topic Re: Service guard queries in Operating System - HP-UX

Service guard queries

masthan_1 — Fri, 04 Mar 2011 18:13:17 GMT

Hi, I am having a bit of conceptual knowledge of MC-Serviceguard and having a few queries. Please help me understanding the following.
=> what are the changes can be done on running cluster and running Package?
=> if one package fails what will happen to running cluster?
=> what are the uses of cluster configuration file & binary file (are both same)?
=> Roles of cluster lock disk?
=> when the cluster id corrupted or gone, what will happen? How can we regenerate it?

Regards,
KS

Re: Service guard queries

Steven E. Protter — Fri, 04 Mar 2011 18:19:39 GMT

Shalom,

=> what are the changes can be done on running cluster and running Package?

Yes, so long as you don't make mistakes cmquerycl/cmcheckconf/cmapplyconf can happen on a running cluster. You should be able to validate the package so you know it works prior to these steps.

=> if one package fails what will happen to running cluster?

Other packages should continue to run assuming they are not dependent on the one that failed.

=> what are the uses of cluster configuration file & binary file (are both same)?

Cluster runs off the binary. It builds the binary in the process outlined above from the text files.

=> Roles of cluster lock disk?

Decides if there is a problem which node gets to keep running and which node gets rebooted(TOC) to prevent data corruption.

=> when the cluster id corrupted or gone, what will happen? How can we regenerate it?

Need more details to answer this question.

SEP

Re: Service guard queries

masthan_1 — Fri, 04 Mar 2011 19:18:37 GMT

Thanks steven for the answers.
I would like to have the package configuration procedurs (with commands if possible). Is there anything like we only exports package control files not package configuration files to other nodes in the cluster.
=> If a package fails, how can we recover the same and please post me the step by step probing procedures. In which case, a down package will affect other packages in the cluster?
=> How many cluster lock disks we can have in a cluster?

Thanks in advance for your answers.

Re: Service guard queries

P Arumugavel — Sun, 06 Mar 2011 06:23:23 GMT

hi,

Package Configuration Procedure:

Create the Package Configuration File:
1. #mkdir /etc/cmcluster/package_name (Create directory for package specific files)
2. #cd /etc/cmcluster/package_name (Change to the directory)
3. #cmmakepkg -v -p pkg.conf (Create package configuration template)
4. #vi pkg.conf (Modify package parameters)

Create the Package Control Script:
1. #cmmakepkg -v -s pkg.cntl (Create package control script template)
2. #vi pkg.cntl (Modify script variables)

Copy Package Control Script and Other Files to Other Nodes:
ftp pkg.cntl to other systems in the cluster (Distribute the control script to all nodes)

Create and Distribute the Binary File:
1. cmcheckconf -v -P /etc/cmcluster/package_name/pkg.conf
(Verify the package configuration file)
2. cmapplyconf -v -P /etc/cmcluster/package_name/pkg.conf
(Compile package into cluster binary file)

Rgds...

Re: Service guard queries

P Arumugavel — Thu, 25 Aug 2011 20:34:30 GMT

hi,

>> => How many cluster lock disks we can have in a cluster?

Read the below thread.

http://h30499.www3.hp.com/t5/Serviceguard/Multiple-PVs-in-cluster-lock-VG/m-p/2934358#M5371

Rgds...

Re: Service guard queries

P Arumugavel — Sun, 06 Mar 2011 06:49:05 GMT

>> => If a package fails, how can we recover the same and please post me the step by step probing procedures. In which case, a down package will affect other packages in the cluster?

hi,

If a package fails, you will get some input from the screen.

check for any errors spitted under /etc/cmcluster/pkg.cntl.log and /var/adm/syslog/syslog.log.

This both will give you the maximum detailed information about the errors.

See the error while halt/start the package.

Verify the health of filesytems associated with package.

>> In which case, a down package will affect other packages in the cluster?

SEP replied already: Other packages should continue to run assuming they are not dependent on the one that failed.

Rgds...