Could not unregister virtual machine $VM: retrying."
sleep 10
vmware_unregister_vm $VMID
VMID=`vmware_get_vid "$RELVMXPATH"`
done
ocf_log info "Virtual machine $VM is stopped"
return $OCF_SUCCESS
fi
}
# Monitor a virtual machine
vmware_monitor() {
if [ -n "$VMID" ] && [ "`vmware_get_status $VMID`" = "Powered on" ]; then
ocf_log debug "Virtual machine $VM (ID $VMID) is running..."
return $OCF_SUCCESS
else
ocf_log debug "Virtual machine $VM is stopped/suspended/not registered"
return $OCF_NOT_RUNNING
fi
}
# Print metadata informations
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="vmware">
<version>0.2</version>
<longdesc lang="en">
OCF compliant script to control vmware server 2.0 virtual machines.
</longdesc>
<shortdesc lang="en">Manages VMWare Server 2.0 virtual machines</shortdesc>
<parameters>
<parameter name="vmxpath" unique="0" required="1">
<longdesc lang="en">
VMX configuration file path
</longdesc>
<shortdesc lang="en">VMX file path</shortdesc>
<content type="string"/>
</parameter>
<parameter name="vimshbin" unique="0" required="0">
<longdesc lang="en">
vmware-vim-cmd executable path
</longdesc>
<shortdesc lang="en">vmware-vim-cmd path</shortdesc>
<content type="string" default="${OCF_RESKEY_vimshbin_default}"/>
</parameter>
</parameters>
<actions>
<action name="start" timeout="600" />
<action name="stop" timeout="600" />
<action name="monitor" timeout="30" interval="300" depth="0"/>
<action name="meta-data" timeout="5" />
</actions>
</resource-agent>
END
}
# See how we were called
#################################################################
case $1 in
meta-data)
meta_data
exit $OCF_SUCCESS
;;
start)
vmware_validate
vmware_start
;;
stop)
vmware_validate
vmware_stop
;;
status|monitor)
if ocf_is_probe; then
vmware_validate_probe
else
vmware_validate
fi
vmware_monitor
;;
usage|help)
vmware_usage
exit $OCF_SUCCESS
;;
validate-all)
vmware_validate
;;
*)
vmware_usage
exit $OCF_ERR_UNIMPLEMENTED
;;
esac
exit $?
diff --git a/tools/sfex_daemon.c b/tools/sfex_daemon.c
index 516c10133..ee0c3147c 100644
--- a/tools/sfex_daemon.c
+++ b/tools/sfex_daemon.c
@@ -1,345 +1,345 @@
#include <config.h>
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>
#include <signal.h>
#include <limits.h>
#include <sys/mman.h>
#include <string.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <syslog.h>
#include "sfex.h"
#include "sfex_lib.h"
#if HAVE_GLUE_CONFIG_H
#include <glue_config.h> /* for HA_LOG_FACILITY */
#endif
static int sysrq_fd;
static int lock_index = 1; /* default 1st lock */
static time_t collision_timeout = 1; /* default 1 sec */
static time_t lock_timeout = 60; /* default 60 sec */
time_t unlock_timeout = 60;
static time_t monitor_interval = 10;
static sfex_controldata cdata;
static sfex_lockdata ldata;
static sfex_lockdata ldata_new;
static const char *device;
const char *progname;
char *nodename;
static const char *rsc_id = "sfex";
static void usage(FILE *dist) {
fprintf(dist, "usage: %s [-i <index>] [-c <collision_timeout>] [-t <lock_timeout>] <device>\n", progname);
}
static void acquire_lock(void)
{
if (read_lockdata(&cdata, &ldata, lock_index) == -1) {
cl_log(LOG_ERR, "read_lockdata failed in acquire_lock\n");
exit(EXIT_FAILURE);
}
if ((ldata.status == SFEX_STATUS_LOCK) && (strncmp(nodename, (const char*)(ldata.nodename), sizeof(ldata.nodename)))) {
unsigned int t = lock_timeout;
while (t > 0)
t = sleep(t);
read_lockdata(&cdata, &ldata_new, lock_index);
if (ldata.count != ldata_new.count) {
cl_log(LOG_ERR, "can\'t acquire lock: the lock's already hold by some other node.\n");
exit(2);
}
}
/* The lock acquisition is possible because it was not updated. */
ldata.status = SFEX_STATUS_LOCK;
ldata.count = SFEX_NEXT_COUNT(ldata.count);
strncpy((char*)(ldata.nodename), nodename, sizeof(ldata.nodename));
if (write_lockdata(&cdata, &ldata, lock_index) == -1) {
cl_log(LOG_ERR, "write_lockdata failed\n");
exit(EXIT_FAILURE);
}
/* detect the collision of lock */
/* The collision occurs when two or more nodes do the reservation
processing of the lock at the same time. It waits for collision_timeout
seconds to detect this,and whether the superscription of lock data by
another node is done is checked. If the superscription was done by
another node, the lock acquisition with the own node is given up.
