ftp transfer stops after 200mb

[Warren griggs]

Hi,

I have a server running HPUX 11. When I try to transfer a large file to another server (hpux 11) after 200mb is transferred (always 200mb) the session hangs with acknowledgement failures and attempting re-transmissions and eventually timesout.

I have tried this transfer to other servers and have narrowed it down to this one server. It has a 4 port etherenet card running auto-port aggregation @ 100MB Full Duplex.

Does anybody know why this is happenening?

Thanks in advance,
Colin.

[U.SivaKumar_2]

Hi,

Have you analyzed the port bandwidth utilization of the swith to which server is connected ?

Any packet drops ?

What mode are you using ? binary or ascii ?

regards,

U.SivaKumar

[melvyn burnard]

This may be COMPLETELY off track, but I had a call that I dealt with a few years ago where the customer could xfer files up to a certain size on his one of many servers, all using FDDI cards. Turned out to be the FDDI card on the server next along in the ring.
Could it be the card, perhaps?
Other than that, one assumes that APA and networking are patched to the latest level?

[Warren griggs]

I'm somewhat confused now. I re-ran the test so that we could sniff the network to check if there were any packets being dropped.... Hey Presto it works now? Nothing has changed on the HP Servers so I guess that this maybe has something to do with the network. My network admin tells me nothing has changed since yesterday afternoon. I'll keep an eye on it and see if I get it again.

Thanks,
Colin.

[Massimo Bianchi]

Hi,
i had a similar problem.

The file went wrong, but if we zipped it it worked.

the zipped was only a few kilobites less...

We think it was kernel issue, related with the stream of bytes under it mis-interpreted.

we solved with latest streams/arpa patch.

does rcp works ?

HTH,
Massimo

[Michael Steele_2]

Since you've isolated it to one server pop in a cross over cable and run some performance tests with large files.

For if its the NIC as you're suggesting then you should be able to isolate it.

However, there have been 'numerous' complaints about 'auto' anything when configuring speed and duplex on NICs. Most common is conflict with the switch. So bear this in mind and consider 100FD or 100HD instead of anything 'autonegotiated'.

lanscan (* note NMID / PPA *)
lanadmin -x NMID/PPA
lanadmin -x 4
lanadmin -X 100HD

Also seen a lot of debate about HD over FD.

[Jeffrey Killian]

Try looking at the packet size on the two systems. We regularly transfer 2-3GB files (and larger) between our HP and Windows servers... We had similar problems with timeouts as you are experiencing until we determined the the default packet size on the HP was 64K and on the windows machines 4K...

[Jeff Schussele]

Hi Colin,

Possibility exists that this is a connection problem. You might consider running
inetd -l
to up the connection logging to see if it catches a dropped connection.

My 2 cents,
Jeff