RE: Connection Timeout/Reset by Peer while in GUI

Here's our setup:

BPCS 6.0.02 PLF FULL C/S March 98 Cum
AS/400 V4R2
Full TCP/IP network

WAN Connections for users in UK, Germany and Ireland connecting to Montreal
(CANADA). Speed of links varies from 112 to 256 Kbps Frame-Relay (CIR) not
dedicated to BPCS. 

We had some problems with timeout and "Reset by peer" problems. Typically
the problem lies on the AS/400 not responding fast enough (in one instance
we had a security issue with NEWILIB's objects that caused the connection to
be rejected quickly with "Connection reset by peer" messages. Typically, the
culprit for us is an ODBC connection (not done via BPCS) that takes too much
CPU on the AS/400 causing it to be sluggish in its response time. You need
to attack those with DBMON or PRTSQLINF and add some index in the database.
Also, we have found that NT, Windows 95 and Windows 98 each works with
different DLLs when it comes to TCP/IP. THere are versions of WINSOCK.DLL
that are cleaner than others, in our experience the following seems to be
true:
The following is a list of WINSOCK.DLL that have been tested
1. Win98  c:\windows\winsock.dll  1/15/99  21504     - Works
2. Win98  c:\windows\winsock.dll 24/08/96 -              - Does not work 
3. Win95  c:\windows\winsock.dll  24/08/96  42368    - Works
4. Win95  c:\windows\winsock.dll  3/20/98    21520    - Works


BPCS Client/server - like many client/server programs out there - is
sensitive to the latency of your network i.e. the ping time between the PC
and the AS/400. We found that a ping time in excess of 150 ms will typically
translate into problems. BPCS is not really bandwidth hungry but does
require good latency.

Doing a NETSTAT, option 3 on the AS/400 is a starting point for looking at
these recurring problems. Check for the connection on port 5000 (DFS apps
like CEA and COM) or port 9007 for OPS applications (if you are mixed mode
you don't care about OPS). Obviously you may have installed your daemon on
different ports check your BPCS startup script.

Once you have located the connection check the details, to find out the
round-trip time and round-trip variance. Also check for high number of
retransmissions. 

Routing issues normally requires you to run TRACERT on the Windows 95,98 box
to find out what is going on. Name resolution is typically not a problem as
BPCS DFS applications only supports IP addresses not Host names. OPS will
however be able to resolve names properly. 

Marc Chartrand

-----Original Message-----
From: George Sagen [mailto:gsagen@primesourcetech.com]
Sent: Wednesday, July 28, 1999 1:00 PM
To: BPCS-L@midrange.com
Subject: RE: Connection Timeout/Reset by Peer while in GUI


Martyn,

Connection timeouts and connections being reset by peer are more likely
networking issues or index problems, not PC hardware performance issues. I
wouldn't use resources on upgrading the PC hardware with an expectation that
timeout issues will go away, though I might do it to speed up CEA and COM in
general.

Regarding network problems with the PC connection timing out, I once saw on
6.0.02 PC's taking a long time to connect or timing out when DHCP was turned
on for internet addressing purposes. When the user wasn't connected to the
internet it would search for the DHCP server and be unable to find it.
Eventually the PC would use some other service for IP address resolution,
but BPCS would sometimes time out before that. Turn off DHCP and suddenly
everything connected fine. I also recently saw a system where the router
table on the AS/400 was incorrectly configured to look for addresses on the
wrong segments across the WAN before looking on the LAN. Slowed down the
connect time so it took over 30 seconds just to load a program (6.1 GUI).

Regarding index problems, there have been many recent discussions on this
list about how having the right indexes can speed up data access and
sometimes avert a fatal timeout by helping the programs avoid sequential
file scans.

How can you tell if your problem is likely network or index problems? If it
takes a long time to load all your programs or they timeout before they can
load, you probably have a network problem. Also, if while in the middle of
working, you are spontaneously disconnected with no apparent pattern, you
may have duplicate IP addresses. Find out if your network uses static IP
addressing or some service that assigns addresses. If static, two or more
nodes likely have the same address. Of course you may have an evil IT weenie
randomly yanking patch cables from the hub and cackling to himself as users
across the building start swearing (I know you're reading this, Nevin).  If
your programs load okay, but some of them are just slow and eventually time
out, you probably have an index problem.

Bottom line...I wouldn't upgrade the PC's for your timeout issue.

Best wishes.

Geo.

George Sagen
BPCS Application Consultant
gsagen@primesourcetech.com <mailto:gsagen@primesourcetech.com>
http://www.primesourcetech.com 

PrimeSource Technologies, L.L.C.
7373 East Doubletree Ranch Road
Suite 150
Scottsdale, AZ 85258
(801)360-6360 Direct & VM 


.  -----Original Message-----
.  From: uucp@Uucp1.mcs.net [mailto:uucp@Uucp1.mcs.net]On Behalf Of Martyn
.  Barratt
.  Sent: Wednesday, July 28, 1999 2:19 AM
.  To: 'BPCS-L@midrange.com'
.  Subject: RE: Connection Timeout/Reset by Peer while in GUI
+---
| This is the BPCS Users Mailing List!
| To submit a new message, send your mail to BPCS-L@midrange.com.
| To subscribe to this list send email to BPCS-L-SUB@midrange.com.
| To unsubscribe from this list send email to BPCS-L-UNSUB@midrange.com.
| Questions should be directed to the list owner: dasmussen@aol.com
+---