Clustering Help Needed

[James Payne]

I need help with the clustering service in Windows 2000. We have two
identical Micron servers and want to use clustering. We don't have any
cluster aware software, just data that needs to fail over in the event of a
server failure. My questions are this, how difficult is it to setup
clustering, and are my expectations of have my data available instantly on
the second node in the event the first fails real?



We have a database that has to stay running at all times. The data must be
real-time at all times. Will clustering keep my data identical on both
servers at all times, so that when we fail-over, our data is current as of
that moment?



Please reply to (e-mail address removed)

 

[k2list]

I need help with the clustering service in Windows 2000. We have two
identical Micron servers and want to use clustering. We don't have any
cluster aware software, just data that needs to fail over in the event of a
server failure. My questions are this, how difficult is it to setup
clustering, and are my expectations of have my data available instantly on
the second node in the event the first fails real?

Clustering is not extremely difficult to setup, but it is not for the
faint of heart. It can be tricky, depending on the configuration.

The two computers would be sharing a common disk system, so the data is
really just written in one place. In case of failure, the second node
takes over the data disks and presents the same data.

We have a database that has to stay running at all times. The data must be
real-time at all times. Will clustering keep my data identical on both
servers at all times, so that when we fail-over, our data is current as of
that moment?

Databases are another story. I would not recommend running non cluster
aware DB software on a cluster. If Node A fails while in the middle of
a transaction, this may lead to data corruption. After fail over, Node
B is going to present you with the same corrupt data.

Running non cluster software on a cluster is possible. Configuration
and reliability of data in case of failure depends on the software itself.

 

[NuT CrAcKeR]

clustering isnt that difficult to setup. It can be very EXPENSIVE though...

First, you need a specific kind of storage system, a SHARED storage system.
These can be SCSI, but it is much easier to configure and is more reliable
when using Fibre Channel.

Next, your database server MUST be cluster aware. SQL 2000 Enterprise will
fit this order.

The reason you must use a cluster-aware database product is that the cluster
has to be able to control the application, and its services.

The other thing you can do is have 2 database servers running in real time,
and have them do realtime data replication between them. This is not ideal,
but then you can have a session-based load balancing device in front of the
DB servers. When a client goes to acecess the DB, the load balancer will
pick one of the two DB servers for the connection, and make sure that the
client connection is maintained only to that server. This can be very ugly,
and is often problematic.

When using a proper cluster solution with a clustered DB, failover takes
about 30 seconds to 2 mintues depending on the hardware you decide to use,
and the size of the databases that you are working with.

If you need more information about this topic, let me know.

NuTs