Re: Can sombody list disadvantages of using Optimal setting for Rollback segs.

rgaffuri_at_cox.net (Ryan Gaffuri) wrote in message news:<1efdad5b.0310201058.1a0147d7_at_posting.google.com>... BIG SNIP as he basically started a new thread ...
> im having trouble sizing my optimal setting for some staging
> instances. Right now they are in default which is 1m. We do about
> 40-50 loads a night, many over lap each other. Cant give you a precise
> rate because they run when the data files come in which varies.
>
> Some loads are very small and run in seconds some loads deal with 10+
> GBs of data and have very large rollback segments. We cant really size
> to specific rollback segment because the size of the individual loads
> varies(we dont know which file will be large on a particular night).
>
> We have tons of disk space and its just a staging database so we dont
> backup rollback. We do our backup off of the transportable tablespaces
> we use to publish data. So if we lose a rollback we just drop it and
> drop the datafiles, then import the tablespaces.
>
> what rules of thumb can i use to size these? im concerned about wait
> issues with growing the large segments. Im thinking of sizing them to
> 50-100m for optimal since maybe 4-5 loads a night grow larger than
> that and on many nights none of the rollback segments get that large.
> Its just the 'occasional' case.
>
> Not a particularly easy thing to set up on the fly. Sorry for the
> vagueness of my response. Im on 8.1.7.4 in Sun Solaris.

Ryan,

(ignoring comments about backups, ... must ... stay ... on ... topic) As you proclaim to have tons of disk space, how about disabling optimal, and see how big they get?

(ARRRRRRGH ... can't resist ...)
if you have loads of disk space, why not backup the rollback tablespaces datafiles also? I mean, its more work to *exclude* tablespaces' datafiles from a hot or cold backup script that it data dictionary driven than it is to include them.

(this subject was just covered by HJR in detail in c.d.o.s).

Pd Received on Mon Oct 20 2003 - 18:39:25 CDT