Hardware Failure Of Disk Drive

Hello,
While researching potential solutions to expand the disk capacity of our
server I have come across a question that I hope someone in here is able to
answer.
For example a database has got 4 file groups, each with critical data on all
of them. These are directly connected to the server either by internal RAID
config or SAN device. While the database is in operation a transaction is
opened across all file groups, updating and inserting data to all 4. During
this process a RAID card or connection to the SAN fails cutting the
connection to file groups 3 and 4.
The transaction being execured then hits an error (as it cannot continue to
update and read 3 and 4) and goes to rollback transaction. The rollback
however is successful on the file groups 1 and 2 but not on 3 and 4. Thus
potentially leaving data inconsistency.
Am I correct in assuming this? If so, is there any way to combat this
possible situation. Though, I do understand that the likelyhood of a RAID
card or SAN failing is remote the chance is still there.
Thanks
Anubis.If a write fails under these conditions then the database will certainly be
marked Suspect at which point a restore from backup is going to be your best
option. Regular transaction log backups are the simplest way to mitigate
this. Log shipping is an additional option.
On the hardware side, it's possible to mirror over redundant controllers for
RAID or use multiple controllers with failover in a SAN, which will prevent
or limit data loss and downtime in the event of a disk or controller
failure.
--
David Portas
SQL Server MVP
--|||My understanding is that if it's not committed when the DB restarts it
should rollback the remaining file group 3 and 4 when their connectivity is
back.|||Hello,
This was my understanding also, though, if a database looses connectivity to
a file group would it stop accepting or working with the other files and go
'offline' or does it continue to work with the other files?
"Peter Yao" <peteryao@.NoSPAMhotmail.com> wrote in message
news:%23AR8ACHcFHA.3280@.TK2MSFTNGP09.phx.gbl...
> My understanding is that if it's not committed when the DB restarts it
> should rollback the remaining file group 3 and 4 when their connectivity
> is
> back.
>