Clients need restart after broker

[Ashutosh singh <ge...@gmail.com>]

Greetings All,

Appreciate if someone can help to address the issue.

Every time  when we restart any of the broker some of the clients are
getting stuck for hours and do not produce or consume messages.  we have 8
node cluster and replication factor is set 2 and 3 for few topics.  We are
in process of changing it to 3.

I know that this is the issue with the clients but how to find what is the
issue with clients ? what settings to look for ?

There are 3 clients which is being used -

kafka-python  - python
kafka-node - node
sarama client - go

For example -  below messages are keep on repeating until unless I restart
my application.

GO -
[Sarama] 2019/11/19 22:19:12 kafka: Failed to produce message to topic
hotels_promo_logs:write tcp 172.17.0.3:54188->10.30.20.17:9092: write:
broken pipe


node -

2019-11-19 22:21:43 [KAFKA_SERVICE][INFO] Kafka message sent failed for
Dispatch: NC750694137205078 for topic: FLEET_OWNER_ACCEPT_DETAILS with
params: {"dispatch_request_id":"5dd36d7468fb13002e0e2bff"}| Error Messages:
Broker not available (loadMetadataForTopics), Broker not available
(loadMetadataForTopics), Broker not available (loadMetadataForTopics)


Thanks
Ashu