TCP Client dies - packet loss over VPN?

[itcamefrommars]

OK... how to trouble shoot a network with intermediate experience.

I have a TCP Client going over a router to router VPN that is freezing up.
I'm told the app is sensitive packet loss.
I've set up this same scenario before many times with no problem. My host side seems fine. The client app (basically a custom telnet) stops screen paints at the remote. Originally I thought it was a bandwidth issue.
But this is apparently a compound problem. Bandwidth is higher now, but it still happens.

Now i get about 4% packet loss from my UNIX box to the remote workstation after about 1000 pings. I have replaced cables and swapped out equipment (Including a test server). I have about 6300 (or was it 63000) collisions in the last month reported by the server. Every time i unplug or reset a router w/ switch or hub/switch I get a few hundred collisions. Is this normal? (I was told I should have 0)

Put a sniffer on and I don't see anything WAY out of the ordinary... using ethereal and I'm not VERY versed in network trouble shooting.

Any idea's on things to try to find my network inconsistencies causing this?



ok the setup:

HOST (1 server, 2 workstations)
Cable->Router->UNIX

Remote (1 workstation)
DSL->Router->WinXP

Routers are Netopia R9100
UNIX is SCO Openserver 5.0.5
DSL is 1.4 down 256 up
Cable is 965 down 348 up (or close to that)
VPN is IPSec w/ MD5 (tried pptp with and without encrypt)

Thank you so very much.... looking for the next foothold!

Marshall

 

[ZSwen]

what did the sniffer and ethereal report? if packet loss , it must report something..

make sure router interface and unix interface full duplex.
dont set it to auto.

what kind of collision?

 

[itcamefrommars]

ethereal is the sniffer im using... didn't see any packet loss there. i'm not an expert at using it though.

i really don't know different kinds of collisions or how to detect them.
SCO openserver's 'netstat -i' shows collisions.

ok... obviously this problem is at least a little beyond me.
where to i start or what resources are there out there to learn me?

tank u.

 

[itcamefrommars]

i may have a solution.... i'll post tommorrow for those interested if it works.

 

[asassin]

Marshall I would like to see your solution as I am experienceing a similar problem at work. I have about 11 VPN users who all connect via cable or DSL to the corporate network through SHIVA client software. The problem they are having is with file transfers from this 1 particular server which is WIN2k Adv server. It has 2 NICS(IP's) that are registered in our corporate DNS to point to the hostname "NCCIC1TFDS01". We have had the networking department run sniffor pro on the nics to anylize the packing information. Presently they can not seem to get a complete file transfer to occur no matter which file it may be or what size it may be. They are not losing their VPN connection but it seems their TCP session is resetting and not allowing them to complete the file transfer. We have already made sure the NIC settings are set to 100/FULL which is the corporate standard.

Any ideas how to fix this problem?