ITEM: FB2401L

Hard takeover fails in a 3 nodes cluster. incorrect serial definitions

ENV:            9076 4.2.1 hacmp 4.2.2

CUSTOMER REP:   Valerie Scott & Jack Anderson

PROBLEM: doing hard failures back & forth in hacmp doesn't recognise
        failed clusters went down. graceful failover ok. poweroff
        won't failover. no numeric error code. Hacmp.out didn't
        see anything. clstat is yellow.
        This is on nodes 5, 6 & 7. 7 still shows interfaces up 
        which is impossible. node 5 is the backup that 6 & 7 
        feeds into. customer is running a lotus notes ap.
                                
        The customer found out what they think is the problem.

        They had cabled the serial connections like this.

                              5
                          tty2 tty1
                           /     \\
                          /       \\
                         /         \\
                      tty1        tty2 
                      6 tty2- - -tty1 7

        According to the diagram the serial networks should be
        serial1 5tty1 - 7tty2
        serial2 7tty1 - 6tty2
        serial3 6tty1 - 5tty2

        However when the customer defined the networks in hacmp
        they were typed in as:
        serial1 Atty2 - Ctty1
        serial2 Ctty2 - Btty1
        serial3 Btty2 - Atty1

        So when system 6 when down, and 5 should takeover, all of the 
        tcpip networks showed down for 6, but the serial2 network
        still showed up because it was getting its heartbeat packets
        from node7.

Action taken: Talked with customer and redefining her serial networks
        worked fine.

Action Plan: close



Support Line: Hard takeover fails in a 3 nodes cluster. incorrect serial definitions ITEM: FB2401L

Dated: December 1997 Category: N/A

This HTML file was generated 99/06/24~13:30:15

Comments or suggestions?
Contact us