RFR: After GC reporting should also report metaspace progress

[Zhengyu Gu]

On 09/26/2018 06:08 AM, Aleksey Shipilev wrote> On 09/25/2018 11:27 PM, 
Zhengyu Gu wrote:
>> Following upstream discussion, Shenandoah should follow other GCs to report metaspace progress after
>> GCs.
>>
>> I also changed logging tag for heap from (gc, ergo) to (gc, heap) to be consistent with others.
>>
>> Webrev: http://cr.openjdk.java.net/~zgu/shenandoah/metaspace_stats/webrev.00/
> 
> I am not OK with changing the tag to (gc, heap). Our adopters use those logging blocks to debug
> ergonomics decisions, and it should stay (gc, ergo).
> 
> The logging seems to be better called from ShenandoahControlThread::run_service, that actually
> drives the cycle. There is "if (gc_requested) ..." block that handles whatever actions that should
> happen after the GC. Put the field tracking "last" metaspace usage in ShHeap, and call into ShHeap
> from ShControlThread.

Other than full and degenerated GC, we really do not report progress of 
whole GC cycle, so there is no point to report metaspace progress.

Therefore, I withdraw this RFR.

Thanks,

-Zhengyu

> 
> -Aleksey
>