What things acutally done in `Ref Enq` in g1 gc

Question

I found a paragraph of gc log in my online hbase service as follow:

2018-11-08T19:12:35.907+0800: 4069334.381: [GC pause (G1 Evacuation Pause) (young)
Desired survivor size 41943040 bytes, new threshold 1 (max 1)
- age   1:   23290768 bytes,   23290768 total
, 1.9809681 secs]
   [Parallel Time: 22.6 ms, GC Workers: 13]
      [GC Worker Start (ms): Min: 4069334380.9, Avg: 4069334381.1, Max: 4069334381.2, Diff: 0.3]
      [Ext Root Scanning (ms): Min: 0.8, Avg: 1.0, Max: 1.1, Diff: 0.3, Sum: 12.8]
      [Update RS (ms): Min: 1.6, Avg: 1.8, Max: 2.0, Diff: 0.4, Sum: 23.4]
         [Processed Buffers: Min: 1, Avg: 4.4, Max: 9, Diff: 8, Sum: 57]
      [Scan RS (ms): Min: 2.2, Avg: 2.4, Max: 2.5, Diff: 0.3, Sum: 30.9]
      [Code Root Scanning (ms): Min: 0.0, Avg: 0.0, Max: 0.0, Diff: 0.0, Sum: 0.0]
      [Object Copy (ms): Min: 6.7, Avg: 7.5, Max: 17.0, Diff: 10.4, Sum: 97.9]
      [Termination (ms): Min: 0.0, Avg: 9.5, Max: 10.3, Diff: 10.3, Sum: 123.1]
         [Termination Attempts: Min: 1, Avg: 2.5, Max: 5, Diff: 4, Sum: 33]
      [GC Worker Other (ms): Min: 0.0, Avg: 0.0, Max: 0.0, Diff: 0.0, Sum: 0.3]
      [GC Worker Total (ms): Min: 22.1, Avg: 22.2, Max: 22.3, Diff: 0.3, Sum: 288.5]
      [GC Worker End (ms): Min: 4069334403.3, Avg: 4069334403.3, Max: 4069334403.3, Diff: 0.0]
   [Code Root Fixup: 0.0 ms]
   [Code Root Purge: 0.0 ms]
   [Clear CT: 0.4 ms]
   [Other: 1957.9 ms]
      [Choose CSet: 0.0 ms]
      [Ref Proc: 4.2 ms]
      [Ref Enq: 1951.7 ms]
      [Redirty Cards: 1.2 ms]
      [Humongous Register: 0.0 ms]
      [Humongous Reclaim: 0.0 ms]
      [Free CSet: 0.3 ms]
   [Eden: 592.0M(592.0M)->0.0B(592.0M) Survivors: 32.0M->32.0M Heap: 6521.8M(12.2G)->5943.3M(12.2G)]
 [Times: user=0.19 sys=0.11, real=1.98 secs]

As you can see most time spend on Ref Enq. What can I do to optimize this.

score 1 · Answer 1 · answered Nov 09 '18 at 06:41

1

Answer for the title :: Information about the time taken for processing of Reference Objects is shown in the Ref Proc and Ref Enq phases. During the Ref Proc phase, G1 updates the referents of Reference Objects according to the requirements of their particular type. In Ref Enq, G1 enqueues Reference Objects into their respective reference queue if their referents were found dead.

Answer for the optimization :: If these phases take too long, then consider enabling parallelization of these phases by using the option -XX:+ParallelRefProcEnabled

For further info --->

https://docs.oracle.com/javase/10/gctuning/garbage-first-garbage-collector-tuning.htm#JSGCT-GUID-0BB3B742-A985-4D5E-A9C5-433A127FE0F6

answered Nov 09 '18 at 06:41

Sandunka Mihiran

556
4
19

1. `ParallelRefProcEnabled` already added to java startup option. 2. Can you give an example to the clause `In Ref Enq, G1 enqueues Reference Objects into their respective reference queue if their referents were found dead.`, I can not understand the `referents ` in your statement. @Sandunka – ganle hu Nov 09 '18 at 07:00
That's what they suggest. Did you consider hardware acceleration ? – Sandunka Mihiran Nov 09 '18 at 07:02

What things acutally done in `Ref Enq` in g1 gc

1 Answers1