Friday, May 19, 2006

Flash Recovery Area on ASM?

In yesterday's post, "VLDB with ASM?", I showed problems with ASM mirroring for high volume database. Today, I want to emphasize that you should be very careful placing flash recovery area (FRA) on ASM diskgroup with normal or high redundancy.

The reason is the same as before. Whenever all or majority of disks in a failure group are "broken", ASM starts immediately rebalancing. If there is no enough space left in failed failure group (well, zero is all disks fail) than ASM will mark those disks as HUNG and no files can be created on that particular disk group. For flash recovery area this is very bad state - no files can be create on this diskgroup, i.e. ARCx processes couldn't backup online redo logs so at some point the DB will just freeze if situation is not resolved promptly. Flash back logs also couldn't be written. Any database backups will fail. Resolution requires manual intervention and might actually take a while.

If FRA needs to be placed on ASM than consider using external redundancy even for relatively small size FRAs. Use mirroring inside storage box or between storage boxes as some vendors support that.

In fact, the latter looks like a good approach for mirroring of large ASM database over different storage boxes - using external redundancy and let SAN boxes to mirror using vendor's technology (though, it might require additional licensing).

0 Comments:

Post a Comment

<< Home