Conference 7.286::fddi

935.0. "massive failover problems with DECbridge 620" by VNABRW::HAINZL_R () Tue Apr 20 1993 14:33

Problem:
        on a customer site with DECbridge 620 and redundant
        connected ethernet segments we experienced serious failover
        problems.

Problems experienced as well with DECbridge620 Firmware V1.2A  and V1.3

Configuration:

        !==========!                    !==========!
        ! VAX7610  !                    ! VAX7610  !
        !==========!                    !==========!
             !                               !
             !   DAS                         ! DAS
             !                               !
        !==========!                    !==========!
        !  DEFCN   !====================!  DEFCN   !
        !==========!                    !==========!
             !!                              !!
             !!                              !!
             !! f on FDDI-LINE               !! b on FDDI-LINE
        !==========!                    !==========!
        !  DEFEB1  !====================!  DEFEB2  !
        !==========!                    !==========!
          !2 !3 !4                        !2 !3 !4
          !f !f !f                        !b !f !b
      ----o-------------------------------o----------- Segm.A
             !  !                            !  !
             !  !                            !  !
      -------o-------------------------------o-------- Segm.B (Root Bridge of
                !                               !              ST on Segment B)
                !                               !
      ----------o-------------------------------o----- Segm.C

f... LINE forwarding
b... LINE in backup state

There a lot of bridges (ca. 40 , not all from Digital, e.g. Vitalink etc)
in one spanning tree domain.

All is working fine in steady state, performance ok).
All started with V1.3 loaded onto the bridges.

Then we did some failover tests:

- disconnecting the transceiver cable from the forwarding line 4 of DEFEB1
  causes  a failover to line 4 on DEFEB2 (shown by the LEDS);
  so far an expected behaviour.
  BUT : DECnet as well as LAT connections from and to this segment
        were extremly slow and timed out regularly... the performance was
        a real mess.
        And not enough ... if we tried to touch the forwarding table on
        DEFEB2 by MCC
          MCC> SHOW BRIDGE DEFEB2 FORWARDING TABLE DYNAMIC ENTRIES * ALL
        the Bridge DEFEB2 reinitialized.

        The above mentioned behaviour was reproducable :
                reconnecting line 4 on DEFEB1 ... all ok
                disconnecting line 4 on DEFEB1 ... all worse.

- at this time we decided ( in best DEC manner) to change all modules of
  DEFEB2.
  Since the new modules obviously had a better address (in  respect of
  spanning tree), the DEFEB2 lines became forwarding after startup.
  New situation:


        !==========!                    !==========!
        ! VAX7610  !                    ! VAX7610  !
        !==========!                    !==========!
             !                               !
             !   DAS                         ! DAS
             !                               !
        !==========!                    !==========!
        !  DEFCN   !====================!  DEFCN   !
        !==========!                    !==========!
             !!                              !!
             !!                              !!
             !! b on FDDI-LINE               !! f on FDDI-LINE
        !==========!                    !==========!
        !  DEFEB1  !====================!  DEFEB2  !
        !==========!                    !==========!
          !2 !3 !4                        !2 !3 !4
          !b !f !b                        !f !f !f
      ----o-------------------------------o----------- Segm.A
             !  !                            !  !
             !  !                            !  !
      -------o-------------------------------o-------- Segm.B (Root Bridge of
                !                               !              ST on Segment B)
                !                               !
      ----------o-------------------------------o----- Segm.C

f... LINE forwarding
b... LINE in backup state


and all the things were vice versa :

        - disconnecting line 4 on DEFEB2
        - failover to line 4 on DEFEB1

        - but massive performance problems connecting from and to
          Segment 4
        - touching DEFEB1 with MCC --> bridge reinitializing.

        - reproducable also with failover tests on the other segments.

strange :

        - it seemed that there's some correletion between FDDI-Forwarding and
          performance behaviour.

        - if we alse got the FDDI-LINE to failover to the "failover-bridge"
          (by disconnecting FDDI) all was fine again.

the above mentioned behaviour remained after a downgrade to V1.2A
        - only the MCC-problem seemed to be fixed after the downgrade.


-----------------------------------------------------------------------

since this is a major customer in Austria (just look at the number of
bridges), and they bought the Bridges for faiover reasons, urgent help is
needed.

The problem is also escalated by DS.

any suggestions ???, rgds Richard