*/
{
unsigned int t = collision_timeout;
while (t > 0)
t = sleep(t);
if (read_lockdata(&cdata, &ldata_new, lock_index) == -1) {
cl_log(LOG_ERR, "read_lockdata failed in collision detection\n");
}
if (strncmp((char*)(ldata.nodename), (const char*)(ldata_new.nodename), sizeof(ldata.nodename))) {
cl_log(LOG_ERR, "can\'t acquire lock: collision detected in the air.\n");
exit(2);
}
}
/* extension of lock */
/* Validly time of the lock is extended. It is because of spending at
the collision_timeout seconds to detect the collision. */
ldata.count = SFEX_NEXT_COUNT(ldata.count);
if (write_lockdata(&cdata, &ldata, lock_index) == -1) {
cl_log(LOG_ERR, "write_lockdata failed in extension of lock\n");
exit(EXIT_FAILURE);
}
cl_log(LOG_INFO, "lock acquired\n");
}
static void error_todo (void)
{
if (fork() == 0) {
- cl_log(LOG_INFO, "Execute \"crm_resource -F -r %s -H %s\" command\n", rsc_id, nodename);
- execl("/usr/sbin/crm_resource", "crm_resource", "-F", "-r", rsc_id, "-H", nodename, NULL);
+ cl_log(LOG_INFO, "Execute \"crm_resource -F -r %s --node %s\" command\n", rsc_id, nodename);
+ execl("/usr/sbin/crm_resource", "crm_resource", "-F", "-r", rsc_id, "--node", nodename, NULL);
} else {
exit(EXIT_FAILURE);
}
}
static void failure_todo(void)
{
#ifdef SFEX_TESTING
exit(EXIT_FAILURE);
#else
- /*execl("/usr/sbin/crm_resource", "crm_resource", "-F", "-r", rsc_id, "-H", nodename, NULL); */
+ /*execl("/usr/sbin/crm_resource", "crm_resource", "-F", "-r", rsc_id, "--node", nodename, NULL); */
int ret;
cl_log(LOG_INFO, "Force reboot node %s\n", nodename);
ret = write(sysrq_fd, "b\n", 2);
if (ret == -1) {
cl_log(LOG_ERR, "%s\n", strerror(errno));
}
close(sysrq_fd);
exit(EXIT_FAILURE);
#endif
}
static void update_lock(void)
{
/* read lock data */
if (read_lockdata(&cdata, &ldata, lock_index) == -1) {
cl_log(LOG_ERR, "read_lockdata failed in update_lock\n");
error_todo();
exit(EXIT_FAILURE);
}
/* check current lock status */
/* if own node is not locking, lock update is failed */
if (ldata.status != SFEX_STATUS_LOCK || strncmp((const char*)(ldata.nodename), nodename, sizeof(ldata.nodename))) {
cl_log(LOG_ERR, "can't update lock.\n");
failure_todo();
exit(EXIT_FAILURE);
}
/* lock update */
ldata.count = SFEX_NEXT_COUNT(ldata.count);
if (write_lockdata(&cdata, &ldata, lock_index) == -1) {
cl_log(LOG_ERR, "write_lockdata failed in update_lock\n");
error_todo();
exit(EXIT_FAILURE);
}
}
static void release_lock(void)
{
/* The only thing I care about in release_lock(), is to terminate the process */
/* read lock data */
if (read_lockdata(&cdata, &ldata, lock_index) == -1) {
cl_log(LOG_ERR, "read_lockdata failed in release_lock\n");
exit(EXIT_FAILURE);
}
/* check current lock status */
/* if own node is not locking, we judge that lock has been released already */
if (ldata.status != SFEX_STATUS_LOCK || strncmp((const char*)(ldata.nodename), nodename, sizeof(ldata.nodename))) {
cl_log(LOG_ERR, "lock was already released.\n");
exit(EXIT_FAILURE);
}
/* lock release */
ldata.status = SFEX_STATUS_UNLOCK;
if (write_lockdata(&cdata, &ldata, lock_index) == -1) {
/*FIXME: We are going to self-stop */
cl_log(LOG_ERR, "write_lockdata failed in release_lock\n");
exit(EXIT_FAILURE);
}
cl_log(LOG_INFO, "lock released\n");
}
static void quit_handler(int signo, siginfo_t *info, void *context)
{
cl_log(LOG_INFO, "quit_handler called. now releasing lock\n");
release_lock();
cl_log(LOG_INFO, "Shutdown sfex_daemon with EXIT_SUCCESS\n");
