If it's not a sweep then I would guess it is a client issue, e.g. a
connection invoking a resource-hogging (or broken) procedure or UDF.
Examine your connections, isolate the client with the rogue thread,
and find out exactly what it is asking the server to do.

-AF

--- In ib-support@y..., "mgastde" <michael.gast@w...> wrote:
> Hi,
>
> we have frequently a curious problem in a productive system running
> Firebird SS version 1.0 (build 796) under Linux: The CPU usage of
one
> of the ibserver processes rises for a _long_ time (up to more than
20
> hours) to about 95%. The machine has really long response times in
> this phase. But there are no visible errors, neither in the client
app
> nor in the ibserver.log.
>
> SWEEP is set to 0
> DIALECT is set to 3
> gfix reports no errors
> The difference between the current transaction number and the OTN is
> small (usually not greater than 100).
>
> Any ideas? For example how to find the reason?
>
> Best regards
> Michael Gast