Re: Snapshot refresh Ora 7.1.6

On Mon, 22 Sep 1997 21:16:24 +0200, Emmanuel SELLIER <esellier_at_fisystem.fr> wrote:

>* Should i recreate a snapshot log after each alter operation on a
>replicated table in the master database ?
Yes, if you mean you alter the structure of the table(like adding or deleting a column). The snapshots cannot handle DDL changes.

>* What if one of the client database (where the above table is
>replicated) has not aplied each rows update contained in the reseted
>snapshot log ?

When you delete the snapshot log, the snapshot goes into "complete" mode until it is synced up again. So, After a DDL change on a master table you will have to: 1. drop the snapshot log 2. drop the snapshot on the client machine 3. Create the snapshot log on the master table 4. . recreate the snapshot on the client machine to reflect the new layout. 5. Once the "new" snapshot is built, you will begin only updating changed rows again. NOTE: If you create the log after recreating the snapshot on the client, you will force 2 complete refreshes.

>* Does a snapshot log reset force client updates to be in "complete"
>mode ?

The answer is yes, but doesn't sound like it applies in this case. If you are changing the table's structure, you must recreate the snapshot.
>* Should i "break" user jobs from the queue before altering master
>tables ?

Yes, or set the 'next' time to later than the change. You will probably be recreating the snapshot, so this setting becomes meaningless.

>* How does the master database clean mlog tables ?
It keeps a date of last refresh for each client snapshot. Once ALL clients(it keeps a list of all clients too) have refreshed as of a certain date/time, all records with DML changes prior are deleted.

>* Can i force the snp process to commit every N row during refresh ?

Good question. I asked the same one myself. The answer is no so far.

>* What command should i use to refresh a BIG table ?
A. "BIG" is realative. B. The same commands used on little ones. I'm not sure what you are asking here.

>* What if the one of the rollback seg. assigned to the snapshot cannot
>be extended ? Will the process continue on another ?

Yes, the job for that snapshot will just continue to fail until something is done about the problem.
>* Known bugs in snapshot system in 7.1.6 or 7.3.2 ?

Who's the master and who's the client? I had a BUCKET load of problems when the master is 7.1.6 and the client is 7.3.2.3. It's a long and horrible story. Post a reply if you are interested.

>Thanks a lot for answering directly by mail at :
>mailto:esellier_at_fisystem.fr
Received on Fri Sep 26 1997 - 00:00:00 CDT