I am running Kafka locally in my MAC machine with single node and 3
brokers, each running in different port.
I was trying the Kafka PoC with having multiple brokers up and running and
sending messages to Kafka and in between failing some nodes (leader nodes).
There are some observations as below:
- huge latency in getting acknowledgment back from Kafka once any leader
broker is down. Also, the latency is observed for every consecutive call
even after bringing back failed brokers. The average latency was ~10 sec.
on my local setup.
- And also, it is observed that whenever a leader broker fails, Kafka
does not automatically allocate a new leader, it only allocates when a call
to Kafka is made, which add up into further latency.
Can you provide some inputs on the above observations?