DMX migration : Running out of instance numbers for ctd devices

[Ra Jose]

Hello friends,

We have a prod oracle database server on rx8640 running HP-UX 11iv2. This is connected to EMC san DMX-3000. We will soon be migrating the DMX-3000 to DMX-4000. I understand that EMC engineers copy the entire DMX-3000 to DMX-4000. Our HP-UX server will be connected to this new DMX-4000. What precautions need to be taken to make sure, all the LUNs needed by the server are discovered on the new SAN DMX-4000.

Before migration, we can take maps of all VGs. After the migration, we can import these maps which will create a new lvmtab with new devices from DMX-4000. We can run ioscan and insf to discover and create device files.
Do we run out of instance numbers for these devices. Since there are a large number of LUNs and each LUN is given an instance number by the kernel (which appears in ioscan o/p under I). Do we remove old ctd devices and then recreate new ones for new devices from DMX-4000.

I believe there is a limit on how many devices can be addressed on one HBA path. 16 times 128
which is 2048 LUNs can be addressed on on HBA
patch. I am sure we would exceed this if we keep old ctd devices in /dev/dsk and new ones get created from new DMX-4000.

Are there any precautions to be taken?

Thank you for your replies.

Ra Jose.