EVA Vdisk "Failover/Failback" Setting

[Jack Trachtman]

Each EVA5K Vdisk can have failover/failback
setting of either:

No preference
Path A-Failover only
Path A-Failover/failback
Path B-Failover only
Path B-Failover/failback

What is the recommended setting
for VMS hosts? Tru64 hosts? And why?

Thanks

[Uwe Zessin]

The supported setting depends on VCS revision (=EVA controller firmware), operating system type and firmware. If you give me the numbers, I can look it up for you.

A 'failover' happens when a controller goes down. 'failback' means that the EVA transfers virtual disk ownership back if the failed controller come up again. Some operating systems can get confused by this or unexpected answers from SCSI inquiries.

You can always force a manual 'failover' of a virtual disk from the host ("SET DEVICE/SWITCH" or something like this on OpenVMS, sending a 'START UNIT' with SCU on Tru64 Unix).

[Jack Trachtman]

OK, I spoke w/HP and now have a clearer understanding of this Failover/Failback option.

In effect, this is only useful if you are using what I'll call an "EVA" centric approach to defining paths to the hosts. We are not. We are allowing VMS to do a rough load-balancing by assigning paths in a round-robing manner at boot. I was hoping the above would automatically put the failed-over paths back the way they were after a controller errors out, but that's not possible.

What HP suggested (that some sites do) is create a script to check on the paths and Switch them to the desired paths if they do not match what's desired. This means that I have to do my own load-balancing by deciding on the paths, and update my table as disks are created and destroyed. Doable but not the simplified solution I was looking for.