I ran into a curious problem with edev today. I think its a bug, but I
thought I'd run it past the list and see what you all think.
We have using edev a lot to define SAN volumes for Linux guests. This past
week, we did a big power shutdown of our data center, so all the VM systems
got recycled. I found, after one of them came up, that I had an edev
device pointing to the wrong LUN. I traced it back to a typo in my system
config file. The virtual device was listed in 2 different edev statements
in my config file:
edev FF0B type fba attr 2107,
fcp_dev 100B wwpn 50050763060886AB lun 4012400000000000 ,
fcp_dev 110B wwpn 50050763061886AB lun 4012400000000000
edev FF0B type fba attr 2107,
fcp_dev 100C wwpn 500507630618C6AB lun 4012400C00000000 ,
fcp_dev 110C wwpn 500507630608C6AB lun 4012400C00000000
When the system came up, I don't recall seeing any error messages, though I
might have missed it. What I find strange is that the FF0B device ended up
pointing to the 2nd edev definition (lun 4012400C) rather than the first.
(The 2nd statement should have been FF0C.)
It seems to me that, in this situation, CP should have rejected the 2nd
definition and posted an error message rather than just overriding the
previous definition with the new one.
What do you think?
Martha
