Search code examples
javatomcatgarbage-collectionnewrelic

New Relic JVM graphs on Tomcat

I am currently trying to optimise my Tomcat on a 64bit sever with 4 GB of memory. Here are my current settings and its ticking over nicely:

-Xms256m
-Xmx1024m
-XX:MaxPermSize=256m
-XX:+UseParNewGC
-XX:+UseConcMarkSweepGC
-XX:+CMSClassUnloadingEnabled

What I don't quite understand is how they relate to the New Relic charts - is the Eden space heap usage optimal - what should I be aiming on seeing in the reports? Relic Charts

Solution

  • The information in the charts don't exactly have a correlation to the parameters specified fro the JVM at startup because the information provided is a trend. Most, if not all, of the metrics used for the charts for the JVM are averages over the period of time specified in the picker.

    To help explain, I'll share the metrics used for each chart:

    Heap Memory Usage (MB)

    • Memory/Heap/Committed with value_fn: average_value
    • Memory/Heap/Max with value_fn: average_value
    • Memory/Heap/Used with value_fn: average_value

    PS Eden Space heap usage (MB)

    • MemoryPool/Heap/PS Eden Space/Committed with value_fn: average_value
    • MemoryPool/Heap/PS Eden Space/Max with value_fn: average_value
    • MemoryPool/Heap/PS Eden Space/Used with value_fn: average_value

    PS Survivor Space heap usage (MB)

    • MemoryPool/Heap/PS Survivor Space/Committed with value_fn: average_value
    • MemoryPool/Heap/PS Survivor Space/Max with value_fn: average_value -MemoryPool/Heap/PS Survivor Space/Used with value_fn: average_value

    PS Old Gen heap usage (MB)

    • MemoryPool/Heap/PS Old Gen/Committed with value_fn: average_value
    • MemoryPool/Heap/PS Old Gen/Max with value_fn: average_value
    • MemoryPool/Heap/PS Old Gen/Used with value_fn: average_value

    Non-heap memory pool usage (MB)

    • MemoryPool/Non-Heap/Code Cache/Used with value_fn: average_value
    • MemoryPool/Non-Heap/PS Perm Gen/Used with value_fn: average_value

    Garbage collection CPU time

    • GC/PS Scavenge with value_fn: average_call_time

    Class count

    • JmxBuiltIn/Classes/Loaded with value_fn: average_value
    • JmxBuiltIn/Classes/Unloaded with value_fn: average_value

    Does this help explain the information in the charts sufficiently? Let us know if you need further assistance with this issue.