This variable is intended for supporting suboptimal internal deployments. It controls how many replicas the partitions in chatter have in Kafka, see training-arch-op-kafka-transientchatter-events for more information.

The default is 3, which works well when deploying to 1 or 3 or more data centers. Because internally CS only has access to two data centers, the default setting of 3 would need to be adjusted, as data services would rather set them to 1. With two datacenters, there is no way to get a perfectly stable, automatically recoverable deployment. Data services are accepting this and will monitor these systems to be able to manually recover on short notice in case something goes wrong.