What is Replication Batch Interval setting and when can it be modified?
Configuration replication works by combining change events into a "batch", which is the unit of replication to subscribers. The minimum size of a batch is determined by the Replication batch interval parameter. The current default value for Replication batch interval is 5 seconds, which means if there is a constant stream of configuration changes, there is one replication batch created approximately every 5 seconds.
When a subscriber is geographically separated from the publisher by a high latency WAN, the time required to apply a batch of changes (Config replication) on to subscriber is influenced heavily by network latency. This can lead to cases where a subscriber goes out of sync periodically, when there are active configuration changes on the publisher and the subscriber using default 5 seconds (batch size) to pull in a batch of configuration changes.
From 6.5.0, we have a new option as a cluster-wide parameter called "Replication Batch Interval" (value range 1-60 seconds), which defines the minimum replication batch interval. Increasing this value should help in deployments where subscriber is across a high latency WAN from the publisher.
There are downsides to increasing the value of this parameter, mainly the apparent time delay before a configuration change made on the publisher get replicated to all the subscribers. For example, the delay before a new Guest account provisioned can authenticate successfully via RADIUS through subscriber. We recommend that you modify this parameter only when instructed by technical support.