Conference netcad::hub_mgnt

1744.0. "DECbridge 900MX and spanning tree problem" by KETJE::VANDENBERG_R () Thu Dec 01 1994 13:31

	Hi,

	How can we explain this.

	1. Customer configuration

                     BACKBONE
                  []------+-------------------+---------------+--------[]
                          |                   |               |
                         STE's        +------+|               |
                                STE's |DEMPR2||           +---+--+    
                                   |  +------+|           |DEMPR3| 
		                 +-+----+ |   |           +-+-+--+
                                 |DEMPR1| |   |             | |
                                 +----+-+ |   |             | STE's
                                      |   |   |             |
                        STE  STE STE  |   |   |             |
                          |   |   |   |   |   |Backup       |
                        +-2---3---4---5---6---7-+           |
		     +--| DECbridge900MX 1      |--+        |
                     | A|                       |B |        |Loop
                     |  +-----------------------+  |        |
                     |                             |FDDI    |
                     |                             |        |
                     | B+-----------------------+A |        |
                     +--| DECbridge900MX 2      |--+        |
                        |         Root          |           |
                        +-2---3---4---5---6---7-+           |
                          |   |   |           |             |
	                 VAX VAX VAX          +-------------+

		  STEs = Cluster's Satelite


	2. The questions
		
	By mistake the customer made a loop between the 2 bridges through
	a DEMPR.
	Until there no problem, line 7 of bridge 1 is in backup.
	All the Cluster statelites on the Bridges and on the DEMPR's (Cluster)
	are running fine.

	Now they' removed the connector of bridge 1 port 7. From this moment
        ALL the stations of the clusters lost their connections (Cluster state 
        transition) (SYSGEN; RECNXINTERVAL = 20 sec)
		   
        I can undertsand that VAXes on the Backbone and on DEMPR3 can't access
        the server during the the pre-forwarding and listening phase of the 
	bridge port (longer than 20 secs), but WHY ALL OTHERS VAXES connected
        on the bridge 1 also were loosing the connection.

	The only way possible for me is that a Topology change on ONE port of
        a Bridge causes a complete spanning tree configuration.
	Is this possible or normal?

	What I need to understand is how to see (calculate) the spanning tree
        algorithm with a Multi-port Bridge, i.e, should I see such a bridge;
        as separate 7 bridges connected to a virtual segment or 7 separate      
        bridges connected together (some white paper or doc available?).

	Thanks in advance,
	Robert
	


			    
				

		   

    
    The algorithm doesn't change with the number of ports.  
    
    If the link that you disconnected was Blocking (was the backup link),
    then you shouldn't have seen any break in connectivity.  If, however,
    you have the root wrong then what was disconnected was the designated
    port for that LAN, and you would experience the delay waiting for the
    backup port to take over.
    
    As for why the cluster members all saw a problem, you should check
    with Cluster gurus; is it not true that all nodes maintain the state of
    connections to all other nodes in the cluster?
    
    You can see the spanning tree configuration with HUBwatch, or any SNMP
    manager using RFC1493 - the Bridge MIB.
    

	Scott,
	
	OK, I 'checked with Cluster guru and I may be possible that other nodes 
of the cluster are reseting if there is a group inconsistancy.
	
	Thanks,