Hello - we have a single Solr 5.2.1 node that (for now) contains four single shard collections. Only two collections actually contain data and are queried. The machine has some unusual latency that led me to sample the CPU time with VisualVM. On that node we see that DistributedQueue$LatchWachter.await() and OverseerAutoReplicaFailoverThread() claiming a lot of CPU time, DistributedQueue takes the most of it. I have set logging to DEBUG but no Zookeeper logging is printed, the connections are stable and updates rarely come in.
Is it because the node runs the overseer? If so, how can i prevent it from taking quite a lot of CPU time on a node with four collections where no state changes occur? Any other thoughts? Markus
