Conference noted::decnis

3553.0. "X.25 DA-circuits: "thunderstorm"" by UTOPIE::BRAUN_R () Fri Feb 21 1997 07:12

Hi all,

we had (or still having) a strange problem with DECNIS and X.25 DA-circuits

The config:



                VAX, VAX    (total about 50, 25 per NIS connected via X.25
                             DA-circuits with DECnet)

                  
                          . . . . . 

                        - - - - - - - - 
                - -  |                  | - 

               |    internal X.25 network   |
                    of the Austrian PTT
               |                         - - 
                - -  |                  |
                        - - - - - - - - 
                        |           |
            
                      NIS 600     NIS 600
                       3.1.8       3.1.8



All the VAXes have DECnet-OSI 6.3.5(6) and an internal DSV-11 controller.

Additionally, the DECNIS are acting as X.25-gateways, too.


During the last weeks, we had strange and intermittent problems:
Suddenly, one of the DECNIS stopped DECnet routing via the X.25-
DA-circuits (X25-gateway (e.g. X.29) still worked o.k). 
Only via a disable and enable of the X25-DA-circuit on the DECNIS
it was possible to clear up the situation. 
After a while, the customer was able to identify a particular
VAX as cause:  when defining this VAX on the other DECNIS
(enabling the X.25 reach addr, disabling on the original DECNIS,
the problem moved to the other DECNIS).
So far, so bad, but not only the particular VAX was in trouble, but
also ALL OTHER reachable addresses on the concerned DECNIS.

In the OPCOM we saw lots of adjacency state changes (up and down)
on probaby ALL reachable addresses, the DECNIS seemed to trash
X.25 call requests to all destinations. At nearly the same time,
lots of illegal packets arrived, some of them being logged to 
OPCOM.

We are not sure, if the problem now is being fixed after having
swapped the DSV-11 in the concerned VAX, but the customer
cannot understand the following:


***********

1.) Why does it affect ALL destinations and not the particular
    one only (if it really was due to the illegal packets)

    This seems to be a DECNIS-bug in every case.

***********

2.) According to our X.25-knowledge, an illegal packet from
    one side (illegal for any reason: CRC, X.25 protocol state
    violation, ...) should NEVER pass the X.25-network, causing
    problems on the other side (DECNIS).

    But this could be due also to a poor implementation of the
    used X.25-switches, not really being CCITT-compliant regarding
    protocol checking, .... 

 

Any help will be highly appreciated.

Thanks,
Ralph