Backup using ARCserve

[Don Spare]

I have a potential issue with ARCserve backups of my Oracle archived redo logs on HP-UX. It seems when there is high redo rollover, and the archived redo log backup is running at the same time, then Oracle gets a message from the archiver process that archiving failed. It then tries again. This continues until the ARCserve backup is finished and then it returns to normal.

My question is what is actually happening here?
Is is possible the ARCserve client has 'stolen' the file lock on the redo archive that is currently being written?
Can ARCserve lock the whole directory?
Is the archive log in question complete?

This is making me a bit nervous as I cannot find any information on ARCserve that goes into this level of detail.

If anyone can shed some light on this for me I would greatly appreciate it. Or if someone can point me to a reliable source for getting these answers.

[harry d brown jr]

Don,

If you have a backup running while you have files opened and being written to, then the backup will keep attempting (most will) to back the files up - or until it just quits.

Why are you backing up your archive logs if you are already backing them up via some other method??

Rule #1: Unix doesn't have locks.
Rule #2: Applications that use a locking scheme must rely on other applications to use the same method to properly work.
Rule #3: When backing up a database using a non-database backup, make sure the database is in some kind of "hotbackup" mode.
Rule #4: Make sure you are current on your patch bundles.
Rule #5: Make sure you have good backups - test them!

live free or die
harry

[Pete Randall · Accepted solution]

Rule #6:
Don't trust products from CA

[Don Spare]

Well, I'm sorta in a Catch-22 here. My redo logs are being archived as quickly as possible during the nightly batch cycle. So fast that they can quickly overflow the allotted disk space reserved for them. Therefore I try to back them up to tape in the middle of all that so the older ones can be deleted and free up space for future files. If I wait for the batch to finish, the archive logs will certainly overflow the allotted 16GB of storage space. That will freeze the database until more space is availabe for the archiver to run. Its a precarious balancing act.

for Rule #3 - this is an archived redo log backup only.

for Rule #4 - I wish.

for Rule #5 - I'm afraid any test I can run will be when my production instance dies and needs recovery.

for Rule #6 - no comment.