Recovery in an IPIC HA environment

A connection into a cluster may fail, leaving units of work (UOWs) that need to be resynchronised at a later point.

When a client region, with outstanding units of work, tries to reconnect to an HA cluster then an attempt is made to restore the connection to the specific region in the cluster that these UOWs relate to. If successful then resynchronisation processing takes place to complete these UOWs. A client region can suffer multiple connection failures, leaving more than one set of UOWs that require resynchronising. If the client region is unable to connect to a cluster region associated with one set of outstanding UOWs, then it will look to see if there is another set of UOWs that need processing and attempt to connect to their associated cluster region. This sequence of events continues until a connection is established with a specific server region, or until there are no more sets of UOWs found, at which point the client region will then attempt to connect to the HA cluster's generic end point.

The client region's IPCONN XLNACTION attribute will be used to decide how any outstanding UOWs are dealt with once a connection becomes acquired. This attribute defaults to KEEP, meaning that any UOWs that could not be resynchronised are retained. If XLNACTION is set to FORCE then those UOWs that could not be resynchronised at that time are forced to complete heuristically, according to the decision indicated by the ACTION attribute of their associated TRANSACTION resource definition.

A client region that has no outstanding UOWs will always connect to the HA cluster's generic end point.



dfht1_ipicha_recovery.html | Timestamp icon Last updated: Sunday, 4 June 2017