All committers have signed the CLA.

Tested by creating a simple docker image:

// a.cpp
#include <iostream>
#include "getNumberOfPhysicalCPUCores.h"
int main()
{
    std::cout << "thread: " << getNumberOfPhysicalCPUCores() << std::endl;
    return 0;
}
// g++ a.cpp

> cat Dockerfile
FROM centos
COPY a.out /a.out
CMD ["a.out"]

> docker build -t='jayson/test_cpu' .

>  docker run --rm --cpus="2" jayson/test_cpu ./a.out
thread: 2
>  docker run --rm --cpus="4" jayson/test_cpu ./a.out
thread: 4
>  docker run --rm --cpus="8" jayson/test_cpu ./a.out
thread: 8
>  docker run --rm --cpus="40" jayson/test_cpu ./a.out
thread: 40
>  docker run --rm --cpus="80" jayson/test_cpu ./a.out
docker: Error response from daemon: Range of CPUs is from 0.01 to 40.00, as there are only 40 CPUs available.
See 'docker run --help'.
> docker run --rm --cpus="8.6" jayson/test_cpu ./a.out
thread: 9

Tested by creating a simple docker image:

By the way, this can be tested using integration test, something similar to https://github.com/ClickHouse/ClickHouse/blob/master/tests/integration/test_jemalloc_percpu_arena/test.py

Tested by creating a simple docker image:

By the way, this can be tested using integration test, something similar to https://github.com/ClickHouse/ClickHouse/blob/master/tests/integration/test_jemalloc_percpu_arena/test.py

I'll try to add some tests later this week

How does this work with cgroupv2 only hierarchies?

@bobrik How does this work with cgroupv2 only hierarchies?

After doing some research, this PR doesn't work under cgroup v2.
However, now both CPU and Memory don't work under cgroup v2, and v2 is not enabled by default as far as I know. Maybe we can open another issue for cgroup v2 limits?

Reference:

• Control Group v2: https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html

I think it's ok to just add v1 and have a separate issue for v2.

"Stress test (debug, actions) — OOM killer (or signal 9) in clickhouse-server.log"

I'm not sure why this test is failed.

@Mergifyio update

update

✅ Branch has been successfully updated

@Mergifyio update

update

❌ Base branch update has failed

I think this PR is ready to be reviewed. Is there anything I need to do?

I am waiting for CI and going to merge it 👍

Is any plans to backport this to 21.8?

@thyn No, as it is not a bugfix.
The only effect is that queries are run slightly less efficient and consume slightly more memory.