I have a 2 node A/P cluster which has not yet gone live. During testing I've noticed that when a network failure is simulated failover is delayed for 3 mins. I've narrowed it down the Information Store causing the delay. If I reduce the 'pending timeout' for the IS resource the overall delay is correspondingly reduced. The worrying part is that if I pause the passive node, so it does not failover, and check the database headers with eseutil they are all shown as 'dirty'. It appears that the IS resource is not going offline cleanly and is being terminated after the 180 sec timeout. Is this expected behaviour for an Exchange 2003 cluster? -- Darylt
