Florian Haas wrote:
How can I get an up to date version of your RA?
Here:
http://hg.linux-ha.org/agents/raw-file/tip/heartbeat/vmware
I made some additional modification and created a patch file against
this script. I moved the "set_env" function inside the "validate-all"
function.
--
Cristian Mammoli
APRA SISTEMI srl
Via Brodolini,6 Jesi (AN)
tel dir. 0731 719822
Web www.apra.it
e-mail c.mamm...@apra.it
--- vmware 2010-03-01 10:58:56.138352097 +0100
+++ VMwareVM 2010-03-01 11:30:59.302347806 +0100
@@ -1,17 +1,17 @@
#!/bin/sh
#
-# VMware OCF resource agent
+# VMware OCF resource agent
#
-# Copyright (c) 2008 Apra Sistemi s.r.l.
+# Copyright (c) 2010 Apra Sistemi s.r.l.
# All Rights Reserved.
#
-# Description: Manages a VMware server 2.0 as a High-Availability
-# resource
+# 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) 2008 Apra Sistemi s.r.l.
+# Copyright: (C) 2010 Apra Sistemi s.r.l.
#
# See usage() function below for more details...
#
@@ -20,13 +20,17 @@
# * OCF_RESKEY_VIMSHBIN (mandatory, full path to th vmware-vim-cmd executable)
#
# Requirements/caveats:
-# * vmware-server 2.0 RC1 installed and autostarted on all nodes
+# * 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
-# dealing with many virtual machines: raise it to something around 300 secs
+# * Default_Action_Timeout stock value (20 sec) isn't enough if you are
+# dealing with many virtual machines: raise it to something around 900 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
#################################################################
@@ -38,12 +42,12 @@
#################################################################
# Path to the virtual machine configuration file
-VMXPATH="$OCF_RESKEY_vmxpath"
+VMXPATH="$OCF_RESKEY_vmxpath"
# Path to the vmware-vim-cmd executable
VIMSHBIN="$OCF_RESKEY_vimshbin"
-# global variables
+# Global variables
VMXDIR=
RELVMXPATH=
VMID=
@@ -69,32 +73,32 @@
vmware_get_msgid() {
$VIMSHBIN vmsvc/message $1 2>/dev/null \
| awk '/^Virtual machine message/ {print $4}' \
- | awk -F : '{print $1}'
+ | awk -F : '{print $1}'
}
# Answers message
vmware_answer_msg() {
- $VIMSHBIN vmsvc/message $1 $2 $3 >/dev/null 2>&1
+ $VIMSHBIN vmsvc/message $1 $2 $3 &> /dev/null
}
# Register a virtual machine
vmware_register_vm() {
- $VIMSHBIN solo/registervm '"'$1'"' >/dev/null 2>&1
+ $VIMSHBIN solo/registervm '"'$1'"' &> /dev/null
}
# Unregister a virtual machine
vmware_unregister_vm() {
- $VIMSHBIN vmsvc/unregister $1 >/dev/null 2>&1
+ $VIMSHBIN vmsvc/unregister $1 &> /dev/null
}
# Start a virtual machine
vmware_poweron_vm() {
- $VIMSHBIN vmsvc/power.on $1 >/dev/null 2>&1
+ $VIMSHBIN vmsvc/power.on $1 &> /dev/null
}
# Suspend a virtual machine
vmware_suspend_vm() {
- $VIMSHBIN vmsvc/power.suspend $1 >/dev/null 2>&1
+ $VIMSHBIN vmsvc/power.suspend $1 &> /dev/null
}
