Time Machine ejecting internal hard drive during backup

Question

I originally posted this in the Time Machine group. It was suggested I post to the Xserve group.
We have an XSERVE running 10.6.4 server. It has 2 internal 1TB drives. The plan is to use one as the main drive and the second as a Time Machine backup. This used to work in 10.5, but since about 10.6.2 or so, it has not worked correctly.
Time machine will come on, begin the backup, and then error off with a message that the drive has been ejected. I have erased the drive, but still no joy. I have partitioned the drive, still no joy. I have verified the drive repeatedly, and all is fine. I can copy files to and from the drive until the cows come home, and all works fine, which means I do not think it is the drive itself. Activate TM, however, and it will fail before the first backup is completed. The only way to get the drive back is to do a complete power down and then power up. Needless to say, that is not a good thing for a server that we rely on.
Any ideas?
10/14/10 10:41:27 AM com.apple.backupd1230 Backup content size: 542.8 GB excluded items size: 474.1 GB for volume Server HD
10/14/10 10:41:27 AM com.apple.backupd1230 No pre-backup thinning needed: 82.46 GB requested (including padding), 930.13 GB available
10/14/10 10:41:27 AM com.apple.backupd1230 Waiting for index to be ready (101)
10/14/10 10:41:27 AM mds70 (Normal) DiskStore: Creating index for /Volumes/backup/Backups.backupdb
10/14/10 10:47:38 AM kernel FusionMPT: Notification = 22 (SAS Discovery) for SCSI Domain = 0
10/14/10 10:47:38 AM kernel Discovery condition = 0x000f0001
10/14/10 10:47:38 AM kernel FusionMPT: Notification = 18 (SAS Phy Link Status) for SCSI Domain = 0
10/14/10 10:47:38 AM kernel SAS Phy Link Status: PhyNum = 0, old link rate = 8, new link rate = 0, SASAddress = 50080007002e8c04
10/14/10 10:48:10 AM kernel SCSIPressurePathManager: Timed out waiting for inactive/error path to become active, loops = 0
10/14/10 10:48:10 AM kernel 
10/14/10 10:48:10 AM kernel disk0s2: media is not present.
10/14/10 10:48:10 AM com.apple.backupd1230 Stopping backupd because the backup volume was ejected!

10/14/10 10:48:10 AM com.apple.backupd1230 Error writing to backup log. NSFileHandleOperationException:* -NSConcreteFileHandle writeData:: Input/output error
10/14/10 10:48:10 AM com.apple.backupd1230 Error writing to backup log. NSFileHandleOperationException:* -NSConcreteFileHandle writeData:: Input/output error
10/14/10 10:48:10 AM com.apple.backupd1230 Error writing to backup log. NSFileHandleOperationException:* -NSConcreteFileHandle writeData:: Input/output error
10/14/10 10:48:10 AM com.apple.backupd1230 Copied 13416 files (4.0 GB) from volume Server HD.
10/14/10 10:48:10 AM com.apple.backupd1230 Error writing to backup log. NSFileHandleOperationException:* -NSConcreteFileHandle writeData:: Input/output error
10/14/10 10:48:10 AM servermgrd64 servermgr_backup: TimeMachinePostBackupHook called.
10/14/10 10:48:10 AM servermgrd64 servermgr_backup: TimeMachinePostBackupHook done.
10/14/10 10:48:11 AM kernel jnl: disk0s2: close: journal 0xffffff8012b15d20, is invalid. aborting outstanding transactions 


Answer 1

No, it is not always at the same point/file/bytes transferred. It will vary from a few GB (say 3GB) to quite a few GB (say 30GB). I would add amongst the things I have done was to:
Perform numerous Disk Utility Checks. Never any errors found.
Erased the backup drive on multiple occasions. Double checked the GUID and format.
Partitioned the drive and tried backing up to a new partition. (No joy there either)
Copied lots and lots of data from the root disk to said partitions to test them. All went fine.
It really is Time Machine that is causing the issue. I can use the drive all day long as long as TM does not run. When it does, I eventually get a failure message that the drive has been ejected.
While I have not tried making a partition on the home drive (I'd rather not, lots of data there, make users mad, and since TM is not working, I have no backups at present), I do limit the data that is being backed up to JUST that of the system. /Users is being excluded at the moment. So I doubt there is anything out of the ordinary in the system. And, yes, I did a disk check on the root disk and all is fine there too.