Service Restart

[Kathir Lakshmanan]

How does the service restart works?

I have a Service pointing to a script /etc/cmcluster//abc.mon and the RESTART is set to "-r 1"

Inisde the script it goes in a loop and checks for certain processes if they are not running, it breaks out of the loop stops the app and starts the application.

My question is, does the script need to return a exit code of 0 or 1, so that the RESTART count will increase or is it sufficient when the loop is broken it comes out of the script to bump the RESTART COUNT Value?

Thanks,
Kathir

[melvyn burnard]

The Srvice Restart i sthere to restart your script in case it dies.
The way a Service is used is that you set up a script to do an infinite loop to check for the existence of required processes, for example the Oracle log writer.
If one of your specific application processes goes missing, your script should exit with an exit value of say 1.
The Service Manager then sees this, and if there is a Restart value set AND the count has not dropped to 0, the Service Manager will restart that service (i.e. your script)
If there is no restart value or it has dropped to a 0, the Service Manager notifies the Package mmanager that a the service has died, and this causes the Package Manager to halt the package and attempt to switch it to an alternate node.

All documented in the Managing Serviceguard manual available from http://docs.hp.com/en/ha

[Kevin Wright]

script should not exit, unless the processes are not running. Restart count cannot be changed within your script, that is defined in the pkg config.

[Kathir Lakshmanan]

I understand that the count is specified in the pkg cntl file. The part I dont understand is - The script goes in a loop, check for availability of processes, but when it doesnt find one of the process, then it breaks out of the loop and should it restart and then return a exit 1 for the Restart count to increase? or the exit 1 is not necessary??

-Kathir

[Kathir Lakshmanan]

If anyone can post a sample mon script with proper exit options, that would be great...

-Kathir

[BPatrick]

Kathir,
A monitor script for oracle db package is attached.

Regards
Patrick

[Carsten Krege · Accepted solution]

It does not play a role how the service script exits. SG monitors the process. If it exits for any reason and the maximum restart value is not reached yet it will try to restart it, otherwise it would halt the package on the node.

It does not play a role if the service script exits with return value 0 or 1 or anything else. As soon as it exits, SG takes action. The restart counter is increased with every exit.

This is different to the package control script where indeed various exit codes have different meaning:

0 - clean exit
1 - exit with NO_RESTART
2 - exit with RESTART (e.g. dependent EMS resource not available)

Carsten

[Kathir Lakshmanan]

See the previous post