Re: rollback after crash

Pradeep wrote:

> redo log1 gets full -> log switch -> redo log 2 gets full -> log switch
> -> redo log 1 is over written here.

This will not happen because you will get a "cannot write redo log" message in your alert log and everything will come to a grinding halt.

Jonathan Lewis wrote:

>All Oracle processes are the same on-disc code,
>so any process could run any bit of the code -
>in theory - so it's perfectly feasible for smon and
>pmon both to have the ability to perform a rollback
>of a transaction that has no parent.

Ahh, that explains things like:

$ oerr ora 281
00281, 00000, "media recovery may not be performed using dispatcher" // *Cause: An attempt was made to use a dispatcher process for media // recovery. Memory requirements disallow this recovery method.
// *Action: Connect to the instance via a dedicated server process to // perform media recovery.

jg

--
@home.com is bogus.
Let's go to the space!  http://www.ritsumei.ac.jp/~akitaoka/warp-e.html