PH13808: SPLIT-BRAIN HADR FAILURE MAY OCCUR WITH AUTOMATED FAILOVER

APAR status

• Closed as fixed if next.

Error description

• Split-brain failures occur when both head nodes are active and
  accepting transactions that cannot later be reconciled because
  of different and incompatible log chains. This may happen when
  there is a transaction on the Primary that has not yet been
  copied over to the Standby before a failover occurs,
  particularly if the failover is performed by force.  When HA is
  enabled and Failover Automation is enabled, the Failover
  Controller may issue a "by force" takeover command, which may be
   vulnerable to split-brain HADR failure.
  
  To reduce the likelihood of split-brain, Failover Automation
  will now use "by force peer window only", to allow time for the
  Standby to catch up to the Primary.  In addition, this APAR
  addresses an issue that may have allowed the Primary head node
  to shut down before the Standby can take over, which would cause
  the Standby to use "by force" unnecessarily.
  
  It is not possible to completely eliminate the possibility of
  split-brain; a split-brain failure could still occur if the
  Primary head node goes down unexpectedly and remains down for
  longer than the HADR peer window
  

Local fix

• If ssh latency is an issue in your cluster, you can increase
  the size of the HADR_PEER_WINDOW with the following command:
  db2 update db cfg for bigsql dbpartitionnum 0 using
  hadr_peer_window <new value>
  followed by a restart of the head node.
  .
  The new value specified would depend on the specifics of the
  cluster and latency.
  

Problem summary

• Please see problem description.
  

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