[jira] [Created] (KAFKA-4150) Facing Message Loss When trying rolling updates

["Avanish Pandey (JIRA)" <ji...@apache.org>]

Avanish Pandey created KAFKA-4150:
-------------------------------------

             Summary: Facing Message Loss When trying rolling updates
                 Key: KAFKA-4150
                 URL: https://issues.apache.org/jira/browse/KAFKA-4150
             Project: Kafka
          Issue Type: Bug
    Affects Versions: 0.9.0.1
            Reporter: Avanish Pandey


Hi ,

I am trying to see the settings that I can keep for my kafka cluster.

The settings I'am trying are:
replication-factor = 3
min.insync.replicas = 2
num of brokers = 3

Scenario:
Bring broker1 down.
Bring it back up.
Bring broker 2 down.
Bring it back up.
Bring broker 3 down.
Bring it back up.

There is a delay of at least 1-2 minutes in between the brokers being stopped.

Other configs:
zookeeper.session.timeout.ms=10000
num.partitions=6
min.insync.replicas=2
log.retention.ms=172800000
message.max.bytes=1000000
log.flush.interval.ms=5000
log.flush.offset.checkpoint.interval.ms=5000
log.roll.hours=24
zookeeper.session.timeout.ms=1000
leader.imbalance.check.interval.seconds=60
unclean.leader.election.enable=false


What I am seeing is variable number of messages lost. Where as , when i don't bring kafka down, or just bring 1 broker down and back up, everything is fine.

Could you let me know if there is some setting which is wrong or needs correction.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)