linux + ServiceGuard + Samba toolkit problem

[Tal Moran]

Hi all,
I install Serviceguard on RHES 3.0. The Serviceguard work Ok. I try to configure the SAMBA toolkit and I can't start the SAMBA package. The SAMBA server runs on both cluster nods OK. I use the Readme file in the toolkit to configure the Samba cluster. When I start the Samba package I get in the log file error: cmrunserv: Unable to complete command: File exists.
In the attached file there is the configuration file: samba_pkg.cntl, samba_pkg.conf, hasmb.mon and the log files: samba_pkg.cntl.log, /var/log/messages.

Thanks Tal.

[Dan Am]

Hi,just as an aside: My version of SG is only certified for RHAS 2.1. Could that be a problem ? Regards Dan

[Tal Moran]

Hi
I am using Serviceguard for Linux A.11.15 and Samba Toolkit version A.01.00-0. According to the web site I didnâ t see any problem to install on RHES 3.0.

http://www.software.hp.com/portal/swdepot/displayInstallInfo.do?productNumber=T2341BA

T

[Dan Am]

Ok , I have 11.14

[Celso Medina Kern]

Hi Tal,

The core of the error seems to be
Jun 14 11:01:28 nod-1 cmcld: Unable to start service samba_pkg (it is already active)

Looking package log, we see that service is starting twice, and fails in the second chance, for obvious reason.

Could you attach /usr/local/cmcluster/conf/samba_pkg/toolkit.sh and the complete /usr/local/cmcluster/conf/samba_pkg/samba_pkg.cntl and /usr/local/cmcluster/conf/samba_pkg/hasmb.mon?

Celso

[Tal Moran]

Hi Celso,

Attached is the entire configuration file for the samba package.

Thanks Tal.

[Celso Medina Kern]

Tal,

My findings:
1 - According to instructions on toolkit header, you should call this script with start or stop only. You are using monitor instead.

2 - The script slice where the prob occurs is:
# Check if cmrunserv should be called the old way without a restart count.
if [ "$SMB_SERVICE_RESTART" = "" ]
then
cmrunserv $SMB_SERVICE_NAME ">> $0.log 2>&1 $service_cmd"
else
cmrunserv $SMB_SERVICE_RESTART $SMB_SERVICE_NAME ">> $0.log 2>&1 $service_cmd"
fi

>> more specific in else statement. It would translate:
cmrunserv -r 0 samba_pkg ">> samba_pkg.cntl.log 2>&1 /usr/local/cmcluster/conf/samba_pkg/hasmb.mon"

This would call
/usr/local/cmcluster/conf/samba_pkg/toolkit.sh monitor
I dont know how cmrunserv runs, but it seems there is a problem once you´re starting monitor again after is has already been started in the normal start of package(hasmb.sh, function launch_monitor_service)

It seems you should not use toolkit.sh to monitor the service.

My suggestions:

- dont use toolkit.sh to monitor the service, use hasmb.mon instead.
- Turn shell verbose, including set -x at the beggining of hasmb.sh. It will tell us much more about what´s been executed.

Good luck!!

C

[Eric Taverny_1]

You are right Celso.

I had the same problem for the linux sendmail service guard toolkit.

I had defined the hasendmail.monitor SERVICE in the the package control script (for the package control script to start it). But the toolkit.sh script launch the hasendmail.sh script that starts the hasendmail.monitor SERVICE.

So when the cluster try to start it (as defined in the package control script), it fails because the service is already started by the toolkit.sh (hasendmail.sh) script.

I have an question : is better to start the SERVICE with the toolkit.sh (hasendmail.sh) or with the package control file ?

Thanks ;-)

Cheers.