Collisions with Full-Duplex

Discussion in 'Cisco' started by Jim, Dec 17, 2003.

  1. Jim

    Jim Guest

    Hello all! Was wondering if someone can clarify something:

    We have a 7206VXR router running Version 12.2(17)IOS with a PA-FE-TX
    (Fast-ethernet (TX-ISL) Port adapter, 1 port ) card. The interface is
    configured for 100Mbps and Full-Duplex as follows:

    interface FastEthernet2/0
    ip address 192.168.3.11 255.255.255.0
    ip directed-broadcast
    ip pim sparse-dense-mode
    duplex full
    no mop enabled

    The switch port it connects to (on a 6509) is also hard-coded for
    100/FULL

    Yet we are seeing collisions and under-runs:

    CHICR01#sh int fa2/0
    FastEthernet2/0 is up, line protocol is up
    36618378 packets output, 3198461469 bytes, 206 underruns
    206 output errors, 206 collisions, 0 interface resets

    Cisco TAC is indicating that "Regardless whether etherent is in full
    duplex or not, ethernet by nature - CSMA/CD (carrier sense multiple
    access with collision detection) still will check to see if the wire
    is busy."

    I always thought that the collision detection mechanism in Ethernet
    was disabled automatically when you configure a port/interface for
    Full-Duplex. Am
    I mistaken?

    TIA
    Regards,
    Jim
    Jim, Dec 17, 2003
    #1
    1. Advertising

  2. In article <>,
    Jim <> wrote:
    :We have a 7206VXR router running Version 12.2(17)IOS with a PA-FE-TX
    :(Fast-ethernet (TX-ISL) Port adapter, 1 port ) card. The interface is
    :configured for 100Mbps and Full-Duplex as follows:

    :The switch port it connects to (on a 6509) is also hard-coded for
    :100/FULL

    :Yet we are seeing collisions and under-runs:

    :CHICR01#sh int fa2/0
    :FastEthernet2/0 is up, line protocol is up
    : 36618378 packets output, 3198461469 bytes, 206 underruns
    : 206 output errors, 206 collisions, 0 interface resets

    Notice that the underrun, output error, and collision counts are all
    identical.


    :Cisco TAC is indicating that "Regardless whether etherent is in full
    :duplex or not, ethernet by nature - CSMA/CD (carrier sense multiple
    :access with collision detection) still will check to see if the wire
    :is busy."

    Even if they are right, the line is never going to -be- busy,
    unless you have some kind of strange tap, or a wiring problem... or,
    of course, a signalling problem. And the chances that every underrun
    would be exactly matched by an output error and colision is very small.
    The only reasonable conclusion is that -something- is incrementing all
    the counters at the same time.

    I seem to recall that I saw someone mention a problem earlier this year
    in which all the counters were incremented at the same time.
    That wouldn't explain why they are incremented at all, but it could
    indicate why the counters are all the same. Unfortunately, I do
    not recall any details about source of the multiple-counters problem.


    The situation reminds me of a problem with the 2900XL and 3500XL that
    was documented in September, about a failing electrolytic capacitor:

    http://www.cisco.com/warp/public/770/fn26174.shtml
    Reported symptoms include network performance degradation and a
    high Frame Check Sequence (FCS) error count due to the use of this
    capacitor, as a filter capacitor, on power, provided to the network
    interfaces for the system. As the capacitor degrades, the network
    interface controllers will be unable to reliably decode packets
    from the network. Eventually the capacitor degradation will lead to
    failure of the system. It is not possible for any data corruption
    to occur because of this issue.

    This cannot -directly- apply to you as you do not have one of those
    devices, but it's an unexpected potential source of problems that
    bears thinking about if nothing else sensible comes forward.
    --
    Rome was built one paycheck at a time. -- Walter Roberson
    Walter Roberson, Dec 17, 2003
    #2
    1. Advertising

  3. Jim wrote:

    > Cisco TAC is indicating that "Regardless whether etherent is in full
    > duplex or not, ethernet by nature - CSMA/CD (carrier sense multiple
    > access with collision detection) still will check to see if the wire
    > is busy."


    That is a fine statement. But it does not clarify why in a full-duplex
    point-to-point link you see collisions. The channel can never be busy,
    simply because there is only 1 station on each wire pair able to
    transmit.

    How long is your cable? My guess is that you see the result of
    occasional cross-talk interference, or another interference problem.

    --
    Joop van der Velden
    Joop van der Velden, Dec 17, 2003
    #3
  4. Jim

    DMc Guest

    Make sure you have not use cat3 cabling by mistake.

