Restore Log Header

[techer]

Hi all,

I am attempting to restore my Information Store (Ex5.5 Sp4) from backup. When I go to start the Info Store, the transaction logs are automatically rolled into the database successfully until the process gets to the LAST log file. Then, I get these errors:

ESE97; ID:124; Desc: (1960) Unable to read the log header. Error -530.

ESE97; ID:0; Desc: (1960) -530

MSExchangeIS; ID:1081; Desc: Unable to recover the database because error 0xc8000212 occurred after a restore operation.

MSExchangeIS; ID:5000; Desc: Unable to initialize the Microsoft Exchange Information Store service. Error 0xc8000212.

I've read as much as I can find about this and found that, while ANY info on how to actually recover from it is non-existent, it is "..the single most common reason that the restoration of an online backup of Exchange Server may not work." (MS KB 247109)

Anyway I was wondering whether anybody's seen this? Is there a utility I can run that will remove the need to roll data in from the logs (I don't mind losing minor amounts of data).

At this stage it appears that my full online backup (which was registered by Arcserve 2000 as being complete and error free) is 56GB of useless data...

Please help if you can.

Thanks,

MW

 

[zbnet]

You haven't given us any history of the event or explained the reason why you need to restore your database, so it's very difficult to offer the corrcet advice for your situation.

Is this a production restore? Test restore? Mailbox recovery restore? Is this your first experience of restoring an Exchange 5.5 database? Have you read the MS DR whitepaper that explains all about how the logs etc works? faq10-6004

 

[techer]

Hi,

We needed to restore the database because of disk IO errors that were popping up in the Application Log. Running CHKDSK revealed bad clusters in Priv.edb and Pub.edb (but zero bad sectors, further tests ruled out possible HW failure, which left the file system). I decided to wipe the drives and reformat.

While we still had a functioning IS, we were lucky enough to be able to migrate all the mailboxes to a temporary server on the same site.

The Public Folders' contents however, which contained most of the corruption, gradually diminished until the IS crashed & wouldn't start at all.

Since then, we've been endeavouring to get the IS database restored from a backup. Our plan has been to restore the IS (both Priv & Pub), then rename Priv.edb (to get a clean priv dbase), then migrate all the mailboxes back.

The problem we encountered is in my original post. This morning I attempted the restore again, this time using the Full backup from the Friday, then restoring the latest Diff backup from the following Thursday over the top.

Whilst reading a KB article, I came across an interesting comment tucked away inside, mentioning (briefly) that if the server was shut down in such a way to leave the database inconsistent, the Edb.log file should not be present, because it "may cause problems during the restoration" (no mention made as to why). The comment goes on to state that if the server was shut down gracefully, include the Edb.log file in the directory prior to rolling forward.

I renamed the Edb.log file to Edb.bak, ran the restore and it worked correctly first time! I've since found that the differential job from that Thursday actually already had the corrupt clusters present, so I'm going back to the original restore (which I couldn't get working before) to hopefully get a bad-cluster-free database directory.

I am quite confident that upon restoring, when I have renamed the Edb.log file, the restore will work as it's supposed to. I will follow up with results after tonight (hopefully, problem solved)...

Cheers,

MW