OS2 World.Com Forum

Subject  :  Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  02 Jul, 2006 on 15:10
I have a PPro-200, 256Mb RAM (4x64), 2xPromise Ultra-133 controller with 4 hard drives attached to them:
500Gb, 2x 250Gb, 1x120Gb.

The eCS itself is on a small 2Gb drive on the motherboard IDE port.

The NIC is an RTL 8169, used for it's 1000Bsp capacity.

The protocols are the followings:

-Netbios.
-Netbios ofer TCP/IP.
-TCP/IP.

The IP is provided by my IP-Cop firewall DHCP server.

The 250Gb drives and the 120Gb drives are used to store datas. The 500Gb drive is used as backup. Once a week, the 500Gb drive is mounted, and a differential copy is done from the 3 drives to the 500Gb drive.

I have actually about 400 Gb in my 3 "small" hard drives.

an old HP LJ4+ is connected by ethernet, and is shared trought eCS. In case of eCS problem, my other PC can access the LJ4 directly trought a TCP port.

A small FTP server and Web/2 server are running on this eCS PC. These are published to the internet.


The problem:
Each about 10 days, there is a network problem. The ressources aren't accessible again...
When I look at my eCS, at first view, everyhing seem ok. The PC is running, my TCP tools seem to be there.
But the DHCP monitor tell me that the DHCP client don't run. When I relaunch the DHCP client (dhcpstrt.exe), then it wait indefinitely to get the IP and the lease.

I have to reboot my PC, and then everything run perfectly again. During about 10 days...

Is there a log or an event viewver somewhere which help me to see what in this #ç%¦@$ doesn't work?

Is this the NIC? Or a memory fault? or the processor which would be too hot? Or my eCS network part which is corupted? Or something else? Anything??? How can I have an idea?

Note: I allready had some problems some time ago, then I reinstalled all the system, and then I used Partition Saving to do a exact copy of my system drive on an identic drive. For this reason, I think it's a hardware problem.

If somebody have an idea to debug this PC, I would be happy to hear it.


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Ben Dragon
Date  :  02 Jul, 2006 on 16:08
Post your setup.cmd from your *\mptn\bin directory.

---

Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  02 Jul, 2006 on 18:33
Here you have:
The "setup.cmd" in c:\mptm\bin
------
route -fh
arp -f
ifconfig lo 127.0.0.1
REM ifconfig lan0
REM ifconfig lan1
REM ifconfig lan2
REM ifconfig lan3
REM ifconfig lan4
REM ifconfig lan5
REM ifconfig lan6
REM ifconfig lan7
dhcpstrt -i lan0
REM ifconfig sl0
===========
The one in c:\tcpip\sample\bin
---
route -fh
arp -f
ifconfig lo 127.0.0.1
REM ifconfig lan0
REM ifconfig lan1
REM ifconfig lan2
REM ifconfig lan3
REM ifconfig lan4
REM ifconfig lan5
REM ifconfig lan6
REM ifconfig lan7
REM ifconfig sl0
ipgate off

Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Fahrvenugen
Date  :  03 Jul, 2006 on 06:38
First, I havn't worked with a system built with your specific hardware, so right now I'm taking a few guesses...

To start, what is your Keepalive set at for your routing table? If it is still the default, OS/2's tcp/ip stack sets an excessively long keepalive value for the routing table, and sometimes the table gets full quickly. To check the current value, use:

[C:\]inetcfg -g keepalive
$Inetcfg: CURRENT DEFAULT MINIMUM MAXIMUM
Keepalive 60 7800 0 7800 KeepAlive (sec)

If your keepalive is still at 7800, then add the following line to your startup.cmd (or your \mptn\bin\setup.cmd):

inetcfg -s keepalive 60

This will help keep your routing table from getting filled with stuff and slowing things down.

Next, are you using the latest driver for your RTL8169? The current build is at 1.11. Check your build level with:

[C:\]bldlevel c:\ibmcom\macs\rtgnda.os2

If your revision isn't 1.11, then head over to:

http://www.os2warp.be/index2.php?name=gigabitdrvs#4

And grab the latest version of that driver and see if it helps (I have no clue what version of the driver shipped with ecs1.1. With 1.2 it is the older 1.06 version)

Hope this helps!


