Troubles with Shenandoah

[Roman Kennke]

I also see it with latest jdk/jdk build.

Also, I notice that latencies go through the roof. Usally in the us
range, we suddenly get up to multi-seconds (!!) range.

     @                _  1.394.016 µs (207208, 2,09%)
                 @    _  2.788.033 µs (710359, 7,16%)
                   @  _  4.182.050 µs (853475, 8,60%)
                 @    _  5.576.067 µs (735210, 7,41%)
                 @    _  6.970.084 µs (732427, 7,38%)
                @     _  8.364.101 µs (694916, 7,00%)
              @       _  9.758.118 µs (593103, 5,98%)
                 @    _  11.152.134 µs (715708, 7,21%) ^50%
                  @   _  12.546.151 µs (769548, 7,76%)
                 @    _  13.940.168 µs (739466, 7,45%)
                   @  _  15.334.185 µs (854111, 8,61%)
                 @    _  16.728.202 µs (739780, 7,46%)
              @       _  18.122.219 µs (584904, 5,90%) ^85%


WTF is going on here?!

Roman


> Hi,
> 
> On Mon, Apr 8, 2019 at 11:13 AM Roman Kennke <rkennke at redhat.com>
> wrote:
> > I'm still failing to reproduce the problem. I am running with those
> > settings:
> > 
> >           <executable>/home/rkennke/progs/jdk-
> > 12+33/bin/java</executable>
> >           <arguments>
> >             <argument>-showversion</argument>
> >             <argument>-Xmx16g</argument>
> >             <argument>-Xms2g</argument>
> >             <argument>-Xlog:gc:stderr:time,level,tags</argument>
> >             <argument>-XX:+PrintCommandLineFlags</argument>
> >             <argument>-XX:+UnlockExperimentalVMOptions</argument>
> >             <argument>-XX:+UseShenandoahGC</argument>
> > 
> > 
> > and I don't get lost messages or anything like that.
> > 
> > I am having troubles with the benchmark though: if I understand it
> > correctly, the server and client are asking some parameters on
> > startup.
> > However, for some reason, I don't see the messages printed out, but
> > apparently it is waiting for input. So I keep pressing ENTER a
> > couple
> > of times, blindly.
> 
> Please see my previous message about this: it's a Maven library bug
> and I explain the workaround.
> 
> For short runs such as ~10s when you press all enters to the
> benchmark, I also don't see any problem.
> They start to appear for longer runs.
> Try to set batch size = 10000 (i.e. 10 times the default). This will
> make the benchmark run for ~100s and for me this was enough to get
> the
> failures.
> 
> > The questions start appearing then. (Missing
> > terminal flushes/syncs?) Also, the client appears to be waiting for
> > something between test runs. Is there a way to automate this? I.e.
> > pass
> > the parameters to the pom.xml and make the client not wait between
> > runs?
> 
> You can't make automatic back to back runs, but you can tune the
> parameters to make a run longer, see above.
>