Troubles with Shenandoah

[Aleksey Shipilev]

On 4/8/19 10:30 PM, Simone Bordet wrote:
> On Mon, Apr 8, 2019 at 10:01 PM Aleksey Shipilev <shade at redhat.com> wrote:
>> You're not likely to get the max latency improvements either, as generational STW GCs would be able
>> to hit the similar 1..3ms pauses consistently.
> 
> Well, G1 is in the 15-25 ms range, while ZGC is < 2 ms.

Aha, cool. G1 performs better for me here, at about 5ms; but that's probably I don't run with huge
heaps and on moderate load.

> Did you run with the default benchmark configuration?
> If so, can you try to increase the "batch count" to 10_000 or more?
> Shenandoah was ok for me for short runs, but not for longer runs.

This was with batch-count=10K. I did run several iterations of it, spanning around an hour (lots of
[Enters] pressed), and there are no lost messages. It used to be every 5-th run ending with lost
msgs. What would be a good stress test to run overnight? batch-count=10M?

>> The whole exercise was the source of interesting bugs, we already fixed a few, and other fixes are
>> in pipeline.
> 
> So what exact version can I use to run this benchmark with Shenandoah?
> The latest 13? Should I wait for the fixes in the pipeline?
> I can build Shenandoah from the sources.

I'd say for evaluation use, you can continue with jdk12u binary and always-committed-heap, unless
you experience another problem. If you are feeling adventurous, then build from jdk/jdk source. That
is JDK development dumping ground^W^W mainline, which is *really* bleeding edge and might have
surprising bugs. Otherwise, the patches should land in jdk12u within a week, as our CIs clear them
for backporting, and we fix other problems too.

-- 
Thanks,
-Aleksey