Useful Links:

• Yale AD home

• TSM at Yale

• IBM's TSM Support Site

This procedure is taken, in a slightly modified form, from an IBM TSM Support document which discusses how to do full restores of a system from TSM. Because this is a topic which there is great interest in here at Yale, we are publishing it for the benefit of the Roundtable members and others.

IMPORTANT:

• This procedure is for skilled administrators only. It has been shown to work but has not been extensively tested.

• For Windows XP or other Windows systems, you should consult the IBM document linked above to see if this procedure may be modified to work on those systems.

• This procedure is only as useful as your most recent successfully-completed backup. Have you verified that your system is backing up properly?

• This process is ONLY for recovery onto identical hardware to the original system.

• If you want to do a "Point in time" recovery (where the system is restored to the state it was in at a given time and date), add the date and time for the restore to the command lines in steps 4, 7, 9 and 10. Remember that TSM only runs once a day and try to find out when the backup actually occurred on the target date.

1. Install the same version of the OS that was on the original machine onto the same partition that it was on before the crash (usually C:\WINDOWS).

   This is a change from previous methods of restoring Windows from TSM which involved a "parallel installation" onto a different partition or different Windows root directory.

2. Fully patch Windows. Washing it thorough Windows Update a few times is fine.

3. Install and configure the latest version of the TSM client. Verify that you can connect to the TSM server as the node name you want to restore.

Tip: When you install the TSM client, add a line which reads:

RESOURCEUTILIZATION 4

to the dsm.config file. This will dramatically improve restore performance.

4. Start the TSM client command line tool.

   IMPORTANT:

   FROM THIS POINT, DO NOT REBOOT THE SYSTEM UNTIL THIS ENTIRE PROCEDURE IS COMPLETE, EVEN IF PROMPTED TO DO SO!

5. Restore the System State's Windows File Protection (WFP) catalogs from TSM with this command:

   restore group "{SYSTEM STATE}\SYSSTATE" c:\systemp\ -showmembers -preservepath=complete -rep=all

6. From a command prompt (not TSM), copy the recovered system state files into the C:\WINDOWS directory tree:

   xcopy /h /s /y c:\systemp\SYSFILES\WINDOWS\system32\CatRoot\* c:\windows\system32\CatRoot\

7. Delete the temporary directory that TSM used for the WFP catalogs

   rmdir /s /q c:\systemp

8. From the TSM command line, restore the system drive (usually C:)

   restore c:\* -sub=yes -rep=all

9. Restore the other partitions on the system using a similar command line as in step 7.

   This can run simultaneously with the restore of the system drive if you start additional TSM command line consoles for each additional partition. This speeds up overall restore time except for very small systems where the entire system is contained on a single TSM tape.

Note: This seems to trigger Windows File Protection errors (pop-up windows) during the restore. It's probably best to "skip" replacing the file but this could use further testing.

When all partitions have been fully restored, proceed to the next step.




10. Restore the system state from TSM:

    restore systemstate

11. Restore the system services (includes IIS metabase, etc) from TSM:

    restore systemservices

12. UNPLUG THE MACHINE FROM THE NETWORK.

13. Reboot the system. This may take longer than normal.

14. Log in as the local Administrator. This may also take longer than usual.

    The system may put up a device detection/installation message and prompt you to reboot, this is OK. Just reboot and continue with the next step.

15. Check that the system seems to be running OK.

    Remember it's unplugged from the network so you may get errors related to that.

16. Reboot and plug the system back into the network during the POST. Log in as local admin.

17. Recommended step: Drop the system from the domain and re-join it. This will prevent problems with the system's password not being in sync with its computer object in the AD.

18. Check that system seems to be running normally by logging in as a YALE user, checking network services, etc.

19. END

Be sure to run tests on all services the machine provides to be sure there are no lingering glitches.