yþythoss
No, I replace both disks, not remove both disks.
OK I think this is an English language issue.
You mean that you "swap" the disks from port 1 to port 2 & from port 2 to port 1, effectively changing their boot order.
I read the bug report.
So the problem is that after the "rebuilt array" disks have been swapped you cannot boot.
IIRC this is a long standing issue (from many years ago) that occurs when the disk partition(s) do not end on the same cylinder.
You say you have that cylinder misnatch problem after rebuilding your array
Somewhere on the wiki (& you can search way back in forums & bugzilla) are instructions to correct the cylinder mismatch. Once corrected either disk in the array should be able to boot the server.
Years ago I think this is an issue related to the particular disks, brand, number of cylinders, size etc, rather than CentOS, although that might no longer be the case.
Can you add my comments to the bug report as I don't have physical access to buzilla at present.