What causes a SQL Cluster to Failover?

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

tjhazel - 28 Jun 2005 16:34 GMT

What causes SQL Server to fail from the active node other than the heartbeat
flat-lining? I have the cluster installed and tested the heartbeat by
unplugging the active machine. Failover works great.

My question is what else can cause the failover? Lets say we have a memory
leak. Is there a point when the heartbeat is still fine, but the memory is
slowly being consumed. Is there a threshold that causes a failover?

What about an application that is hung and consuming excessive cpu cycles.
Assuming the heartbeat is still alive, is there a point when the active node
will fail over?

Is there a good white paper or kb article that can help shed some light on
this question?

Thanks,
John

Reply to this Message

Mike Epprecht \(SQL MVP\) - 28 Jun 2005 20:33 GMT

Hi

When the connectivity is lost at both IP and SAN level to the other node, a
failover occurs.

If the server that is running the instance takes too long to respond to a is
alive.

Clustering does not care about a bad application that is bleeding RAM, as
long as the server responds to is alives, it stays in the cluster.

If the application is cluster compliant, then clustering services can manage
it, otherwise clustering just cares that the server is up. Clustering
compliance involves WMI and other API requirements from the application.
MSDN has the information on it.

SQL Server is cluster compliant.

Regards
--------------------------------
Mike Epprecht, Microsoft SQL Server MVP
Zurich, Switzerland

IM: mike@epprecht.net

MVP Program: http://www.microsoft.com/mvp

Blog: http://www.msmvps.com/epprecht/

> What causes SQL Server to fail from the active node other than the
> heartbeat
[quoted text clipped - 17 lines]
> Thanks,
> John

Reply to this Message

Charles Tolento - 29 Jun 2005 14:59 GMT

TJHazel

Here is a link to some SQL Cluster Failover tests.

http://www.sql-server-performance.com/sqlserver2000_clustering_install_part4.asp

Also if you are able to unplug the heartbeat from the current active node
and failover occurrs you need to setup your public nic for "All
communications". This will allow the public nic to handel the hearbeat if
the private network fails.

Check out KB 258750 for heartbeat configuration.

Regards

CT

> What causes SQL Server to fail from the active node other than the heartbeat
> flat-lining? I have the cluster installed and tested the heartbeat by
[quoted text clipped - 13 lines]
> Thanks,
> John

Reply to this Message