# Get virtual machine power state
@@ -106,27 +110,27 @@
# Get vid of missing virtual machines
vmware_get_broken() {
$VIMSHBIN vmsvc/getallvm 2>&1 \
- | awk -F "'" '/^Skipping/ {print $2}'
+ | awk -F \' '/^Skipping/ {print $2}'
}
-# Variables depending on the above functions
+# Variables depending on the above functions
#################################################################
-set_environment() {
- # Directory containing the virtual machine
- VMXDIR="`dirname "$VMXPATH"`"
+vmware_set_env() {
+ # Directory containing the virtual machine
+ VMXDIR="`dirname "$VMXPATH"`"
- # Basename of the configuration file
- RELVMXPATH="`basename "$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"`
+ # 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"`"
+ # Virtual machine name
+ VM="`awk -F '"' '/^displayName/ {print $2}' "$VMXPATH"`"
- # msg.autoAnswer value in config file
- VMAUTOMSG="`awk -F '"' '/^msg.autoAnswer/ {print $2}' "$VMXPATH"`"
+ # msg.autoAnswer value in config file
+ VMAUTOMSG="`awk -F '"' '/^msg.autoAnswer/ {print toupper($2)}' "$VMXPATH"`"
}
# Main functions
@@ -134,41 +138,44 @@
# Print usage summary
vmware_usage() {
- cat <<END
+ 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 dependencies
# Check for mandatory files presence and consistency
vmware_validate() {
if [ -z "`pidof vmware-hostd`" ]; then
- ocf_log warn "vmware-hostd is not running"
- return $OCF_ERR_GENERIC
+ ocf_log err "vmware-hostd is not running: aborting"
+ exit $OCF_ERR_GENERIC
fi
if [ ! -x "$VIMSHBIN" ]; then
- ocf_log warn "vmware-vimsh executable missing or not in path ($VIMSHBIN)"
- return $OCF_ERR_INSTALLED
- fi
+ ocf_log err "vmware-vim-cmd executable missing or not in path ($VIMSHBIN): aborting"
+ exit $OCF_ERR_ARGS
+ fi
if [ ! -f "$VMXPATH" ]; then
- ocf_log warn "Specified vmx file ($VMXPATH) does not exist"
- return $OCF_ERR_ARGS
+ ocf_log err "Specified vmx file ($VMXPATH) does not exist: aborting"
+ exit $OCF_ERR_ARGS
fi
-
+
+ # Now we can safely setup variables...
+ vmware_set_env
+
+ # ... and verify them
if [ -z "$VM" ]; then
- ocf_log warn "Could not find out virtual machine name"
- return $OCF_ERR_ARGS
+ ocf_log err "Could not find out virtual machine name: aborting"
+ 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
+ # $VMID is allowed to be empty in case we are validating a
# virtual machine which is not registered
return $OCF_SUCCESS
@@ -189,28 +196,28 @@
find "$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
+ 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"
+ vmware_register_vm "$VMXPATH"
VMID=`vmware_get_vid "$RELVMXPATH"`
if [ -z "$VMID" ]; then
ocf_log err "Could not register virtual machine $VM: aborting."
exit $OCF_ERR_GENERIC
fi
ocf_log info "Virtual machine $VM registered with ID $VMID"
- fi
+ fi
ocf_log info "Powering on virtual machine $VM"
