Testing HACMP

[Mag0007]

I have a managed system with 2 LPARs. Both of these LPARs have HACMP. How can I test HACMP with this? How can I simulate a hardware failure?

TIA

 

[mrn]

Use smit hacmp & perform a takeover on the standby node.

Mike

"A foolproof method for sculpting an elephant: first, get a huge block of marble, then you chip away everything that doesn't look like an elephant."

http://www.airport-parking-site.co.uk/

 

[DSMARWAY]

Hi

Testing your cluster , you should allocate an outage and test for all possible failures

i.e. Network card failure

1. unplug primary cable
2. unplug standby cable
3. unplug both primary and standby
4. unplug heart beat cable
5. unplug disk heartbeat cable


Controlled failover

1. Perform simple failover
2. Perform take over , then fail it back


Server crash

1. Echo something to /dev/kmem ( server will crash and failover )
2. Unplug the power cable from the server ( server should failover )
( Make sure you have valid backups before performing server crash)

Most of the above are physical tests , i.e. you unplug cables from your server

 

[Mag0007]

DSmaraway:

Thanks for the reply.

How would I test a cable failure on a LPAR? Everything is virtual, how would I simulate a failed cable?

 

[DSMARWAY]

You have 2 LPARS HACMPed , i.e. in effect 2 servers sharing the resources , so LPAR 1 will have , the same resources allocated to it as lpar 2

so 1 lpar may have 2 x cpus ,4 GB RAM , 3 network interfaces
and lpar 2 the same

( you need to check what resources are allocated to each LPAR )

 

[Mag0007]

Correct, but how would you simulate a 'failed hearbeat'? There is not physical Rs232 or SCSI cable to do this....

Is there a LPAR command from the HMC I can issue to break this heartbeat?

 

[DSMARWAY]

what version of HACMP are you running ?
because you needed a non ip heartbeat ( well you did on versions 5.1 and earlier) i.e. either a physical serial cable between nodes or via disk

Your heartbeat also uses your network interfaces

what does cllsif show ?