Logstash JVM GC "Allocation Failure"

[ebuildy (Thomas Decaux)]

I run Logstash 2.2 with head size of 8G:

export LS_HEAP_SIZE=8G

exec setuser logstash logstash agent
--pipeline-workers 4
--pipeline-batch-size 2500
--pipeline-batch-delay 10
-f /etc/service/logstash/logstash.conf

ps -aux gives:

http://b3.ms/P5Z9ZQxDMlbr

Saying memory is 11Mo. But, running "jstat -gccause 98 2000", it complains:

"Allocation Failure"

Is it normal? (also I always get high CPU as soon as I send few events via any input plugins, tried with TCP, UDP, RabbitMQ).

Thanks,

[warkolm (Mark Walkom)]

Why so high? Unless you are pushing massive messages through LS, you don't really need more than 4GB.

[ebuildy (Thomas Decaux)]

I don't know as I see "Allocation Failure" Java error (JVM debug), I though more heap could help.

I try to reduce the CPU logstash is consuming (very high compared to Fluentd, for almost same pipeline TCP => filter => elasticsearch).

Is it something normal because the jRuby?

Thanks,