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  03 Jul, 2006 on 10:30
I created a small batch file which does the
"inetcfg -s keepalive 60" at each PC start, as I read that if I put it in the setup.cmd, then this modification will be discarded the next time I use the "TCP/IP Configuration".

For the NIC drivers:
- I downloaded the last driver on the site you gave me. On the MPTM card and protocol configuration.
- I did "Have Disk" and pointed on the folder with the unzipped files. No new card apppears, probably (I hope) becasue the card allready existed, it's just the old driver which were overwritten.
- I selected the "Realtek 8169 MAC driver" in the stage 1 and stage 3, and pressed the "Change" button.
- Exit and reboot.
I can just hope that the driver was really updated...

Small note:
- As you see, I am not really used with eCS/OS2 network configuration...

Other small related question:
- How may I see the version of an installed driver???
- Isn't there any log file to see what didn't went wrong somethere?

About the impact on the stability of these fine-tuning... Well, we can only see in 1-2 weeks how it (doesn't ? ) crash.


Thank for your help.


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Fahrvenugen
Date  :  03 Jul, 2006 on 17:35
Hi again,

The easiest way to find out the installed version level of the driver is with the bldlevel command. At an OS/2 command prompt type:

bldlevel c:\ibmcom\macs\rtgnda.os2

It should give an output something like:

Build Level Display Facility Version 6.12.675 Sep 25 2001
© Copyright IBM Corporation 1993-2001
Signature: @#chagrin:1.11#@Realtek RTL8169 OS/2 NDIS 2.0.1 MAC Driver
Vendor: chagrin
Revision: 1.11
File Version: 1.11
Description: Realtek RTL8169 OS/2 NDIS 2.0.1 MAC Driver

*BUT* If the output is:


Build Level Display Facility Version 6.12.675 Sep 25 2001
© Copyright IBM Corporation 1993-2001
Signature: @#chagrin:1.06#@Realtek RTL8169 OS/2 NDIS 2.0.1 MAC Driver
Vendor: chagrin
Revision: 1.06
File Version: 1.6
Description: Realtek RTL8169 OS/2 NDIS 2.0.1 MAC Driver

Then you have the older version (version 1.06, or if the Revision level is older then 1.06 - there is also a 1.05 build) of the driver installed, you'll want to try the newer version.

One way to *manually* install the new driver is,

1. Unzip the rtgnda11.zip into a temporary directory

2. Copy the files rtgnda.os2 and rtgnda.nif into the following directory:

c:\ibmcom\macs\

3. Check your file version with the bldlevel command as above.

4. Reboot

Again, I've never worked with that particular network card chipset under OS/2, so I don't know if this is the difficulty or not. But it is worthwhile to give it a try.

Another thought to try is to dump the DHCP part, and just assign a static IP address to your network card (if you can do that on your network). I doubt DHCP is causing the difficulty, but if it is, a static IP might solve the difficulty.


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Fahrvenugen
Date  :  04 Jul, 2006 on 00:07
I just had another thought.

What TCP/IP stack are you running? From a command prompt, type:

inetver

You shoudl get an output like:

Version numbers of TCP/IP protocol drivers:
SOCKETS.SYS: 6.3100
AFOS2.SYS: 6.3100
AFINET.SYS: 6.3100

If you have an earlier stack, try the stack update, newstack.zip at:

http://hobbes.nmsu.edu/cgi-bin/h-search?sh=1&button=Search&key=newstack&stype=all&sort=type&dir=/

instructions for installing are in the readme in that zip.

As for your other questions regarding a potential memory or processor heat issue, I doubt these are at fault. Generally if you have a memory issue, you'll get the machine crashing or freezing up. From what you've described, the machine is still running, it is just the network which has stopped working.

Hope this helps.


