> Has anyone experienced anything like this. I have a firebird
> 1.5.2 version running on a dual cpu machine running windows
> 2000 server. There are approximately 30 users that logon via
> my application, an accounting/ inventory/ purchasing/ ordering
> etc system.

I have experienced apparent freezing (with 0 cpu use) on dual
CPU machines but only under specific circumstances...

- doing a transfer from one database to another

- lasting from a few seconds to a minute or more

- any other activity over the network to and database
on that server unfreezes the system for a time

- only occurs on boxes that are HT (hyperthreading) capable
(may still happen on dual systems even with HT turned off
but does not happen on dual systems that are not HT
capable.)

On single CPU systems disabling HT is a reliable way to get
around the problem. On dual CPU systems with HT capabilities
it actually seems to help to have HT enabled.

The problem you describe sounds quite different to what I have
just said. I have never seen the problem happen during "normal"
activity against the server (indeed "normal" activity is one
way of breaking the apparent deadlock described above).


Check what you have set for sweeping. Perhaps disabling it and
scheduling it up to run only at night - dont forget that heavy
disk activity can slow (but not usually stop) a system without
showing much CPU. So also check for other regular activity
that may be consuming the disk.

I guess the other thing to consider are hardware issues, could
the network card on the system be suffering from problems?

In the short term you could try moving to another box to see
how it behaves. Set up on a single CPU system with no HT and
see whether the problems persist. (I do know that this is far
from convenient, its like telling users to reinstall Windows
to see if that fixes their problems. :-)

Not sure what else.

--
Geoff Worboys
Telesis Computing