INC-220174 · Issue 716384
Improved cleanup for data from joining nodes
Resolved in Pega Version 8.6.5
A Cassandra node that is down for longer than the grace period (default 10 days) can introduce zombie data that creates instability when it returns to the cluster. This can include VBD partition summary data that was deleted and can break loading rules and cause the service to fail to start up. To resolve this, additional logic has been implemented to detect zombie summary records, including summary records without field descriptors and summary records without dictionaries having already been provided, and to read dictionaries from the latest summary record in case there are preceding zombie records with dictionaries.