Re: Cluster and WriteBack Caching

Thanks to both of you. The reason I asked the question was that I was a bit
confused by reading the BOL section about Troubleshooting SQL Clusters, which
states:

To optimize performance when using failover clustering, consider the
following:

If your disk controller is not external to your clustered computer, you must
turn off write-caching within the controller to prevent data loss during a
failover.


Write-back caching cannot be used on host controllers in a cluster without
hindering performance. However, if you use external controllers, you continue
to provide performance benefits. External disk arrays are not affected by
failover clustering and can sync the cache correctly, even across a SCSI bus.


My setup is two Dell 6650's with 2 HBA's each. These are fibered into a
Dell/EMC CX300 disc array with SPS. All mdf's and log's are on the SAN disc
array. As there is no switch, I am direct-attached. My thinking is that I
should stick with the write-back, but wanted to get some expert input.

Thanks.
--
John


"Mike Epprecht (SQL MVP)" wrote:

> Hi
>
> Yes, I agree with you, Internal SAN caching is OK, but since the poster did
> not specify, I assume HBA caching was being referred to here.
> What really scares me is that some HBA vendors allow for caching on their
> products.
>
>
> Regards
> --------------------------------
> Mike Epprecht, Microsoft SQL Server MVP
> Zurich, Switzerland
>
> IM: mike@xxxxxxxxxxxx
>
> MVP Program: http://www.microsoft.com/mvp
>
> Blog: http://www.msmvps.com/epprecht/
>
> "Geoff N. Hiten" <sqlcraftsman@xxxxxxxxx> wrote in message
> news:uhNfTgrnFHA.2152@xxxxxxxxxxxxxxxxxxxxxxx
> >I am gonna disagree here... Sort of. Never use Write caching on local disk
> >controllers in a cluster. However, Internal SAN write caching is a good
> >thing that allows significant performance improvements. This is assuming
> >that the SAN vendor has a very strong data protection (batteries,
> >de-staging, mirrored cache, etc.) design.
> >
> > --
> > Geoff N. Hiten
> > Senior Database Administrator
> > Microsoft SQL Server MVP
> >
> > "Mike Epprecht (SQL MVP)" <mike@xxxxxxxxxxxx> wrote in message
> > news:equSLKrnFHA.576@xxxxxxxxxxxxxxxxxxxxxxx
> >> Hi
> >>
> >> Yes. Write caching can not guarantee that the data is on disk if the
> >> server dies. In that case, your "high availability" solution could turn
> >> into a "no-availability" solution as you will probably have database
> >> corruption after the failover.
> >>
> >> Read caching, that is OK, but never write caching.
> >>
> >> Regards
> >> -------------------------------
> >> Mike Epprecht, Microsoft SQL Server MVP
> >> Zurich, Switzerland
> >>
> >> IM: mike@xxxxxxxxxxxx
> >>
> >> MVP Program: http://www.microsoft.com/mvp
> >>
> >> Blog: http://www.msmvps.com/epprecht/
> >>
> >> "JT" <Jthayer@xxxxxxxxxxxxx> wrote in message
> >> news:95F802D0-13AE-4A7B-B17B-59FE8A1FDDB6@xxxxxxxxxxxxxxxx
> >>> Hi,
> >>> I have a 2 instance sql failover cluster on two nodes. Currently, write
> >>> and
> >>> read back caching are enabled on all disks in the SAN. Could this lead
> >>> to
> >>> problems? Thanks.
> >>> --
> >>> John
> >>
> >>
> >
> >
>
>
>
.

Relevant Pages

  • Re: Clustering with NAS
    ... Whilst I appreciate that clustering with NAS is ... >> Mike Epprecht, Microsoft SQL Server MVP ...
    (microsoft.public.sqlserver.clustering)
  • Re: Do I need Enterprise to do clustering?
    ... MVP - Windows Server - Clustering ... But we could upgrade SQL2000 Enterprise to SQL2005 Enterprise in place ... Microsoft SQL Server MVP ...
    (microsoft.public.sqlserver.clustering)
  • Re: configure sql cluster with one NIC (two nic but teamed)
    ... Microsoft SQL Server MVP ... > MVP - Windows Server - Clustering ...
    (microsoft.public.sqlserver.clustering)
  • Re: Cluster and WriteBack Caching
    ... Yes, I agree with you, Internal SAN caching is OK, but since the poster did ... Mike Epprecht, Microsoft SQL Server MVP ...
    (microsoft.public.sqlserver.setup)
  • Re: SQL and Majority Node Set
    ... Double Take's product rides on MS clustering ... > Don Wilwol ... >> Mike Epprecht, Microsoft SQL Server MVP ...
    (microsoft.public.sqlserver.clustering)