Why would shutdown abort cause any uncertainty in data integrity? It
seems no less certain than shutdown immediate, in that uncommitted
transactions are aborted, and committed transactions are safely in the
online redologs. In the case of shutdown immediate, you have to sit
around waiting for uncommitted transactions to roll back (and whatever
else to happen). In the case of shutdown abort, uncommitted
transactions are rolled back after the database is back up, in the
background. It seems to me abort would be better for a cluster failover.
--
Jeremiah
www.wolfe.net/~jeremiah
Joe wrote:
>
> May be my description was not clear enough.
>
> In my case, I am running Legato cluster software with 2 Solaris and 1
> shared disk.
> When the cluster is up or switch-over from one machine to
> another machine, it will first to shutdown the Oracle and then
> bring up the Oracle database.
>
> In some case, the Oracle is shutdown (normally) already before
> the switch-over occurs. This means that there is NO Oracle Instance
> or processes running when issuing the shutdown command. But the command,
> as mentioned before, is hanged. It is NOT waiting for the incomplete
> transaction rollback, but something else (something I don't know
> or understand).
>
> Of course, I can always use shutdown abort but I do not like to issue
> shutdown abort because it may cause some uncertainty in the data
> integrity when there is database running.
Received on Thu Jan 06 2000 - 14:58:44 CST
