Helen,
As I see in most of this answers, there is a common suspect about UDF...
Developers use the Rfunc that has any years without upgrading (since 2003),
when FB 2.1 was not in the field.
How can we check if there is the problem of the random crashes? We have
downloaded in the Suse Linux and compiled there before to install it.

> b) you're trying to use the non-NPTL package on a Linux where NPTL is the

active threading model.
We have installed a non-NPTL package while the OS supports them. I thought
that there could be no problem with this, just threading would work in the
normal speed, but not to crash the FB server. Isn't it?

Ramiro

2009/2/3 Helen Borrie <helebor@...>

> At 11:42 AM 3/02/2009, you wrote:
> >We have a pool of more than 600 users accessing some FB 2.1.1
> >databases on a Open SUSE Linux server from different win32 apps
> >(Delphi 7 and VB6).
>
> Superserver?
>
> >We have migrated those databases from FB 1.5 since there was a
> >performance problem with large DB sql queries and stored procedures
> >(one of this DBs has a single file of more than 9 GB with a table of
> >more than 25 millions of records).
> >After the migration, the server begins with random crashes
> >("terminated abnormally (-1)" message on firebird.log).
> >Unfortunatelly, this log is not enough to detect what is the problem.
> >After some surfing through several forums we detected that problems
> >could be among:
> >1) incompatibility of clients (gds32.dll, fbclient.dll, etc.)
>
> No
>
> >2) incompatibility of UDFs (i.e.: rfunc)
>
> Yes
>
> >3) Network disconnection of some clients (errors 104, 32, etc. in the
> >log).
>
> No - these are symptoms, not causes
>
> >4) Incompatibility of components used in the source code (Zeos,
> >Intersol, IProvider, etc.)
>
> No
>
> >5) Queries that are restricted from FB 2.0 (table alias, etc.)
>
> No. Exceptions, yes.
>
> >6) Linux rights over some files or folders? (errors while writing to
> >FDB datafiles).
>
> No. If permissions were the problem, it wouldn't cause random crashes. It
> just wouldn't work.
>
> >7) Any other I forget?
>
> Assuming Superserver (because of the Guardian message you're seeing in the
> log)
>
> 8) OS threading model. You'll get "random crashes" if
>
> a) you're trying to use the NPTL package on a Linux where NTPL is broken.
> b) you're trying to use the non-NPTL package on a Linux where NPTL is the
> active threading model.
>
> 9) You're getting an out-of-memory condition under load with 600 users
> threading off one Superserver process.
>
> 10) You are running some admin tool alongside regular user applications,
> that is keeping a MON$ transaction running and consuming resources
>
> ./heLen
>
>
>

--
Ramiro Barreca
rbarreca@...


[Non-text portions of this message have been removed]