IJ08822: COMMIT_LVL TO HANDLE MIS-MATCHED CAA AND POWERHA LEVELS

APAR status

• Closed as fixed if next.

Error description

• COMMIT_LVL is a process used by CAA to ensure that
  all nodes support the same set of capabilities. All
  nodes must agree on the proposed level. Once that
  coordinated configuration change is complete, the
  instances of CAA on all nodes can be confident of the
  capabilities that are supported across the cluster.
  
  The COMMIT_LVL operation can fail if an older version
  of PowerHA is running on a newer version of CAA: the
  older version of PowerHA may not be able to support
  the facilities provided by the newer version of
  PowerHA, and fail the request.  This can happen if
  the customer has upgraded AIX before upgrading
  PowerHA.  This can prevent a CAA cluster - and hence,
  a PowerHA cluster - from being formed.  While running
  latest PowerHA on latest CAA (AIX) will always work,
  support for node at a time migration requires that
  the cluster at least function with mixed levels, even
  if it runs with lowest level of capabilities
  supported cluster wide.
  
  To allow this, two facilities are added to CAA:
  1) The "clctr -commit" command is enhanced to support
     an explicit level.  With no level given, it will
     continue to attempt to commit to the local node
     maximum capabilities.  With an explicit level, it
     will attempt to commit to that explicit level.
  2) The "auto-commit" run when the last node to join
     the cluster performs a JOIN_NODE should respond to
     a failure in the COMMIT_LVL by trying again at the
     next lower level.
  

Local fix

• Upgrading all nodes to latest PowerHA, and then rebooting
  them will frequently resolve this problem.
  An explicit "clctrl -commit" will sometimes allow the
  nodes
  to successfully form a cluster.  However, some
  combinations
  of PowerHA and CAA will not accept this.
  

Problem summary

• CAA always tries to commit to latest CAA level (set of
  capabilities).  When it fails, there is no mechanism to commit
  to an intermediate CAA level (one that exists between the
  current effective level and maximum CAA level).
  

Problem conclusion

Temporary fix

Comments

• This APAR is being closed FIN. This means that a solution to
  this APAR is expected to be delivered from IBM in a release
  (if any) to be available within the next 24 months.
  

APAR Information

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

  IJ07833

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

Fix information

Applicable component levels