9.7.2 Guidelines for selecting the GC
The processing that is to be specified using the pre-statistical GC selection functionality differs depending on the purpose of investigating the class-wise statistical information to be output.
The guidelines for selecting the processing depending on the purpose of investigation is described here.
Specify the processing that can be selected using the pre-statistical GC selection functionality by the argument of the jheapprof command. Depending on the object that is to be investigated or depending on the class-wise statistical information used in the type of investigation, you can select the processing.
The following table lists the guidelines for selecting the processing.
|
Processing (argument of the jheapprof command) |
Investigation target |
Example of how to investigate the class-wise statistical information |
|---|---|---|
|
Execution of a Full GC (-fullgc) |
Changing the object by Full GC |
You can identify the object that causes the memory leak of the Java heap, since the used object is collected by Full GC. |
|
Execution of a copy GC (-copygc) |
Changing the object by copy GC |
With the copy GC, you can identify the object with a long life that migrates to the Tenured area. From this information, you can identify the object that causes increase in the occurrence count of the Full GC. |
|
No execution of any GC (-nogc)# |
Object with short life is being collected by executing GC |
Information on all objects including the used object is output. You can identify the object with a high memory sharing that becomes the cause (super-large object) when the GC collection occurs frequently. |
- #
-
When -nogc is specified in the argument of the jheapprof command, the information on all short-lived objects is output, and therefore, the log output volume increases.