We have Kafka clusters in three datacenters with Mirror Maker replicating across three data centers. We are monitoring JMX metrics of our clusters. I have question regarding FetchRequestPurgatory (kafka.server:type=FetchRequestPurgatory,name=PurgatorySize), description on Wiki says "Number of requests waitingin the fetch purgatory. This is high if consumers use a large value for fetch.wait.max.ms " Most of our brokers seem to have very high value for FetchRequestPurgatory ranging from 10k to just 2. My understanding is once the Fetch Request is fulfilled, the value should go down but why the value is so high on some nodes whereas some nodes keep-up? Naidu Saladi
