Hi, all
Recently, I upgrade my Kafka cluster to 0.8.1.1 and set replication with
num.replica.fetchers=5. Last night, there's something wrong with the
network. Soon, I found the server.log files (not data log!) on every node
reached 4GB in an hour.
I am not sure if it's my inappropriate
This looks very similar to the error and stacktrace I see when
reproducing https://issues.apache.org/jira/browse/KAFKA-1196 -- that's
an overflow where the data returned in a FetchResponse exceeds 2GB. (It
triggers the error you're seeing because FetchResponse's size overflows
to become negative,
