hsperfdata causing long GC/safepoint times: Don't use mmap?

[Vitaly Davidovich]

Could this also be the stable pages effect at play? I'm not familiar with
the mechanics of hsperfdata logging, but if it's writing to same pages
repeatedly, it could be that the page is scheduled for writeback and
subsequent writes to it will block until the writeback completes.

On Wed, Apr 1, 2015 at 9:22 AM, Florian Weimer <fweimer at redhat.com> wrote:

> On 03/26/2015 04:14 PM, Evan Jones wrote:
> > At Twitter, I recently discovered that the hsperfdata file that is
> created
> > by default by the JVM causes long safepoint and GC pauses. It turns out
> > that writes to mmap-ed files can block until disk I/O completes, even if
> > the I/O is to another disk. For details see:
> > http://www.evanjones.ca/jvm-mmap-pause.html
>
> Can you try to get a kernel backtrace when this happens?  Preferably on
> ext4 or XFS.
>
> Scheduling a page for writeback probably requires a kernel memory
> allocation, so this behavior isn't too surprising.
>
> --
> Florian Weimer / Red Hat Product Security
>