Solaris 2.6 5/98, ADSM 3.1.2.20 Server - After a night of successful backups to
the primary disk pool, during a morning Migration to tape my ADSM server began
reporting read-errors from a specific diskpool volume. My diskpool just acts as
a transcient pool for the prior night's backups, is not mirrored and is not
cached. Because migration did not complete before the errors began, some
portion of data from the night's backup was irretrievable. To determine and to
recover what was retrievable, and to reconcile the DB with the readable
diskpool, here's what I did:
1. 'Audit Volume Fix=No' - some but not all areas of the disk/volume were
damaged/unreadable;
2. 'Audit Volume Fix=Yes' - completion state=FAILURE/FAILURE (attempted
twice)
3. 'Move Data' (off the damaged volume) - completion
state=FAILURE/FAILURE/SUCCESS (attempted 3 times)
4. 'Vary Volume Offline' / 'Update Volume Access=Destroyed' (manual commands)
5. 'Restore Volume' - completion state=SUCCESS
6. 'Audit Volume Fix=Yes' - completion state=SUCCESS
I cancelled the process in Step 1 after it had listed over 1000 damaged files.
I expected Step 4 to recover the readable/recoverable files, moving them to
other volumes in diskpool, expected Step 5 to recover any that had made it
successfully to Copypool during the original Migration, and expected Step 6 to
reconcile the diskpool and DB (remove any still-unreadable/unrecoverable
diskpool entries from the DB).
I thought that Step 6 was the end of the recovery process, to the extent we
could recover.
However, since then two unexpected things happened:
1. To verify step 6, I ran 'dsm' from an affected node and checked a
known-damaged filename (one of the files listed as damaged in Step 1)
- the backup for that date was still in the versions list for that
file
2. After step 6, and again this morning (4 days later), I induced another
Migration
- both Migrations paused/failed trying to access the 'Destroyed'
volume, until I did a
'Update Volume Access=ReadOnly' / 'Vary Volume Online'
KEY QUESTIONS,
Q1. Why did the first two 'Audit Volume Fix=Yes' attempts fail ?
Q2. Why is Migration still prompting for the 'Offline'/ 'Destroyed' volume,
and how do I correct this ?
Q3. Is my DB 'clean' (reconciled) yet ?
-rsvp
Kent Monthei
CNT Data Recovery Services
SmithKline Beecham Pharmaceuticals R&D
King of Prussia, PA
[EMAIL PROTECTED]
