IFINCREMENTAL backup of VM guest fails with ANS9351E

[EpsiloN]

Hello

I have TSM 7.1.4 with TSM for VE VMware. Backups of VM-s goes to "fake tape" pool. So vmware control files and data reside on the same storage pool.

After some time backup failed with ANS9351E and ANS1314E errors. I found this article - http://www-01.ibm.com/support/docview.wss?uid=swg21961789 that says that problem will be solved with

"When the VMware control file is not available on the server, run a FULL VM backup to recreate the control file. In this example, instead of using "-mode=IFINCREMENTAL", run a backup with "-mode=IFFULL". The IFFULL backup will recreate the control file so that future IFINCREMENTAL backups can succeed."

It really works, but here is my question. I know that VMware control files should be on disk-based storage and data could be on tape and "fake tape." Could it possible that control files were deleted/moved after "fake tape" expiration/reclamation?

And I need to follow this recommendation, right?
"The vmctlmc option is required if virtual machine data files are backed up to tape. Virtual machine control files must be backed up to a disk-based storage pool that does not migrate to tape. The storage pool can be composed of random access volumes and sequential file volumes; the storage pool can also be a deduplicated pool. Use the vmctlmc option to specify a management class that stores data in such a storage pool."

So, from TSM point of view it is tape.
If I am right, is it possible to extract/migrate only control file to another storage pool?

Thank you.

Regards
Igor

 

[moon-buddy]

You have to store the control files to another tape and not just a 'fake tape'.

I have defined a separate tape pool for the control files so the restore will be faster.

 

[RecoveryOne]

Before putting VE into a Directory Container pool, I was much in the same situation. VE was going to a disk pool, that would migrate to tape. I found backups to be extremely slow, and restores to be unbearable after we started to use tape storage. Each time we started a backup or restore for a VM, it would have to read the control files from tape(s). My solution at the time was to create a small disk pool about 600gb in size and use the vmctlmc option. After doing that, vm backup/restores were much quicker, but still limited by my physical tape mounts. Now, you say "Fake Tape" and not sure what that is here in this context. As long as the vm is active, I never seen ALL the control files get expired. Could it be possible that you may have volumes unavailable that contains the control files? Perhaps not enough mount points to service the request of the vm backup plus reading of the control file? If "Fake Tape" is VTL, sadly I have no experience how volumes in VTL behave.

So, from TSM point of view it is tape.
If I am right, is it possible to extract/migrate only control file to another storage pool?

When we first made the switch to using the vmctlmc, we actually found it MUCH easier to do a new full vm backup of all our nodes. As it stands today in our VE environment with about 400 VM's there's 1529717 control files.

Hope it helps!