Kafka uses Zookeeper to store metadata information about topics, partitions, brokers and system coordination (such as membership statuses). Unavailability or slowness of Zookeeper makes the Kafka cluster unstable, and Kafka brokers do not automatically recover from it. Cloudera recommends to use a 3-5 machines Zookeeper ensemble solely dedicated to Kafka (co-location of applications can cause unwanted service disturbances).

• zookeeper.session.timeout.ms is a setting for Kafka that specifies how long Zookeeper shall wait for heartbeat messages before it considers the client (the Kafka broker) unavailable. If this happens, metadata information about partition leadership owned by the broker will be reassigned. If this setting is too high, then it might take a long time for the system to detect a broker failure. On the other hand, if it is set to too small, it might result in frequent leadership reassignments.
• jute.maxbuffer is a crucial Java system property for both Kafka and Zookeeper. It controls the maximum size of the data a znode can contain. The default value, one megabyte, might be increased for certain production use cases.
• There are cases where Zookeeper can require more connections. In those cases, it is recommended to increase the maxClientCnxns parameter in Zookeeper.
• Note that old Kafka consumers store consumer offset commits in Zookeeper (deprecated). It is recommended to use new consumers that store offsets in internal Kafka topics (reduces load on Zookeeper).