Very long gc pause caused by ES

Elastic Stack Elasticsearch
1 
      [2015-09-11 11:49:22,429][WARN ][monitor.jvm              ]

[node-173.65] [gc][ParNew][47112][14482] duration [11.3h], collections
[1]/[11.3h], total [11.3h]/[11.4h], memory [4.5gb]->[3.2gb]/[15.7gb],
all_pools {[Code Cache] [12.4mb]->[12.4mb]/[48mb]}{[Par Eden Space]
[1.5gb]->[157.6mb]/[1.5gb]}{[Par Survivor Space]
[67.5mb]->[79.2mb]/[227.5mb]}{[CMS Old Gen]
[2.9gb]->[2.9gb]/[14gb]}{[CMS Perm Gen]
[45.2mb]->[45.2mb]/[128mb]}

JVM parameters:

-Xss256k

-Xmn2048M

-XX:PermSize=128M

-XX:MaxPermSize=128M

-XX:SurvivorRatio=7

-XX:GCTimeRatio=49

-XX:MaxTenuringThreshold=12

-XX:+UseConcMarkSweepGC

-XX:MaxGCPauseMillis=200

-XX:+UnlockExperimentalVMOptions
-Xms:16g

-Xmx:16g
JDK version: 1.6.0.45

2

What do you want to know here, because all you have done is post some text and given us no context as to what you are looking for.

3

any clues why this happens or any suggestions how I can diganose this?

4

Are you really running Java 6?
What ES version are you on?

5

yes, java6. 6u45
es version: 0.90.2

6

Wow, that's really old and you should upgrade ASAP.

How much data in the cluster, how many indices, how many shards, how many nodes.