Failover and pools with single machines

[Simon Hobson]

Tina Siegenthaler wrote:

>  > On the other hand, with non-failover (independent)
>  > pools, as
>>  Simon suggested, the lease files can more often hold differing
>>  information: one
>>  lease file showing the address assignment while the other does not.

>The setup has now been running for ten minutes, and it seems that the 
>leases are logged identically on both servers, even the "fixed" ones. 
>But it's only ten minutes... I'll have to check again on Monday and
>compare the logs and lease files.

The leases files will NOT be the same.

What will happen is that both servers will respond to a broadcast 
discover, and enter a lease into their database with a life of 2 
minutes. The client will then request the lease from ONE of the 
servers and that server will enter a new lease entry with the real 
expiry time. The server that didn't get to hand out the full lease 
will expire it's offer after two minutes and mark the address as free.

So when you analyse the leases to see what's active etc, you need to 
aggregate the leases files from both servers and look only at the 
latest leases. IFF you analysis program looks at the timestamp for 
the lease entries and correctly ignores ones that are earlier than 
it's already found, then you may be able to get away with 
concatenating the leases files from both servers and analysing the 
result.