Richard Lu (3/26/2008)

---

I agree follow-on article(s) should be added.

As to the unexpected fail-over, it may indicate that there is a problem/situation in the cluster causing the is-alive check giving false alarm. Usually there could be a problem with the private (team) heart-beat, or the system was in a heavy load, which timed out the is-alive connections during the time span specified. And obviously event logs and the cluster log should be examined for further findings. The automatic failover configuration itself doesn't cause the problem but is a reaction to the is-alive's result. And as a result, the failover was unexpected because of the false alarm. This is a bad example for choosing the automatic failover option. In this case, either resolve the false alarm issue, or disable the automatic fail-over option should be done. This article simply laid out different options and the expecting behaviors along the triggering flow to help people design a proper failover scheme to fit their own needs.