File transfer corruptions but no error indicatios - Network

This is a discussion on File transfer corruptions but no error indicatios - Network ; I have a worrying problem transferring files. The probability of transferring a 15MB file successfully is about 50%. In each failed transfer the correct number of bytes are received, but one or more are wrong. The thing that worries me ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: File transfer corruptions but no error indicatios

  1. File transfer corruptions but no error indicatios

    I have a worrying problem transferring files. The probability of
    transferring a 15MB file successfully is about 50%. In each failed transfer
    the correct number of bytes are received, but one or more are wrong. The
    thing that worries me is that no error is reported, either by the command
    (ie ftp, rcp, or rdist) or by the network adapter (eg netstat -i ro ip -s -s
    link ls eth0).
    This problem exists when copying files between Linux machines (Red Hat 8 on
    Acer 4100) or between IBM machines (AIX 4.3.3 on p-Series).
    The common issue seems to be that the network includes a 3Com Switches
    (Super Stack 2 model 3300) which have worked fine for years but now are
    corrupting the data.
    It surprises me that I can transfer a 15MB file apparently with success but
    that the resultant file contains a corruption. Surely the transport layer
    should catch this problem and request the data to be sent again?
    Can anyone explain how this might occur?



  2. Re: File transfer corruptions but no error indicatios

    "Jeffrey Ross" wrote in message
    news:42536ac3$1@news.iconz.co.nz...
    > I have a worrying problem transferring files. The probability of
    > transferring a 15MB file successfully is about 50%. In each failed

    transfer
    > the correct number of bytes are received, but one or more are wrong. The
    > thing that worries me is that no error is reported, either by the command
    > (ie ftp, rcp, or rdist) or by the network adapter (eg netstat -i ro

    ip -s -s
    > link ls eth0).
    > This problem exists when copying files between Linux machines (Red Hat 8

    on
    > Acer 4100) or between IBM machines (AIX 4.3.3 on p-Series).
    > The common issue seems to be that the network includes a 3Com Switches
    > (Super Stack 2 model 3300) which have worked fine for years but now are
    > corrupting the data.
    > It surprises me that I can transfer a 15MB file apparently with success

    but
    > that the resultant file contains a corruption. Surely the transport layer
    > should catch this problem and request the data to be sent again?
    > Can anyone explain how this might occur?
    >

    Problem turned out to be a double corruption in the same packet where the
    TCP checksum (sum of the values of each of the packet's bytes) was not
    affected. Looks like we must have lots of network errors on a Switch and
    occasionally one (ie the double corruption type) gets past the checks.
    Nasty.



+ Reply to Thread