serviceguard monitoring toolkit can not restart service when killing PID?

[Kurt_33]

I am configurating a seviceguard with apache/tomcat, oracle...etc. And I am using the standard cluster toolkit to start/stop/monitor the application.

My customer wants to local restart/retry the application before failover. For instance, if the httpd was mannully killed, it should restart the httpd again.

However, when I manully kill the httpd PID, the PID went away and in pkg.cntl.log, it showed the it retries to keep search for PID before failover, but NOT restart the httpd. So after 3 times retry to search the PID, it failover to other node

Is there any standard kit that I can make use of instead of writing my own script to restart the application when I kill the PID?

[Olivier Masse]

I'm not familiar with the Apache toolkit but what is doing this is probably a Serviceguard monitor. I don't use the supplied monitors with HP's toolkit because, simply put, I don't like how they react to tribial problems. The behaviour of some of them is to initiate a failover if they detect an error.

Thesem monitors are usually standard scripts so you can change yours to do a retry before initiating a failover. Check for a file named xxxx.mon in your package directory. These monitors are normally enabled with the keyword SERVICE_NAME in the .conf file.file.

[Kurt_33]

SERVICE_NAME[0]="hkuesd_http"
SERVICE_CMD[0]="/etc/cmcluster/hkuesd_http/toolkit.sh monitor"
SERVICE_RESTART[0]="-r 3"

The retry is only for retry to see if PID still not, but not restart the application.

So if I manually kill the application PID, it will retry to found the PID. And surely it will not able to find it if there is no restart on the application.

Just want to check if there is standard script to restart application in mcsg package