Hi all, Why does the cleanup.policy of the topic __consumer_offsets
is COMPACT? Is there a specific reason?
Because kafka only needs the latest consumergroup offset per topic-partitions, and it needs it for all topic-partitions, even if the last commit for the topic-partition was a long time ago. the compaction keeps all keys (consumergroup-topic-partition-offset), but only the lates message for each key. this is exactly what kafka requires.
you can find more info here
Thanks, I am asking this because I am using kafka 2.0.0 and there are partitions of the topic __consumer_offsets
that are very big ( 500-700GB ). There are thousands of segments of that partitions that are older than 2-3 months. How is this possible? Thanks
I have already restarted the broker and this happen only for some partitions of that topic __consumer_offsets