exit(EXIT_SUCCESS);
}
int main(int argc, char *argv[])
{
int ret;
progname = get_progname(argv[0]);
nodename = get_nodename();
cl_log_set_entity(progname);
cl_log_set_facility(HA_LOG_FACILITY);
cl_inherit_logging_environment(0);
/* read command line option */
opterr = 0;
while (1) {
int c = getopt(argc, argv, "hi:c:t:m:n:r:");
if (c == -1)
break;
switch (c) {
case 'h': /* help*/
usage(stdout);
exit(EXIT_SUCCESS);
case 'i': /* -i <index> */
{
unsigned long l = strtoul(optarg, NULL, 10);
if (l < SFEX_MIN_NUMLOCKS || l > SFEX_MAX_NUMLOCKS) {
cl_log(LOG_ERR,
"index %s is out of range or invalid. it must be integer value between %lu and %lu.\n",
optarg,
(unsigned long)SFEX_MIN_NUMLOCKS,
(unsigned long)SFEX_MAX_NUMLOCKS);
exit(4);
}
lock_index = l;
}
break;
case 'c': /* -c <collision_timeout> */
{
unsigned long l = strtoul(optarg, NULL, 10);
if (l < 1 || l > INT_MAX) {
cl_log(LOG_ERR,
"collision_timeout %s is out of range or invalid. it must be integer value between %lu and %lu.\n",
optarg,
(unsigned long)1,
(unsigned long)INT_MAX);
exit(4);
}
collision_timeout = l;
}
break;
case 'm': /* -m <monitor_interval> */
{
unsigned long l = strtoul(optarg, NULL, 10);
if (l < 1 || l > INT_MAX) {
cl_log(LOG_ERR,
"monitor_interval %s is out of range or invalid. it must be integer value between %lu and %lu.\n",
optarg,
(unsigned long)1,
(unsigned long)INT_MAX);
exit(4);
}
monitor_interval = l;
}
break;
case 't': /* -t <lock_timeout> */
{
unsigned long l = strtoul(optarg, NULL, 10);
if (l < 1 || l > INT_MAX) {
cl_log(LOG_ERR,
"lock_timeout %s is out of range or invalid. it must be integer value between %lu and %lu.\n",
optarg,
(unsigned long)1,
(unsigned long)INT_MAX);
exit(4);
}
lock_timeout = l;
}
break;
case 'n':
{
free(nodename);
if (strlen(optarg) > SFEX_MAX_NODENAME) {
cl_log(LOG_ERR, "nodename %s is too long. must be less than %d byte.\n",
optarg,
(unsigned int)SFEX_MAX_NODENAME);
exit(EXIT_FAILURE);
}
nodename = strdup(optarg);
}
break;
case 'r':
{
rsc_id = strdup(optarg);
}
break;
case '?': /* error */
usage(stderr);
exit(4);
}
}
/* check parameter except the option */
if (optind >= argc) {
cl_log(LOG_ERR, "no device specified.\n");
usage(stderr);
exit(EXIT_FAILURE);
} else if (optind + 1 < argc) {
cl_log(LOG_ERR, "too many arguments.\n");
usage(stderr);
exit(EXIT_FAILURE);
}
device = argv[optind];
prepare_lock(device);
#if !SFEX_TESTING
sysrq_fd = open("/proc/sysrq-trigger", O_WRONLY);
if (sysrq_fd == -1) {
cl_log(LOG_ERR, "failed to open /proc/sysrq-trigger due to %s\n", strerror(errno));
exit(EXIT_FAILURE);
}
#endif
ret = lock_index_check(&cdata, lock_index);
if (ret == -1)
exit(EXIT_FAILURE);
{
struct sigaction sig_act;
sigemptyset (&sig_act.sa_mask);
sig_act.sa_flags = SA_SIGINFO;
sig_act.sa_sigaction = quit_handler;
ret = sigaction(SIGTERM, &sig_act, NULL);
if (ret == -1) {
cl_log(LOG_ERR, "sigaction failed\n");
exit(EXIT_FAILURE);
}
}
cl_log(LOG_INFO, "Starting SFeX Daemon...\n");
/* acquire lock first.*/
acquire_lock();
if (daemon(0, 1) != 0) {
cl_perror("%s::%d: daemon() failed.", __FUNCTION__, __LINE__);
release_lock();
exit(EXIT_FAILURE);
}
cl_make_realtime(-1, -1, 128, 128);
cl_log(LOG_INFO, "SFeX Daemon started.\n");
while (1) {
sleep (monitor_interval);
update_lock();
}
}

File Metadata

Mime Type
text/x-diff
Expires
Wed, Feb 26, 10:23 PM (9 h, 15 m ago)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1465820
Default Alt Text
(122 KB)

Event Timeline