Paulo Fagiani wrote:

> ...
>
> Something else that I would apreciate if someone has an answer is that
> since we began to work with Interbase and now Firebird that every
> other week we gotta do a backup/restore to keep the db server
> healthy... otherwise it will get nuts somepoint and start to not
> finishing the connection's processes and the server will overload
> memory and processor... then only a backup/restore to solve it...

I don't have an answer to your "cannot start transaction" problem, so
I'll just comment on the "get nuts" part.
"get nuts" isn't a description that tells us a lot of what users are
doing and what is happening, but my guess is that you observe
"garbage being handled". Quite often, the problem is cured by
ensuring that no transaction stays open for too long, and "commit
retaining" is hardly ever used (if at all).
Could you post the "header statistics" just when the server "get nuts"?

--
Aage J.