[jdk8u-dev] RFR: 8230305: Cgroups v2: Container awareness

[Jonathan Dowland]

On Mon, 3 Oct 2022 10:05:05 GMT, Jonathan Dowland <jdowland at openjdk.org> wrote:

>> This is a backport of 8230305 (Cgroups v2: Container awareness) for JDK8u, working from the jdk11u-dev backport as part of an effort to backport cgroups v2 to jdk8u-dev.
>> 
>> The patch does not apply clean after unshuffling and path fixing:
>> 
>>  * mostly copyright line issues
>>  * different package name for jdk.test.lib.process.OutputAnalyzer
>>  * Most of hotspot/src/os/linux/vm/cgroupSubsystem_linux.cpp failed due to context changes, manually resolved (mostly deletions)
>> 
>> A few other changes were made
>> 
>>  *  Rework use of newer logging API (log_debug etc) to use guarded tty->print_cr
>>  *  replace os::strerror with the libc strerror (which is thread unsafe, but that should be addressed by backporting 8148425 separately)
>> 
>> patch touched hotspot/test/runtime/containers/docker/TestCPUAwareness.java which passes afterwards; I'm running further tests now (and awaiting GitHub CI doing the same)
>
> Local test failure observed in runtime/containers/docker/TestMemoryAwareness.java:
> 
> JavaTest Message: Test threw exception: java.lang.RuntimeException: 'Memory Soft Limit.*524288000' missing from stdout/stderr 
> 
> this might be [JDK-8253714](https://bugs.openjdk.org/browse/JDK-8253714), which is on the list to backport, I'll look at this next.

> @jmtd Any particular reason this PR depends on the Metrics PR? In JDK 11u and JDK head this was the first one of the cg v2 patch series. If you agree, please drop the dependency, targeting master directly.

I collapsed the patch dependency tree into a linear list, and as a consequence some patches have GitHub PR dependencies which are artificial. This one (depending on  8231111 (pr/121)) is an example; it's because the downstream bug 8253714 (pr/130) depends on both.

I could flip the artifical dependency around, and make 8231111 (metrics, pr/121) depend on this (8230305, pr/127) instead (and follow that change down the chain). Would you prefer that?

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

PR: https://git.openjdk.org/jdk8u-dev/pull/127