Failover issues

[Simon Hobson]

Shlomo Dubrowin wrote:
>OK,  So I have failover configured and the primary is giving out leases as
>it should.  Now, I want to test the failover.  What I have been doing is
>shutting down the service on the primary and expecting the secondary to
>automatically take over.  However, in the leases file on the secondary, I'm
>seeing that it believes it's own connectivity is down and not that of the
>primary.  So, when the primary fails, the secondary thinks it has a problem
>with itself and doesn't give out leases.  Any ideas why this isn't working
>as expected?

Does it by any chance go into "Communications interrupted" mode ?

I'm sure one of the failover experts will correct me if I'm wrong, 
but the servers will not go into partner down mode automatically - 
this is something you have to do manually. The reason given is that 
there are failure modes (in particular, routing errors or failure of 
an out-of-band inter server link) which can cause communications 
interrupted but leave both servers able to communicate with clients - 
having both server go into partner down and offering leases across 
the whole range would result in "a certain amount of confusion".

Dig back through the archives, there's been a number of suggestions 
for methods to monitor the peer and put the local server into partner 
down via omapi (IIRC).

Simon