RFR: JDK-8255450: runtime/ThreadCountLimit.java causes high system load

[Thomas Stuefe]

On Tue, 27 Oct 2020 13:40:36 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> Hi,
>> 
>> this is rather trivial:
>> 
>> runtime/ThreadCountLimit.java, introduced with JDK-8222671, caused problems on our test machines (JDK-8222671 is private and cannot be accessed from outside Oracle, so all I know comes from its review thread [1]).
>> 
>> The test creates massive amount of threads in order to hit some OS limit which would manifest as an OOM. This affects unrelated processes, unless the test is executed in a jail or with specific limits set.
>> 
>> This test probably should be executed with a specific limit, but for now lets mark it as @stress and remove it from tier1.
>
> This makes sense to me. On my TR 3970X with Linux x86_64 release, it peaks at 70G VIRT, 3G RSS, and fails half of the time. The test itself should probably made more resilient, but even then it does not look like `tier1`-grade test.

Following the trivial rule I will integrate this patch.

-------------

PR: https://git.openjdk.java.net/jdk/pull/876