solaris cluster 3.2 application failover

creating a simple application-failover for SC3.2

Normally you would want to create a script that is

called by the start_method to initiate the application,

and a script that is called by the monitor_method

to check whether your application is still running.

 

In sun cluster 3.2 however, it is sufficient to just

create a start script. Monitoring is done by 'pmfadm'

if the application fails, pmfadm calls an action that

will fail-over the application to another node.

 

If you do want to monitor the application yourself, you

will have to change the start-method and change the

following line:

 

pmfadm -c $PMF_TAG -a "$action -R $RESOURCE_NAME -G $RESOURCEGROUP_NAME  -T $RESOURCETYPE_NAME" $start_cmd_args

to this:

 pmfadm -c $PMF_TAG -n 2 $start_cmd_args

 

This results in that the application will be restarted (-n 2) twice.

No action will be called. Your monitor script will detect that

the application has failed and a Give_Over will be initiated.

 

In this document pmfadm is left untouched, so there is no need

for a monitor script.

----------------------------------------------------------------------------

 

Step 1

Should be performed on all nodes that should be able to run

the application!

The application needs a configuration directory for its

scripts and other files.

mkdir /opt/SUNWklok

The application will get a startupscript and a monitorscript

vi /opt/SUNWklok/startklok

(startklok)

/usr/openwin/bin/xclock -display <aws-ip>:0

The script should be made executable

chmod +x /opt/SUNWklok/*

 

Step 2

Run the wizard "scdsbuilder" from /usr/cluster/bin

specify the vendor and the name of the resource-type

e.g. vendorname=SUNW resourcetype_name=klok

specify the directory to build the environment

for your ha-application

e.g: /opt/XCLOCK

click on "create"

then go on to the next screen.

specify the application-binary or script.

e.g. /opt/SUNWklok/startklok

This will create the resource-type package.

Make sure that you get the environment created in

/opt/XCLOCK/SUNWklok on the other nodes as

well. So for example go to /opt/CLOCK, tar

the tree from SUNWklok onwards and ftp to the other

nodes where you untar them again in /opt/XCLOCK.

 

Run pkgadd to install the resource type

pkgadd -d /opt/XCLOCK/SUNWklok/pkg

NOTE! pkgadd on all nodes!

On both nodes, change the following file:

/opt/SUNWklok/bin/klok_svc_stop.ksh

Find the following lines:

((SMOOTH_TIMEOUT=$STOP_TIMEOUT * 80/100))

((HARD=$STOP_TIMEOUT * 15/100))

and change them into:

((SMOOTH_TIMEOUT=5))

((HARD_TIMEOUT=5))

step 3.

register the resource-type

clrt register SUNW.klok

create the resource-group

clrg create klok-rg

create the resource

clrs create -g klok-rg -t SUNW.klok klok-rs

To be able to switch between nodes within the hour after a

failover, change the pingpong interval

clrg set -p Pingpong_interval=10 klok-rg

On the display server disable access control

xhost +

Manage and Online the resource group.

clrg manage klok-rg

clrg online klok-rg

 

Switch the klok resource group to the second node and back

 

clrg switch -n secondnode klok-rg

clrg switch -n firstnode klok-rg

 

Kill the clock on the first node:

 

pkill xclock

 

The clock will be restarted on the same node.

Kill it again. It will be restarted on the same node

for a second time. If you kill it a third time, it

will fail_over.