This way when kafka_controllers is updated the pillar value gets updated and any non-controllers get updated to revert to 'broker' only role.
Needs more testing when a new controller joins in this manner Kafka errors due to cluster metadata being out of sync. One solution is to remove /nsm/kafka/data/__cluster_metadata-0/quorum-state and restart cluster. Alternative is working with Kafka cli tools to inform cluster of new voter, likely best option but requires a wrapper script of some sort to be created for updating cluster in-place.
Easiest option is to have all receivers join grid and then configure Kafka with specific controllers via SOC UI prior to enabling Kafka. This way Kafka cluster comes up in the desired configuration with no need for immediately modifying cluster
Signed-off-by: reyesj2 <94730068+reyesj2@users.noreply.github.com>
reyesj2
m0duspwnens