Only leaders report the message-in-rate metrics. For balancing the leaders, take a look at https://cwiki.apache.org/confluence/display/KAFKA/FAQ#FAQ-Whypartitionleadersmigratethemselvessometimes ?
Thanks, Jun On Wed, May 28, 2014 at 1:05 AM, Arjun <ar...@socialtwist.com> wrote: > Hi, > > I just checked the list topic. There is data in the kafka logs directory. > My set up as said has 3 brokers, with replication factor 2. > One thing i see in the output of the list topic is, no partition has this > node as a leader. Is this the reason for bean not showing up anything? > > Thanks > Arjun Narasimha kota > > > > On Monday 26 May 2014 06:09 PM, Devanshu Srivastava wrote: > >> Hi Arjun , >> >> What is the configured number of logical partitions per topic per server . >> With the partition size as 2 , Only two partitions would be created >> distributed over 2 brokers with one partition each , you can confirm the >> same by checking the log file directory for each broker (default - >> /tmp/kafka-logs). Probably the 3d broker does not have any Topic-partition >> for data to be written/read. >> >> You may also use tool : bin/kafka-list-topic.sh –zookeeper >> localhost:2181 –topic >> {topic-name} >> >> >> On Mon, May 26, 2014 at 4:02 PM, Arjun <ar...@socialtwist.com> wrote: >> >> Hi, >>> >>> We have a Kafka cluster which contains 3 kafka brokers. Data is pushed >>> into kafka queue continously. >>> To monitor the kafka brokers, we use the JMX metrics given by kafka. >>> One thing i observed is there is one broker which is returning*"0"* all >>> the time for the bean >>> *"kafka.server":type="BrokerTopicMetrics",name=" >>> AllTopicsMessagesInPerSec" >>> . >>> * >>> Is there some thing i should worry about? i am able to get the data from >>> kafka and push the data into kafka with out a glitch. >>> >>> >>> Thanks >>> Arjun Narasimha Kota >>> * >>> * >>> >>> >>> >