- vmware_poweron_vm $VMID
+ vmware_poweron_vm $VMID
# Give the VM some time to initialize
sleep 10
-
- if [ "$VMAUTOMSG" != "TRUE" ]; then
+
+ if [ "$VMAUTOMSG" != "TRUE" ]; then
# msg.autoAnswer is not set: we try to deal with the
- # most common question: msg.uuid.altered
+ # 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`
@@ -218,14 +225,14 @@
fi
fi
- # Check if the VM is running. We don't bother
+ # Check if the VM is running. We don't bother
# with timeouts: we rely on the CRM for that.
while :; do
- vmware_monitor && break
+ 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 $OCF_SUCCESS
fi
@@ -237,14 +244,14 @@
if [ -z "$VMID" ]; then
ocf_log info "Virtual machine $VM is not registered"
return $OCF_SUCCESS
- else
+ 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
+ # 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
+ vmware_suspend_vm $VMID
sleep 5
while vmware_monitor; do
ocf_log info "Virtual machine $VM is still running: delaying 10 seconds"
@@ -253,12 +260,20 @@
else
ocf_log info "Virtual machine $VM is already stopped"
fi
+ # VMware randomly fails to unregister VMs,
+ # so we send the command twice
ocf_log info "Unregistering virtual machine $VM"
- vmware_unregister_vm $VMID
+ vmware_unregister_vm $VMID
VMID=`vmware_get_vid "$RELVMXPATH"`
if [ -n "$VMID" ]; then
- ocf_log err "Could not unregister virtual machine $VM: aborting."
- exit $OCF_ERR_GENERIC
+ ocf_log warn "Could not unregister virtual machine $VM: retrying."
+ sleep 10
+ vmware_unregister_vm $VMID
+ VMID=`vmware_get_vid "$RELVMXPATH"`
+ if [ -n "$VMID" ]; then
+ ocf_log err "Could not unregister virtual machine $VM: aborting."
+ exit $OCF_ERR_GENERIC
+ fi
fi
ocf_log info "Virtual machine $VM is stopped"
return $OCF_SUCCESS
@@ -278,11 +293,11 @@
# Print metadata informations
meta_data() {
- cat <<END
+ cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
-<resource-agent name="vmware">
-<version>0.1</version>
+<resource-agent name="VMwareVM">
+<version>0.2</version>
<longdesc lang="en">
OCF compliant script to control vmware server 2.0 virtual machines.
</longdesc>
@@ -301,69 +316,60 @@
<longdesc lang="en">
vmware-vim-cmd executable path
</longdesc>
-<shortdesc lang="en">vmware-vimsh path</shortdesc>
+<shortdesc lang="en">vmware-vim-cmd path</shortdesc>
<content type="string" default="/usr/bin/vmware-vim-cmd"/>
</parameter>
</parameters>
<actions>
-<action name="start" timeout="300" />
-<action name="stop" timeout="300" />
-<action name="monitor" timeout="30" interval="300" depth="0" />
+<action name="start" timeout="900" />
+<action name="stop" timeout="900" />
+<action name="monitor" timeout="30" interval="300" depth="0" start-delay="60" />
<action name="meta-data" timeout="5" />
</actions>
</resource-agent>
END
}
-# See how we were called
+# See how we were called
#################################################################
-case "$1" in
-meta-data)
+case $1 in
+meta-data)
meta_data
exit $OCF_SUCCESS
;;
-usage|help)
- vmware_usage
- exit $OCF_SUCCESS
- ;;
-esac
-
-set_environment
-vmware_validate
-rc=$?
-if [ $rc -ne 0 ]; then
- case "$1" in
- stop) ocf_log info "validate failed, vmware considered stopped"
- exit $OCF_SUCCESS;;
- monitor) exit $OCF_NOT_RUNNING;;
- status) exit $LSB_STATUS_STOPPED;;
- *) # it's safe to bail out now
- ocf_log err "vmware_validate failed"
- exit $rc
- ;;
- esac
-elif [ "$1" = "validate-all" ]; then
- exit $OCF_SUCCESS
-fi
-case "$1" in
start)
+ vmware_validate
vmware_start
;;
stop)
+ vmware_validate
vmware_stop
;;
-status|monitor)
+status|monitor)
+ vmware_validate
vmware_monitor
;;
+usage|help)
+ vmware_usage
+ exit $OCF_SUCCESS
+ ;;
+
+validate-all)
+ vmware_validate
+ ;;
+
*)
vmware_usage
exit $OCF_ERR_UNIMPLEMENTED
;;
esac
+
+exit $?
+
_______________________________________________
Pacemaker mailing list
Pacemaker@oss.clusterlabs.org
http://oss.clusterlabs.org/mailman/listinfo/pacemaker
