Service Guard package spawns defunct processes

[Jim Krol]

On a 3-node cluster, running a package which starts several web services, proxy services, and samba services, I find that after a short time, some of the services spawn lots of defunct processes. I can successfully manually stop and start these individual services outside the control script, but, defunct processes get spawned against a few of the services which are started in the control script.

We are running iPlanet-WebServer-Enterprise/4.1 and Netscape-Administrator/3.5, as well as HP's version of SAMBA.

Following is an excerpt of the customer defined functions:

function customer_defined_run_cmds
{
/opt/samba/bin/stopsmb
wait
/netscape/server4/https-admserv/start
/netscape/server4/proxy/start-admin
wait
/netscape/server4/https-service1/start
/netscape/server4/https-service2/start
/netscape/server4/proxy/proxy1/start
/netscape/server4/proxy/proxy2/start
wait
/etc/cmcluster/web/applicationscript start
wait
/opt/samba/bin/startsmb
test_return 51
}

The customer halt function is just a reverse of the above.

I should be able to do the above? Right? Or do I need to take all those actions in the function and roll them up into 1 script, and ONLY execute 1 script in the function?

Has anyone else seen package spawn defunct processes?

[Bill Hassell]

The Service Guard package is just a script which then spawns other processes and scripts which then...you get the picture.

A defunct process is one which was improperly detached from it's parent, probably due to kill -9 (NEVER use kill -9, especially on systems that must be reliable). You'll have to discover which script or process is terminating the parent processes improperly. If a program is starting the defunct processes then the calling sequence and wait for completion is incorrectly written.