    Dave.

    "Jim" <> wrote in message
    news:...
    > Hello all! Was wondering if someone can clarify something:
    >
    > We have a 7206VXR router running Version 12.2(17)IOS with a PA-FE-TX
    > (Fast-ethernet (TX-ISL) Port adapter, 1 port ) card. The interface is
    > configured for 100Mbps and Full-Duplex as follows:
    >
    > interface FastEthernet2/0
    > ip address 192.168.3.11 255.255.255.0
    > ip directed-broadcast
    > ip pim sparse-dense-mode
    > duplex full
    > no mop enabled
    >
    > The switch port it connects to (on a 6509) is also hard-coded for
    > 100/FULL
    >
    > Yet we are seeing collisions and under-runs:
    >
    > CHICR01#sh int fa2/0
    > FastEthernet2/0 is up, line protocol is up
    > 36618378 packets output, 3198461469 bytes, 206 underruns
    > 206 output errors, 206 collisions, 0 interface resets
    >
    > Cisco TAC is indicating that "Regardless whether etherent is in full
    > duplex or not, ethernet by nature - CSMA/CD (carrier sense multiple
    > access with collision detection) still will check to see if the wire
    > is busy."
    >
    > I always thought that the collision detection mechanism in Ethernet
    > was disabled automatically when you configure a port/interface for
    > Full-Duplex. Am
    > I mistaken?
    >
    > TIA
    > Regards,
    > Jim
    DMc, Dec 18, 2003
    #4
  5. Jim

    mikester Guest

    "DMc" <> wrote in message news:<>...
    > Make sure you have not use cat3 cabling by mistake.
    >
    > Dave.
    >
    > "Jim" <> wrote in message
    > news:...
    > > Hello all! Was wondering if someone can clarify something:
    > >
    > > We have a 7206VXR router running Version 12.2(17)IOS with a PA-FE-TX
    > > (Fast-ethernet (TX-ISL) Port adapter, 1 port ) card. The interface is
    > > configured for 100Mbps and Full-Duplex as follows:
    > >
    > > interface FastEthernet2/0
    > > ip address 192.168.3.11 255.255.255.0
    > > ip directed-broadcast
    > > ip pim sparse-dense-mode
    > > duplex full
    > > no mop enabled
    > >
    > > The switch port it connects to (on a 6509) is also hard-coded for
    > > 100/FULL
    > >
    > > Yet we are seeing collisions and under-runs:
    > >
    > > CHICR01#sh int fa2/0
    > > FastEthernet2/0 is up, line protocol is up
    > > 36618378 packets output, 3198461469 bytes, 206 underruns
    > > 206 output errors, 206 collisions, 0 interface resets
    > >
    > > Cisco TAC is indicating that "Regardless whether etherent is in full
    > > duplex or not, ethernet by nature - CSMA/CD (carrier sense multiple
    > > access with collision detection) still will check to see if the wire
    > > is busy."
    > >
    > > I always thought that the collision detection mechanism in Ethernet
    > > was disabled automatically when you configure a port/interface for
    > > Full-Duplex. Am
    > > I mistaken?
    > >
    > > TIA
    > > Regards,
    > > Jim


    Basic stuff fellas - Full duplex does not have collision detection. If
    you're seeing them it is because of some other problem - faulty wire
    or connector or port or even as suggested too long a cable or the
    wrong kind of cable. It's probably a combination. You don't state how
    much of a problem this if for you. If it isn't much of one then I
    would change the cable to a brand new one of the correct type (cat 5),
    clear the counters and then wait. See what happens then.
    mikester, Dec 18, 2003
    #5
  6. Jim

    reshman Guest

    I think your TAC guy is smokin' something -- ask him to share! You don't
    have collisions in Full Duplex.

    Now, my input on your problem: I have the same exact issue on a Pix --
    everything shows full duplex, but I am registering collisions under high
    traffic load. I believe there is something wrong with the interface as the
    switch port shows errors (but not collisions!) as well.

    If there is a bad cable or switch port, I'd expect to see errors, but not
    collisions. The fact that you are seeing collisions leaves me with the gut
    feel
    that there is something wrong with that interface.

    I'd suggest having the interface swapped out -- sounds to me like it might
    be flakey.

    Let us know what you find out.

