Re: Help! Oracle server clustering

Saikat Chakraborty wrote:
>
> I beg to differ in a point of your post.
> If you resize a data file in primary, the standby is not broken.
> I am talking of 8.1.5.
> I believe resizing of datafile also generates some redo information
> that gets passed to standby database. Hence the data file size
> of standby also increases automatically.

Assuming, of course, that you have room on the standby filesystem for that file to grow, which isn't necessarily the case.

> Adding a datafile is a problem. It stops the recovery process from
> happening. You need to *cold* copy the datafile from primary
> to standby and add it to your standby database by the command
> ALTER DATABASE CREATE DATAFILE '<datafile name>' AS '<datafile name>';
> After that, recovery can proceed as normal.

No - you do one or the other, but not both. Either you 1) copy over a valid backup (either cold OR hot) of the affected datafile and add it to the standby site OR 2) you create the datafile on the standby site using "alter database create datafile". Either way you should refresh the standby's controlfile from production.

> The problem in standby is as you pointed out the unrecoverable
> activities. If any so-called poweruser performs that,
> your standby database is broken.

Broken is a bit strong - I'd characterize the standby as having diverged from the primary temporarily. But the whole standby isn't broken, just the datafiles which contained the particular nologging transactions need to be refreshed from production. And, in 9iR2, you can prevent nologging transactions at a global level for just this reason.

Regards,
Sean M Received on Fri Jun 28 2002 - 00:42:23 CDT