One of the scenarios which Kafka users will come across often is the restart of Kafka servers in a cluster to deploy new changes with minimal down time. When chef-client is run in batch on all the nodes (scheduled as daemons), there is a possibility that many kafka servers are brought down at the same time due to restart for config changes resulting in availability/performance issue. The worst case scenario would be all but one Kafka server is brought down.
Another way to alleviate this scenario will be to control the number of servers which can be restarted at a time which can vary by users based on how many replication they set for the topic partitions. Can we provide some feature to be able to control the restart of the particular Kafka node. One way would be to

• conditionally include a new recipe file instead of the "service" resource in " _configure.rb
• The default "new recipe" file will have a dummy recipe which will get executed on any condition that currently triggers the Kafka restart.
• When the dummy recipe gets executed, it notifies the restart "kafka service" resource to perform the restart.

This will help users to override the default "new" recipe file with their own recipe file and logic to control whether the Kafka server can be restarted or not. Let me know if you have any questions/thoughts.