The pruning windows of the Replica and Client nodes are independent. Therefore, the Replica or Client node pruning can be configured for different pruning windows.
Replica Node Pruning
An operator initiates the Replica node pruning from the reconfiguration tool. Pruning starts only after the consensus of N out of N Replica node is reached. This consensus is required to maintain data integrity across the Replica Network. The operator can prune by the number of blocks or by setting the days to prune during the defined pruning window. The default pruning time frame is two weeks. When the pruning begins, any stale key-values data saved in the local persistent storage outside the set pruning window are pruned.
Pruning can only be performed if the following criteria are met:
Data has been safely replicated on the Replica, Client, and Full Copy Client nodes.
Data is outside of the pruning window.
Data is earmarked as obsolete.
Pruning progress can be monitored using relevant metrics such as pruned keys per second.
Client Node Pruning
For the Client node, pruning can be initiated to reduce the Index DB size. A System Operator configures the pruning window using a predefined offset.
For additional pruning configuration information, see https://docs.daml.com/ops/index.html.