Re: Alternative solutions to NLB?

"ggcluster" <> wrote in message
news:...
>
> Hi everyone,
>
> We are currently using Microsoft NLB for the clustering of our sql
> databases, which is not ideal.

How does that even work? You can't have two copies with the data identical
if both copies are writable.

> From what I understand, NLB only works is
> one of the servers in the clusters go down, this doesnt help us our
> problem at the moment is that for some reason, certain services on one
> of the two clustered servers goes down which NLB does not detect,
> therefore NLB thinks that the server is still fine and working when its
> not.

NLB does not do health checking of services. The server has to completely
fail for a node to stop receiving packets.


> What other clustering solutions are there? Whats currently the best on
> the market? I want to recommend something to my manager but I want to
> study and evaluate certain solutions first.

The appropriate solution is to use MSCS (2003) or Failover Clustering (2008)
for SQL.

Russ is absolutely right (as usual). NLB is intended for Web services, like
IIS and MSCS (2003 and earlier) or Failover Clustering (2008 and later) are
intended for making other types of applications (like SQL and Exchange)
highly available.

--
Ryan Sokolowski
MVP - Clustering
MCT, MCITP x3, MCTS x8, MCSE x2, CCNA, CCDA, BCFP


"Russ Kaufmann" <> wrote in message
news:A5359685-DD7C-4153-B632-...
> "ggcluster" <> wrote in message
> news:...
>>
>> Hi everyone,
>>
>> We are currently using Microsoft NLB for the clustering of our sql
>> databases, which is not ideal.
>
> How does that even work? You can't have two copies with the data identical
> if both copies are writable.
>
>> From what I understand, NLB only works is
>> one of the servers in the clusters go down, this doesnt help us our
>> problem at the moment is that for some reason, certain services on one
>> of the two clustered servers goes down which NLB does not detect,
>> therefore NLB thinks that the server is still fine and working when its
>> not.
>
> NLB does not do health checking of services. The server has to completely
> fail for a node to stop receiving packets.
>
>
>> What other clustering solutions are there? Whats currently the best on
>> the market? I want to recommend something to my manager but I want to
>> study and evaluate certain solutions first.
>
> The appropriate solution is to use MSCS (2003) or Failover Clustering
> (2008) for SQL.
>
>