PH14925: SPLIT BRAIN OCCURRED ON BIGSQL USING AUTOMATIC FAILOVER DUE TO TEMPORARY DISCONNECTION

APAR status

• Closed as fixed if next.

Error description

• In a Big SQL cluster with HA and automatic failover enabled,
  if the entire primary head node disconnects temporarily from
  the rest of the cluster, resulting in the interruption of both
  Zookeeper and Big SQL HA connections, the standby may initiate
  a takeover by force.  If the primary head node then reconnects
  and is still online, the result can be a split brain scenario.
  .
  This vulnerability can be mitigated by introducing a small delay
  before the standby initiates a takeover by force, to ensure that
  the primary head node is actually down.
  

Local fix

• A split brain situation is an undesirable but not-unexpected
  possibility when the standby head node takes over while the
  primary is disconnected.  If the reliability of a cluster
  may trigger frequent takeovers, the best course of
  action to avoid split brain situations will be to disable
  failover automation (with the downside being that the head node
  goes down in an actual failure until manual intervention
  occurs).
  

Problem summary

• Please see problem discription.
  

Problem conclusion

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

Applicable component levels