SFHACFS vs Hitachi-VSP & Host mode option 22

[Matej_Susta]

Hi!

I'm installing several new SFHACFS clusters and during the failover testing I run into annoying problem - when I fence one node of the cluster, DMP logs high number of path down/path up events which in the end causes the disk to disconnect even on other active nodes.

We've found out that our disks were exported without the host mode option 22, so we fixed this on storage. Even after this the clusters bahaved the same. Later I've read somewhere on internet, that it's good idea to relabel the disks, so I've requested new disks from storage and did vxevac to new disks. This fixed just two clusters we have, but the other two are behaving still the same.

Have anybody experienced anything similar? Do you know anything I can test / check on the servers to determine the difference?

The only difference between the enviroment is that the not-working clusters have the disks mirrored from two storage systems, while the working ones have the data disks only from one storage.

[Gaurav_S · Accepted solution]

Hello,

This may need some investigation though .. would like to understand some basics ..

1. what is OS & SF version ?

2. Is array configured as per recommended settings in HCL ?

have you had a look at HCL below: (taking SF 5.1 in consideration)

http://www.symantec.com/docs/TECH74012

There is a firmware level mentioned for VSP array, does your array have that firmware level ?

3. Read out  http://www.symantec.com/docs/TECH47728  

Above article talks about Host and Storage Configuration Guide for DMP. Have a look at page 23 for VSP array settings which veritas recommends ...

for e.g for solaris

Solaris
50-07-66-00 (USP/NSC)  >>>>>>>>>>>>>> min. firmware level
09 (Solaris) >>>>>>>>>>>>>>>>>>>>>>> Host mode
02 (Veritas Storage Foundation for Oracle Rac), 22 (Veritas Cluster Server)  >>>>>> host mode options
A/A >>>>>>>>>>>>>>>>>>>>>>>>>>>>> DMP Failover mode ..

G

Gaurav