[prev in list] [next in list] [prev in thread] [next in thread] 

List:       solr-dev
Subject:    Re: Shenandoah GC?
From:       Robert Muir <rcmuir () gmail ! com>
Date:       2018-07-31 21:39:51
Message-ID: CAOdYfZUtroV80y=9=ZWuc7XtvKFgvdK3t_2g1uBavN1=x3e8uA () mail ! gmail ! com
[Download RAW message or body]

I think the best approach is that: when its deemed ready and in the
JDK mainline and available in EA releases, we look into adding the
appropriate GC switches to the jenkins randomization.
Until that time, if you want to build the thing yourself and help them
out by reporting tests found by Lucene bugs, that's great.

But keep in mind that our jenkins fails every so often with compiler
bugs from real release JDKs, so it may be better to focus on
stabilizing what is actually supposed to be stable for production. I
responded about a few of those failures in the last month myself.
Honestly I think my replies get ignored because everyone is filtering
out jenkins failures due to the noise of solr tests.


On Tue, Jul 31, 2018 at 2:56 PM, Mike Drob <mdrob@apache.org> wrote:
> Thanks, Robert.
>
> I agree with your concerns about correctness; saying "workloads" was prob=
ably inaccurate on my part.
>
> Approximately how long ago do you think you tried running the test suite =
against the shenandoah collector? Were there bugs reported against the JDK =
tracker (I looked but couldn't find any instances)?
>
> I'm still trying to evaluate if it is worth my time to even _attempt_ to =
verify correctness here, or if there is enough consensus from the community=
 that it should wait another 6-12 months.
>
> Mike
>
> On 2018/07/31 14:56:19, Robert Muir <rcmuir@gmail.com> wrote:
>> I think your biggest concern needs to be correctness. try first
>> getting the test suite to pass before worrying about any workloads.
>> The last time i tested one of the shenandoah builds, basically all
>> tests related to points failed.
>>
>> On Tue, Jul 31, 2018 at 10:49 AM, Mike Drob <mdrob@apache.org> wrote:
>> > Hi Devs,
>> >
>> > Has anybody tried running Lucene or Solr workloads with the Shenandoah=
 Garbage Collector[1]? I searched through the list archives and saw somebod=
y reported issues with it back in 2014, but I imagine that there's been a l=
ot of change since then.
>> >
>> > We've had some initially hopeful results over on HBase, but I know tha=
t when there are any sorts of GC issues that Lucene usually surfaces them f=
airly quickly. Wanted to see if there was already a trail blazed down this =
road before investing too much further.
>> >
>> > Thanks,
>> > Mike
>> >
>> > [1]: https://wiki.openjdk.java.net/display/shenandoah/Main
>> >
>> > ---------------------------------------------------------------------
>> > To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
>> > For additional commands, e-mail: dev-help@lucene.apache.org
>> >
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: dev-help@lucene.apache.org
>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org

[prev in list] [next in list] [prev in thread] [next in thread]