Strange G1 behavior

[Guillaume Lederrey]

On 20 October 2017 at 15:27, Thomas Schatzl <thomas.schatzl at oracle.com>
wrote:

> Hi Guillaume,
>
> On Fri, 2017-10-20 at 15:03 +0200, Guillaume Lederrey wrote:
> > Thanks for your continuing interest in our issue!
>
>   thanks for giving us the opportunity to look at your workload, and
> your time you spend on our requests.
>

Always a pleasure!


> >
> > I have been firefighting another issue with a user sending a bit too
> > much traffic our way. Good news, this allowed us to tune our
> > throttling and will probably result in a slightly smoother load in
> > the future, which can only help...
> >
> > I have prepared a change with your suggested improvements [1]. But I
> > will wait until next Monday to deploy it. I'll send the new logs as
> > soon as I have them.
> >
> > Kirk suggested earlier to double the size of the heap (from 16GB to
> > 32GB). I have not yet implemented that suggestion. Do you think it
> > make sense to bundle that change with the changes suggested by
> > Thomas? Or should I keep it for later?
> >
>
> I would prefer if we did not change too many variables at the same
> time. Heap size has (obviously) a huge impact on behavior, although in
> this case we will always see whether the heap usage spikes/length of
> marking etc.
>
> The only problem is that it seems we won't get to see that user
> generating these loads anymore, so it might be that some of these
> suggestions may be unneeded (and to some degree detrimental to overall
> performance).
>

We saw this unusual load only this week. We already had long GC pauses and
high GC overhead under our "normal" load conditions. So I expect we will
still see the same behaviours.


> But it's your setup :)
>
> Thanks,
>   Thomas
>



-- 
mobile : +41 76 573 32 40
skype : Guillaume.Lederrey
Freenode: gehel

projects :
* https://signs-web.herokuapp.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20171020/d62d36ed/attachment.htm>