MC/Service guard

[Prashant Zanwar]

Hi,

I work in production environment and presently is allocated to a production system testing project.

I am going to test infrastructure against the customer requirements and some known standards.

I am going to test HP systems. Systems are L-Class. I want to test systems throughly. Systems have MC/Service guard installed.
SG is only used for purpose VG transfer from one system to another and not for HA.

Please forward your suggestions about how best to test the systems? What all I should cover in configuration test.

Best regards
Prashant

[Tim D Fulford]

As I understand your question you are not using ServiceGuard to help mitagate hardware/software failure merely to transfere the relavent volume groups & the data within it.

As such I would say this is not the best use of SG, but that said...

1 - As a "control case" Make sure VG transfere is successful & timely
2 - Test for single path failure, e.g. loosing a SCSI card or fibre card/cable fibre switch AND check transfere is sucessful
3 - test for disk faiulures & that data is available
4 - Test backup & restore proceedures
5 - test for failure of lock disk
6 - test for LAN heartbeat failre (& see what happens to the package)

If you were using the full functionality of SG you could then extend the above to network failures etc. H/W failures etc.

However, just becasue you are NOT using this functionality be aware theat SG DOES require redundant LAN cards & the failure of the LAN (well heartbeat) will result in a failover. Thus DO NOT try and specify machines with say 1 LAN card, especially in the live environment.

Regards

Tim

[Kent Ostby]

Prashant --

The following test WILL cause the system that it is run on to reboot itself, but you should be able to test the failover of the volume group to the other machine if SG is properly configured.

Locate the pid for the cmcld daemon (ps -ef | grep cmcld).

Kill this pid:

kill

This will cause SG to TOC the box that you issue the command on and the package to switch over.

[Prashant Zanwar]

I am attching output of cmviewcl...which will be helpful to know the information about cluster.

Can someone suggest, how best to check for HP-UX and Cluster configuration. Are there any scripts available for it...

Please reply immediate.

Regards
Prashant

[Kent Ostby]

The command "cmscancl" available on most but not all versions of ServiceGuard out there will create a file called /tmp/scancl.out which is quite useful for looking at the configuration.

One thing to check in this file is for the node timeout value to be set to between 8 to 12 seconds and for the heartbeat interval to be set to 1 second.

---------------------------------------

I will also look at the cmviewcl that you posted and see if I spot anything there that needs checking.

[Kent Ostby]

The following are a couple of things to note from the cmviewcl. They are NOT errors, but just an explanation of things are currently set up on your system.

Both of the packages are currently up and running on cmrun50 and will failover to different nodes if there is a problem.

If the cluster is halted (cmhaltcl) and then restarted (cmruncl), the packages will not restart based on the values in the cmviewcl.

Also, if the packages failover to cmrun10 and cmrun30 and cmrun50 becomes available again, the packages will not fail over AUTOMATICALLY to cmrun50.

All of those are fairly normal settings for a ServiceGuard cluster.