Conference 7.286::fddi

1597.0. "Ultrix 4.4, problems with DEFTA controller" by CSC32::B_GRUBBS () Wed Feb 22 1995 12:39

    
    We've got two customers now that seem to be experiencing some kind
    of funny business with FCS check, frame length, and frame alignment
    errors on Ultrix R4000 decstation 5000's with DEFTA controllers.
    In both cases there are OSF DEC3000 alphas on the same ring with
    the DEFTA that NEVER report any of the same errors.
    
    A typical scenario is this:
    
            +---------+ fta - .8db loopback fiber loss
            | andeker |-----------\
            +---------+            \
                                    \
            +---------+ fta - 1.2db  \
            | point   |-----------\   \
            +---------+            \   \    +-----------------+
                                    \   \---| DEC concentator |
            +---------+ fta - 1.8db  \------| 500             |
            | allmalt |---------------------|                 |
            +---------+              -------| ring purger off |
                                    /   ----|                 |
            +---------+ fta - .9db /   /    +-----------------+
            | blatz   |-----------/   /
            +---------+              /
                                    /
            +---------+ fta - 1.5db/
            | miller  |-----------/
            +---------+
    
    In this example ALLMALT is an OSF machine which shows NONE of the
    errors found on the other four machines (all ulrix 4.4).  If you
    look at the heaviest loaded ultrix machine, MILLER, you'd see this in
    it's netstat fta counters.
    
    fta0 FDDI counters at Wed Feb 22 09:32:39 1995
    
           65535 seconds since last zeroed
       122233632 ANSI MAC frames count
               0 ANSI MAC frame errors
               0 ANSI MAC frames lost
      4294967295 bytes received
      4294967295 bytes sent
        46105749 data blocks received
        57434696 data blocks sent
        35664897 multicast bytes received
          223004 multicast blocks received
        15285032 multicast bytes sent
           56613 multicast blocks sent
               0 transmit underrun errors
               0 send failures
              17 FCS check failures
               0 frame status errors
              64 frame alignment errors
            1683 frame length errors
               0 unrecognized frames
               0 unrecognized multicast frames
               0 receive data overruns
               0 system buffers unavailable
               0 user buffers unavailable
               0 ring reinitialization received
               0 ring reinitialization initiated
               0 ring beacon process initiated
               0 duplicate tokens detected
               0 duplicate address test failures
               0 ring purge errors
               0 bridge strip errors
               0 traces initiated
               0 traces received
               0 LEM reject count
               0 LEM events count
               0 LCT reject count
               0 TNE expired reject count
               1 Completed Connection count
               0 Elasticity Buffer Errors
    
    These errors above are after 15 days of uptime.  The customers have applied
    the fta patches for ultrix that let the controller reinitialize, but
    it's during the reinit, caused by these errors, that they get a
    momentary loss of network connectivity.  This can wreak havoc with 
    a bunch of students doing things on nfs mounted drives.
    
    any ideas?  I find it strange that the osf box reports none of these
    errors and it's the main nfs server.  The ultrix boxes are mostly
    clients, with MILLER serving some things via nfs.
    
    any help is appreciated...
    
    --Bert Grubbs
      Colorado CSC