I've got a Big-IP pool managing a couple SQL servers in an active/inactive rollover approach - there's only ever one in the pool at a time.
Initially I was enabling/disabling the nodes in the pool, but I found that due to connection pooling, active connections did not cut over.
Instead, I've used their MSSQL monitor to flag which server is active/inactive, so the nodes are actually taken "Down" instead of just "Disabled". Things work better, except that the very first query after a rollover errors out, as the connection is dropped. I've played with the Pool's "Action On Service Down" settings, but none of the options (None/Drop/Reject/Reselect) seems to solve the problem - first call for each active pool after the switch always returns a connection error.
Short of turning off connection pooling, any ideas for solving this?