Hello,

We are running Pervasive server Engine 2000i SP4 v7.94 with
SQL/ODBC Hotfix on the following system :
os : Windows NT 4.0 sp6
memory : 1GByte

Server configuration :
Allocate Resource at startup : on
Back to minimal state if inactive : off
System cache : off

Cache allocation size : 299999232
Communications threads : 64
Number of Input/output threads :20

Database size : 250 Btr files (# 4 GByte).

Clients side :
200 Dos clients running on Windows NT 4.0 Workstation or Windows XP sp2
Workstation
10 JDBC clients running on Windows NT 4.0 Workstation or Windows XP sp2
Workstation

The problem:
In the task manager, the handles object from NTDBSMGR are periodicaly
increased from 1400 at startup to approx. 45000
and then NTDBSMGR crashes with DrWatson32 message.

Has someone any idea or has had the same problem?
A. Didiot

Sounds like a bad application. Check through each of the active users,
and see if the handles for each user connection is going up or not. I
am guessing that you will find one (or perhaps ALL) users are
increasingly using handles.

This can happen if an app developer opens a file, then fails to close
it, and runs this is some sort of processing loop. (By watching the
user/handle counts, you can usually figure out which process has bad
code.) When you track it down, get that info to the developer and ask
them to fix it.
Goldstar Software Inc.
Building on Btrieve(R) for the Future(SM)
Bill Bach
BillBach (AT) goldstarsoftware (DOT) com
http://www.goldstarsoftware.com
*** Chicago: Pervasive.SQL Service & Support - July 19-22 ***
*** Chicago: Pervasive DataExchange Class - July 27-29 ***

didiota (AT) wanadoo (DOT) fr wrote:

Thanks a lot for you answer.
With the a process viewer tool (from sysinternals) , we could see that
the handles are "Security Token" and not Btr file handles.
These "Security Token" remain open and new one are periodicaly open
.... we don't know why.
Thank again you for your intervention.
A. Didiot

That actually sounds more like you need the client patch than the SQL
engine patch. See:
43502 Session count could increase on your server engine if clients
are creating sessions and failing in their Winsock closeSocket calls
on their local client. Because Winsock is not working correctly and
the CloseSocket is failing, the session with the server is left open.

At least if I remember correctly. That was some time ago.

Leonard

On 30 Jun 2005 01:46:12 -0700, didiota (AT) wanadoo (DOT) fr wrote: