PI17596: Reconnecting containers might immediately stop healthy containers during container reconnect.

APAR status

• Closed as program error.

Error description

• Reconnecting containers can abruptly stop healthy containers
  during container reconnect.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:  All WebSphere eXtreme Scale users running   *
  *                  embedded container and catalog servers in   *
  *                  WebSphere Application Server.               *
  ****************************************************************
  * PROBLEM DESCRIPTION: Containers on the primary catalog       *
  *                      server side of a brownout are           *
  *                      mistakenly being directed to perform    *
  *                      container reconnect.                    *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  During container reconnect, stable containers can be torn down.
  For instance, during a brownout, the containers on the side
  with the primary catalog server should persist and not be torn
  down. However, those servers are torn down and reconnected
  after the island containers have finished reconnecting. This
  is due to a timing window that occurs when the island
  containers reconnect to the catalog server and have coalesced
  into their own group, but before they have coalesced into a
  group with the non-island servers. The island servers are able
  to send a heartbeat or view that does not contain the
  non-island servers to the catalog server. The catalog server
  mistakenly allowed the heartbeat to force the non-island
  servers to reconnect, since they were not yet in the island
  server view.
  

Problem conclusion

• The code was updated to remove the potential timing window.
  

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

  WS EXTREME SCAL

• Fixed component ID

  5724X6702

Applicable component levels

• R711 PSY

     UP

• R850 PSY

     UP

• R860 PSY

     UP