    -Mike

    "Jim" <> wrote in message
    news:...
    > Hello all! Was wondering if someone can clarify something:
    >
    > We have a 7206VXR router running Version 12.2(17)IOS with a PA-FE-TX
    > (Fast-ethernet (TX-ISL) Port adapter, 1 port ) card. The interface is
    > configured for 100Mbps and Full-Duplex as follows:
    >
    > interface FastEthernet2/0
    > ip address 192.168.3.11 255.255.255.0
    > ip directed-broadcast
    > ip pim sparse-dense-mode
    > duplex full
    > no mop enabled
    >
    > The switch port it connects to (on a 6509) is also hard-coded for
    > 100/FULL
    >
    > Yet we are seeing collisions and under-runs:
    >
    > CHICR01#sh int fa2/0
    > FastEthernet2/0 is up, line protocol is up
    > 36618378 packets output, 3198461469 bytes, 206 underruns
    > 206 output errors, 206 collisions, 0 interface resets
    >
    > Cisco TAC is indicating that "Regardless whether etherent is in full
    > duplex or not, ethernet by nature - CSMA/CD (carrier sense multiple
    > access with collision detection) still will check to see if the wire
    > is busy."
    >
    > I always thought that the collision detection mechanism in Ethernet
    > was disabled automatically when you configure a port/interface for
    > Full-Duplex. Am
    > I mistaken?
    >
    > TIA
    > Regards,
    > Jim
    reshman, Dec 18, 2003
    #6
  7. Jim

    Jim Guest

    We're going to Fluke the Ethernet cable end-to-end tomorrow night
    after the market closes. On a side note, the switch port does show FCS
    errors but no collisions. I'll keep you guys posted. Thanks again to
    all of those that have responded.

    "reshman" <marjunk*NO-SPAM*@charter.net> wrote in message news:<>...

    > Let us know what you find out.
    >
    > -Mike
    >
    Jim, Dec 18, 2003
    #7
  8. In article <>,
    Jim <> wrote:
    :We're going to Fluke the Ethernet cable end-to-end tomorrow night
    :after the market closes.

    Excellent idea.

    Fluke network equipment can be pretty expensive to purchase,
    and it might end up sitting on the shelf a lot, but it sure is *useful*
    when you have a problem.
    --
    This signature intentionally left... Oh, darn!
    Walter Roberson, Dec 18, 2003
    #8
  9. Jim

    Jim Guest

    Well folks, it ended up NOT being the cable. The interface in slot 2
    of the 7200 (Fa2/0) was apparently sharing the same backplane bus as
    the interface in slot 1 (Fa1/0). As such, during times of heavy
    network traffic, the bus was overwhelmed causing the errors. Moving
    the module in slot 2 to slot 3 fixed the problem.

    -cnrc.gc.ca (Walter Roberson) wrote in message
    >
    > Excellent idea.
    >
    > Fluke network equipment can be pretty expensive to purchase,
    > and it might end up sitting on the shelf a lot, but it sure is *useful*
    > when you have a problem.
    Jim, Dec 23, 2003
    #9
  10. Jim

    mikester Guest

    (Jim) wrote in message news:<>...
    > Well folks, it ended up NOT being the cable. The interface in slot 2
    > of the 7200 (Fa2/0) was apparently sharing the same backplane bus as
    > the interface in slot 1 (Fa1/0). As such, during times of heavy
    > network traffic, the bus was overwhelmed causing the errors. Moving
    > the module in slot 2 to slot 3 fixed the problem.
    >
    > -cnrc.gc.ca (Walter Roberson) wrote in message
    > >
    > > Excellent idea.
    > >
    > > Fluke network equipment can be pretty expensive to purchase,
    > > and it might end up sitting on the shelf a lot, but it sure is *useful*
    > > when you have a problem.


    Thanks for providing a conclusion.

    =-)
    mikester, Jan 5, 2004
    #10
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Jay B

    Cisco 2900 Full-duplex

    Jay B, Dec 2, 2003, in forum: Cisco
    Replies:
    1
    Views:
    626
  2. Sam
    Replies:
    2
    Views:
    6,977
    Dan Lanciani
    Jan 5, 2004
  3. chuck

    full-duplex on a 804

    chuck, Apr 5, 2004, in forum: Cisco
    Replies:
    4
    Views:
    475
    AnyBody43
    Apr 6, 2004
  4. Guy DeStefano

    Setting WS-2950-24 to full duplex

    Guy DeStefano, Dec 11, 2004, in forum: Cisco
    Replies:
    5
    Views:
    2,017
    Walter Roberson
    Dec 11, 2004
  5. DigitalSierra

    ECP Half-duplex or Full-Duplex?

    DigitalSierra, Oct 18, 2004, in forum: A+ Certification
    Replies:
    0
    Views:
    631
    DigitalSierra
    Oct 18, 2004
Loading...

Share This Page