Cloning two K class servers

[Brian Butscher]

I have two identical K-460 servers that are connected through two Brocade 2800 switches to a Clariion CX600, along with a varitey of NT and Sun servers. Both K-460's use A3404 FCAL adapters (bummer). One server has been working fine with the Clariion for about two months, using EMC Power Path 3.0.3. I took a make-tape-recovery backup from the first K-460 and restored it to the second K-460, changing some file system sizes, IP addressing, and hostname. The restore was flawless. When I booted up the second K-460 the performance on the first K-460 was terrible at best. I was getting emcp:MP:Error: Path to cXtXdX is dead, FCP:Reset triggered due to missed Login event, Missed event:0x3 errors on both servers. On the Clariion, the initiator information for both servers is:system type=HP no auto-trepass, failover=1, arraycomm=enabled. I did remove the file /etc/log/HostIdFile.txt on the second K-460 so when Navisphere was stopped and restarted this file would be recreated with the proper IP address on the second K-460. The zoning on the switches is simple, port 3, 10, 12; port 1, 10, 12. Port 3 and 1 are the two K-460's and ports 10 and 12 are the Clariion. To add a little humor to this, in all seriousness, the two servers are abbott and costello (to those of you who know of this old comedy act - these two are playing it to the hilt)! This looks like an addressing type of problem. Any ideas where there might be a hostname, VGID, FCAL-WWN, that might be the same on both of these servers?

Thanks for any suggestions or ideas you might have...

[Scot Bean]

Sounds like the clone worked a little too well. Now the settings for your configuration are too similar.

Can you simply reconfigure the restored machine's applications? Not being familiar with your exact setup, hard to give advice from this end.

Perhaps you can compare the contents of /etc/rc.config.d/ on both machines, looking for configuration items that are the same but should not be.

Also, paths to storage may want to redundant, instead of the same.

[Brian Butscher]

The solution to this problem was to configure one server on one Brocade switch and the second server to the other Brocade switch. Even though EMC stated they supported FCAL with multiple servers on the same switch, I beleive there is a conflict with the internal SCSI id of the servers on the same quick loop. The ultimate solution to my problem was to replace the FCAL with fibre port adapters. With this new configuration I was able to cross connect each server across both switches.

Thanks,

Brian