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  04 Jul, 2006 on 11:32
Ok.
I have the last version of the TCP/Ip and of the nic driver

Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Fahrvenugen
Date  :  04 Jul, 2006 on 19:05
Sounds good, let's hope that solved the problem! (guess it'll take a couple of weeks to find out... let us know the results. )

Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  05 Jul, 2006 on 09:04
We will see. And yes, I will tell you.

Anyway, thanky you really for your help.


V.


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  13 Jul, 2006 on 12:02
Again, after 7 days... The network seem down. No IP, no DHCP client, the shared ressources arn't visible any more...

On my OS" PC, at the first view, verything seem to be ok. Simply the networks elements don't seem to be there any more.

I did:
"net stop peer" followed by a "net stop req". I have to admit that I didn't had a look at the eventual error messages...$

Then I did:
"net start req" -> no problem.
"net start peer" -> error message internal error ... type HELP NET3057

Then I typed "help net3057" and read the error message. Except rebooting the PC an calling IBM for help (Really? I can do that!?) there is no explenation or solution.

I asked eCS to reboot, and he did it withotu problems. Now the net works elements of my eCS should run fine and quickly... during at least 5 days...

I got an email from a friend about a serie of the RT8169 Gigabite NIC had problems... I am not sure if this info is trustable.

When I bought my rt8169 cards, I bought 2 cards. One for my eCS fileserver and one for my WinXP work PC. Interrestignly, the MAC adress were 2 following numbers

The card installed on my Windows PC work only during 5 minutes. As the card becomes hot quickly, I thought that maybe this card was damaged, and I put in a standard 8139. Then everything works fine. At only 100BPs...

I will try, as soon as I have time, to put a 8139 again in my eCS file server. And see if that correct the situation...

Did anybody heard about a defect serie of 8169 NIC?


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Fahrvenugen
Date  :  13 Jul, 2006 on 17:56
Hmm... very strange.

The things which come to mind at this point is to either try a static IP address (dump the DHCP portion of the setup), try different network card (maybe something which isn't based on the RTL8169) or to try and remove and re-install all the networking stuff (TCP/IP, LAN / Peer services, MPTS). Even if you've just got an older 10/100 network card kicking around (I can't speak for others, but I somehow seem to end up with a few of these spare, no matter how many I give to other people or install in machines...), it may be worthwhile to drop it into that machine for a week or so and see if the problem sitll occurs. It'll run slower on the network, but if it runs longer then a week or so, then you'll know the card is causing the difficulty.

There are some newsgroup posts of people with that NET error message running HPFS386, but I don't think that's an issue, unless you've installed HPFS386 (which isn't on the base eCS).

I'm almost wondering if someting is messed up with the installation. But it may also be the card. When I googled for "rtl8169 network freeze", there seemed to be a few pages I found where people have experienced problems getting this card to work properly under Linux and the 64 bit Windows. Of course OS/2 is different from Linux, but it may indicate potential hardware issues.


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  vingfel vingfel@excite.com
Date  :  13 Jul, 2006 on 20:40
Did I mentionned that I tried long time ago to put my eCS in fixed IP adress? Well, no, it seem that I didn't... Well, no, in fact I didn't told that...

Anyway... The peer service worked during 5-12 days, and then it'snot present any more, and it refuse to start...
The peer service isn't related to the TCP/IP, right???

Then it seem to me that the problem isn't just a TCP/IP config problem. It's something deeper than that...

Okay. Putting back a 10/100 PCI card in my eCS.
Let's see...


Subject  :  Re:Crashing network of eCS1.1 each 8-12 days.
Author  :  Fahrvenugen
Date  :  14 Jul, 2006 on 04:46
You're right. If it is causing problems with both peer and TCP/IP, then it isn't likely a TCP/IP problem (unless it is something wrong with the base network stack, which is a part of MPTS). But you've already looked at the stack. So in that case, a static IP won't likely make a difference.

I hope the 10/100 will solve the problem... it may not give the faster speed, but it may identify the problem.


Powered by UltraBoard 2000 <www.ub2k.com>