>
> Deadlocks have been described in great detail - essentially,
> a deadlock occurs when two or more transactions are waiting
> for resources held by the others. All are waiting, none can
> release resources, so the database management system must
> intervene to unblock the system.
>
> If you're encountering situations where a transaction is
> blocked and nothing happens for more than 10 seconds, then
> the problem is likely to be a long-lived update transaction
> that blocks subsequent updates.

Will this happen during the start of the transaction or later during the
updates? We have the problem that the call to isc_start_multiple will hang
for more than 2 min. After this period i try to close the connection from
another thread but this needs more than 7 min.

Could this happen because we have one firebird instance serving 4 different
databases? Would it better to have a single firebird instance per database?

We do not have log-lived updates - but we have a lot of short running insert
statements for all of them (4 * 14 inserts per sec). This inserts are
surounded from StartTransaction, commit.

Could this lead to the above mentioned lock too?


Thanks for the help.
Henner Kollmann