Cluster reboot best practice?

[rick26]

I want to reboot an active/passive cluster (W2K3) on a regular basis and unattended. Looking for best practice solutions to do this. Application (Exchange 2003) must stay up an running.

 

[xmsre]

reboot the inactive node. Why would you want to reboot the cluster on a regular basis?

 

[pgoss]

I agree reboot the inactive node then transfer resources then reboot the other node. What is the point of rebooting the cluster? I have a Active/Active Win 2003 cluster and the only time i reboot it is for critical patchs.

 

[raz2xa]

Windows (despite the w2k3 hype) is not yet a true 99.999 operating system. A regular reboot can ensure you are not caught out by untrapped memory leaks etc etc. Tho i am not aware of many places that do this with clusters! You should also be aware unless the SAN is configured properly a server may boot up with different fibre channel address which may have an effect on LUN numbering.

I dont have an immediate answer to that question. Windows now has powerful scipting capabilites that should allow you to use scheduler to set events that ensure you reboot the right node in the cluster at the right time. Search the web for scripting solutions that you could use or have a look at MSDN web site on scripting solutions.

 

[xmsre]

You'd be much better off tracking down the leaky app and taking the issue up with the vendor. If you identify the leaky app, and the vendor is no help, then consider service restarts instead of rebooting the box.

Datacenter is the version that advertises the 5 9s. They do this by only supporting specific builds. If you change the build in any way, it's not supported. It may seem limiting, but it's the only way to ensure that a given combination of MS an third party products will perform as advertised.