[THIN] Re: Slightly OT: SQL Redundancy

  • From: "Bray, Donovan (ESC)" <BrayD@xxxxxxxxxxxxxxxxxx>
  • To: thin@xxxxxxxxxxxxx
  • Date: Tue, 16 Aug 2005 13:41:32 -0700

Your absolutely right, but in our case; If my memory serves me right,
upgrading to Enterprise processor licenses was going to cost us another 18
thousand dollars (not including any additional clustering hardware that
would have been required). In our SPECIFIC case we opted to have additional
hardware on hand because our SQL server while majorly problematic for part
of our business, is not a mission critical application. It essentially
effects one department (who didn't want to spend the extra $18k).  Our
Exchange server on the other hand IS a mission critical application, and I'm
not allowing management to be so miserly with it.
While my SQL solution isn't a fully HA solution, it is a much better and
survivable solution than a single server in a rack that's running SQL.
We've recently added another special purpose SQL server and we've been able
to use essentially the same hardware. So I now have two production SQL
Standard servers, and a full working spare in case of emergencies.  Now that
I have a couple production servers I can start considering buying some more
drives and replicating databases between them.  
Depending on how my experience goes with Legato I may consider using it and
retrofitting my SQL production servers to clusters as long as I can continue
using the Standard Processor Licenses. I'll do more research on it once I
have some practical experience with the product from our Exchange


From: Evan Mann [mailto:emann@xxxxxxxxxxxxxxxxxxxxx] 
Sent: Tuesday, August 16, 2005 11:21 AM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Re: Slightly OT: SQL Redundancy

If one of your failover situations requires restoring from backup, then
that's not a very high level of redundancy.  I wouldn't consider it to be
high availability, and only "half" of failover.

Drives are going to be the most common thing to fail in a heavy use server.
The processor, power supplies, and mainboard doesn't get the type of
mechanical abuse HDD's do.


From: thin-bounce@xxxxxxxxxxxxx [mailto:thin-bounce@xxxxxxxxxxxxx] On Behalf
Of Bray, Donovan (ESC)
Sent: Tuesday, August 16, 2005 2:18 PM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Re: Slightly OT: SQL Redundancy

Put "EMC/Legato's FullTime AutoStart/SE (formerly Co-StandbyServer
AAdvanced)" on your list to check out.
I have it on the way to replace our MSCS Exchange Cluster.
The advantage of Autostart is that we can use block level mirroring between
nodes and eleminate the single point of failure of the shared scsi resource.
We opted not to use clustering for our SQL boxes due to the steep processor
licensing as you have to graduate to  Enterprise processor licenses.
What we opted to do is buy two hardware boxes that are totally identical,
Each loaded with 2gb RAM (Max for Standard), we licensed the ACTIVE box for
its Dual Processors. We use the spare box just as a test/development box.
Both boxes are configured with two 146gb ultra3 hot-swap scsi drives in a
hardware mirror.  24gb system partition, the rest is for SQL on a secondary
partition. If there is a failure on the primary server, I have dedicated
hardware already in the rack, powered on, ready to recieve the drives  from
the failed box.  If the failed box's drives are whats suspect, then I have
to restore from backup.  
Your business realities may be different and not allow you to have this kind
of manual failover. But the alternative in our case was just too expensive
to justify.


From: Joe Shonk [mailto:joe.shonk@xxxxxxxxx] 
Sent: Tuesday, August 16, 2005 8:55 AM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Slightly OT: SQL Redundancy



What a success (or failure) has the group encountered in making SQL servers
Highly Available?  Either through clustering (Microsoft or third-party) or
replicated partners.  We are looking to move the Citrix DataStore and a few
application databases to a HA solution, while minimizing the impact to farm
in event an HA partner goes down.



Other related posts: