SQL Cluster more than just hardware protection

jibbajabba

I assumed that a cluster is more like a protection against hardware failures. It loses the heartbeat - it fails over - job done.

Now someone told me that a SQL Cluster also fails over when the services itself fail ..

Is that true ? If so, how does that work ?

bertieb

I assumed that a cluster is more like a protection against hardware failures. It loses the heartbeat - it fails over - job done.

Kinda, but there are lots of instances when a Microsoft cluster will initiate will failover and not necessarily just because a box blows up in a puff of smoke (You can also set up multiple 'heartbeat' connections and specify using a 'public' network as a backup to send the cluster heartbeat traffic for more resilience in this area, you want cluster node comms to be as reliable as possible to help stability and uptime)

Now someone told me that a SQL Cluster also fails over when the services itself fail .. Is that true ? If so, how does that work ?

Depends on the setup, but if SQL was to crash with the service failing it would initiate a failover. Infact if any items within the SQL resource group were to 'fail' (i.e. a SQL disk, sql server service, sql agent, ip address etc with the cluster services marking them as failed) it would initiate a failover in order to attempt recovery*. We had an ancient SQL2000 cluster do this last night after the SQL service crashed.

*You can do a lot of tweakage to define how the cluster reacts and threshold levels for a wide range of areas so this statement shouldnt be taken as a 100% guarantee. It all depends on the particular setup. You should (IMO) always take the notion of a Microsoft cluster as a means of high availability and protection against hardware failures though.