The latest consumer changes to read data from Zookeeper during rebalance have made the consumer rebalance code incompatible with older versions (making rolling upgrades without downtime hard). The problem relates to how partitions are ordered. The old code seems to have returned the partitions sorted:

... rebalancing the following partitions: ArrayBuffer(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19) for topic produce-indexable-views with consumers: ...

the new code instead uses:

... rebalancing the following partitions: List(0, 5, 10, 14, 1, 6, 9, 13, 2, 17, 12, 7, 3, 18, 16, 11, 8, 19, 4, 15) for topic produce-indexable-views with consumers: ...

This causes new consumers and old consumers to claim the same partitions. I realize that this may not be a big deal (although painful for us since it disagrees with our deployment automation) since the code wasn't officially released, but it seems simple enough to sort the partitions if you'd take such a patch.

/Sam
 
NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB