May 5, 2011 at 7:03 pm
Hi Guys,
Just got a question about clustering solutions for sql servers.
Whenever I refer to any clustering article I saw that disadvantage is the single point of disk failure becuase clustered servers shares a single disk.
I wanted to know are there any solutions for this? Can you please points to me any articles or whitepapers or blogs?
Thanks for your help guys.
May 5, 2011 at 9:04 pm
Using appropriate RAID levels will always minimize the chance of data loss, to a great extent. You can serach for RAID levels in google and you will get many articles.
Nag
------------------------------------------------
How to get the best help on a forum
http://www.sqlservercentral.com/articles/Best+Practices/61537
May 5, 2011 at 10:32 pm
from the way you have worded your question we need to clarify the real issue.
The single point of failure is really the SAN, which may include disk, power or connectivity. All of these can be covered with good design.
As mentioned by Nagesh, the disk can be covered by proper RAID configuration, typically use RAID 10 (prefered) or RAID 5 (if you must). If possible have a hot spare inthe SAN so if a disk goes it starts rebuilding immediatly.
You can prevent power failure by having a well configured UPS, with 2 power sources for the SAN.
For the conectivity have dual HBAs with Load ballancing.
This pretty much covers everything that can go wrong with the SAN short of somone driving a truck over your data centre, or as I have seen a technician physically powering the SAN off.
Cheers
Leo
Leo
Nothing in life is ever so complicated that with a little work it can't be made more complicated.
May 6, 2011 at 9:10 am
There are multiple disk use to make a single partition or drive with the help of RAID levels,Cluster is use for High Availability if 1 node fail then second node will work
Its depend on Raid Level ,if 1 disk failure then other will work in the same Raid Level,If you have policy for the Disastor of Storage System then you have to plan DRP (Disastor Recovery Plan) ,DRP onsite or offsite its depend on your requirement
Regards,
Syed Jahanzaib Bin Hassan
BSCS | MCTS | MCITP | OCA | OCP | OCE | SCJP | IBMCDBA
My Blog
www.aureus-salah.com
May 9, 2011 at 6:06 am
In all sql servers the "disks" are a single point of failure, regardless of clustering. Clustering is hardware failover for the server ( e.g. your box with cpu, memory and local disk ) To have a cluster you have to connect your nodes to the same storage: An alternative is to use full two way replication that way you have two of everything including disks.
That said all solutions are to protect against various types of failure and you have to choose the balance of cost, availablity and complexity which suits your requirement. Just as servers have multiple power suplies to guard against failure so your storage will use raid and redundant bus and power supplies, you will use multiple network cards and multiple paths to your storage and you'll raid your actual disks. You have to do all of these, raiding disks without multiple paths to the storage still gives you a single point of failure. Having the lot with an alternative power source is a single point of failure too, a ups which maintains the server but not the network switches is still a point of failure - in other words don't get hung up on disks as a "point of failure". The reason you pay so much for a SAN is never for performance, it's the availability features and management.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
May 9, 2011 at 6:07 am
oops! sorry should have said "without" alternative source of power.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
May 10, 2011 at 4:00 pm
you can convert clustering from an HA solution to a true DR set up by:
setting it up as a geo-cluster (SAN replication and windows 2008 required)
log shipping to a third server (or another cluster)
database mirroring to a third server (or another cluster)
replicating to a third server (or another cluster) - ok that one is a joke. 🙂
---------------------------------------------------------------------
May 10, 2011 at 9:15 pm
That answers my question. Thanks for your help guys.
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply