IC72534: SDS NODES IN SDS CLUSTER WILL BE IN INCORRECT STATUS AFTER NETWORK WAS BROKEN FOR A WHILE

APAR status

• Closed as program error.

Error description

• There are two database servers in SDS cluster, DB server A and
  DB server B;
  
  DB server A resides on box A, DB server B resides on box B;
  
  There are two CM servers to construct a CM cluster, CM1 reside
  in the same box with DB server A, CM2 reside in the same box
  with DB server B;
  
  At first, DB server A is the primary server in SDS cluster and
  CM1 is the only one CM Arbitrator, DB server B was in SDS
  read-only status, all client connection request will be
  redirected to DB server A;
  
  Sometime, break the network on the box A which DB server A was
  resided in to simulate network outage (like:take off the network
  cable);
  
  After the time specified by EVENT_TIMEOUT and FOC timeout, CM2
  notified the break in cluster and issue the fail over command.
  After that DB server B will be new primary server and CM2 will
  be new CM Arbitrator;
  
  At this time, all new client connection request will be
  redirected to DB server B; All old connections were in stuck
  status;
  
  After some while, network on box A will be recovered. Now DB
  server A and B are all in primary status, CM1 and CM2 are all CM
  Arbitrator; DB servers don't communicated each other again;
  
  New client connection request will be redirected to DB server A
  again, and any data operation can be happened on DB server A,
  but any changes happened on DB server A can not be wrote to
  disk, data was lost. Changes happened on DB server B can be
  remained.
  
  After restart DB server A, it will be switched to SDS secondary.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Users of SDS cluster in MACH11 environment.                  *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * Due to loss of network on the primary server in a SDS        *
  * cluster, the connection manager will attempt to perform a    *
  * failover on the SDS node. After the primary network          *
  * connection is re-established, the SDS cluster can have 2     *
  * primary nodes, each potentially corrupting the shared disks. *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Upgrade to 11.50xC9 when available.                          *
  ****************************************************************
  

Problem conclusion

• Problem fixed in 11.50xC9. See CQ218767 remarks for additional
  information on the fix approach.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  IBM IDS ENTRP E

• Fixed component ID

  5724L2304

Applicable component levels

• RB15 PSY

     UP