SRDF - And bringing up a server for the other side

[John Ford_1]

I have a DR site we are brining up. And I have a K580 on each side. My question is that the luns that the EMC is giving to the machine on the other side is not the same as the luns that I have on the local side. My original plan was to do a make_recovery tape and then apply that at the DR side. How do I get it to recognize the new disks on the other side appropriately?? To add another twist we have Sybase running in a raw partitioned environment striped accross disks in VG02.

[Alan Riggs]

You will want your SCSI paths and LUNs to be identical. The best way to do this is to have identical hardware layouts (card expansions) on each machine. If you don't have this (or if cards were added later and appear in different order) you can manipulate the SCSI Ids with the ioinit process. Read teh man page on ioinit and make sure you are clear on the process. Generally, you will have to remap twice. The first remapping shifts everything to "high numbers" to clear the IDs you need. The second remapping sets the permanent values. Remapping does require a reboot.

If you read the man pages and need teh procedure outlined in detail, post again and I (or someone quicker, most likely) will help you out. But I strongly recommend understanding the procedure well before following someone else's step-by-step.

[Dennis Raudenbush · Accepted solution]

John,
Last year I did alot of Failover/DR with EMC BCVs and SRDF. For a group of servers that act as Failover/DR we also have tried/reworked to arrange I/O cards in the same slots (vclass servers) and same with patch levels. The beauty of LVM vgexport -s -p in creating map files is you just have to copy them to your DR server, then vgimport those shared map files and you don't have to worry about what controller number/target/disk (LUN) it maps to. vgimport will find the disk ID on the SRDF disks for you (as these are track for track identical to your production disks). I'm assuming the SRDF vg(s)/lvol(s) are application/database ones here and that you've done an EMC SRDF split of the production group/hypers first. Also you have to pre-create the DR vgs group files on the DR box prior to vgimporting them.

If you need details I can email you portions of our DR scripts. Hope this helps. My first response so be gentle.

john creighton

[John Ford_1]

Rich, I like the answer you proposed. Can you email me the scripts that you spoke of.

fordj@nrucfc.org...

Thanks