Feasibility to backport new allocation event and event throttling to JDK 8,11
Erik Gahlin
erik.gahlin at oracle.com
Wed Dec 16 13:44:12 UTC 2020
Hi Mukilesh,
The Object Allocation Sample event is completely new and before considering a backport, it should be thoroughly stress tested. If it turns out to be issues, we may have to pull it out of JDK 16, or at least disable the event by default.
Then we should probably wait 6 months after JDK 16 has been released, so we can make adjustments to the configuration or sampling mechanism, after feedback from users. Then it can be decided, if it should be backported.
In my opinion, it shouldn’t. Having different type of events in different update releases, or distributions, creates confusion. Users don’t want things to fail due to a security upgrade. Those who really want allocation sampling in an old release can turn on the TLAB events. There will be perhaps 1% additional overhead (YMMV), but that price is only paid by those who really want the feature.
Everybody else can enjoy the stability.
Thanks
Erik
> On 11 Dec 2020, at 23:52, Mukilesh Sethu <mukilesh.sethu at gmail.com> wrote:
>
> Hey,
>
> I could see new allocation event and event throttling was added recently(
> https://github.com/openjdk/jdk/pull/1624). I was wondering if we could
> backport these changes to JDK 8,11. I am just checking if there are any
> concerns in backporting these changes to JDK 8,11 as it introduces new
> event and new event functionality (throttling). From the initial look I
> think backport should be trivial, please let me know what you think ?
>
> Thanks,
> Mukilesh
More information about the hotspot-jfr-dev
mailing list