timed results in network disaster

[elsen at imecom.imec.be]

   The following is a query in order to find out whether anyone else
   has seen the following happen on ultrix 4.n systems.

   Problem / description : timed running in slave mode on 20
   decstations booting from DS5400 server. 
   Timed also running in slave mode on 2 DECstation 5000

   Master timed runs on a VAX3900 (Ultrix 3.1) : timed exe on this
   systems is the Ultrix 4.1 version.


   
   What happens : sometimes all stations seem to receive an invalid
   time from the timed server (this only happens about every 2 weeks).
   This causes all systems to hang since apparently automounter dumps core
   when the system time is being set to a value which corresponds
   to nonsens , see below 
   
********************************* ENTRY     2. *********************************

----- EVENT INFORMATION -----

EVENT CLASS                             OPERATIONAL EVENT
OS EVENT TYPE                  310.     TIME CHANGE
SEQUENCE NUMBER                  6.
OPERATING SYSTEM                        ULTRIX 32
OCCURRED/LOGGED ON                      Thu Oct 19 01:45:43 1911 MET DST
                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
OCCURRED ON SYSTEM                      isis
SYSTEM ID                 x0A000005     MICROCODE REV = 5.
SYSTYPE REG.              x01530102
                                        FIRMWARE REV = 83.
PROCESSOR TYPE                          KA650

********************************* ENTRY     3. *********************************
(this entry in fact belongs to a vaxstation , but the same can be
seen on the decstations)



   Further more when this incident occurs we can see that files in 
   /dev are corrupt for each and every decstation booting from
   the ds5400 server.
   Special device files involved are mainly pty devices but sometimes
   also "mem" and "kmem".
   After remaking those devices with MAKEDEV the station itself
   can operate correctly again.



    Question : has anyone else seen problems with timed  of
               this nature ?
    Remark   : this incident has occurred twice on us when each of
               the DECstation 5000's were being installed for the
               first time.
               Would there be any reason for the timed protocol to
               fail when a stations boots into the network for
               the first time ?

    Is it "bad" I  mean "real bad" to run the ultrix041 timed 
    on  a 3900 (vax) running ultrix 3.1 ?
    Due to time constraints it was not yet possible to upgrade
    this machine.



                               Thanks for any help ,

                                                    Marc (elsen at imec.be)