Logstash error stack

I am using logstash 2.2.4, elasticsearch 4.5, filebeat 1.1.1

Beats input: unhandled exception {:exception=># Zlib::BufError: buffer error >, :backtrace=>["org/jruby/ext/zlib/ZStream.java:134:in finish'", "org/jruby/ext/zlib/JZlibInflate.java:72:ininflate'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/lumberjack/beats/server.rb:380:in compressed_payload'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/lumberjack/beats/server.rb:251:infeed'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/lumberjack/beats/server.rb:462:in read_socket'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/lumberjack/beats/server.rb:442:inrun'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/logstash/inputs/beats_support/connection_handler.rb:33:in accept'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/logstash/inputs/beats.rb:211:inhandle_new_connection'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/logstash/inputs/beats_support/circuit_breaker.rb:42:in execute'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/logstash/inputs/beats.rb:211:inhandle_new_connection'", "/opt/logstash/vendor/bundle/jruby/1.9/gems/logstash-input-beats-2.2.7/lib/logstash/inputs/beats.rb:167:in `run'"], :level=>:error}

Can anybody interpret this error stack for me?

According to my diagnoses, my logstash nodes needs more cpu power.
The logstash process delayed when cpu is not enough , buffer filled up and broke the pipeline, that's why the error is showing.
In newrelic monitor, the moving average cpu consumption was 80% when this error shows, it might reach 100% actually.

.

This is the newrelic histogram showing a node with not enough CPU. Network I/O is zigzag, looks like it needs to wait for cpu to finish processing a batch, then flush the output to the network

.

This is the newrelic histogram showing a node with enough CPU. The Network I/O is smooth.

Is it correct interpretation?