Allow disabling of PSI_*_* async metrics collection

[MikhailBurdukov]

Changelog category (leave one):

• Improvement

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

Allow disabling of PSI_*_* async metrics collection

Documentation entry for user-facing changes

• Documentation is written (mandatory for new features)

Details

We run ClickHouse in a pretty specific containerization infrastructure and we have faced issues where /proc/pressure/* files actually exist:

 ~ # ls -la /proc/pressure/
total 0
dr-xr-xr-x    5 root root 0 Oct  2 13:00 .
dr-xr-xr-x 3055 root root 0 Oct  2 13:00 ..
-r--r--r--    1 root root 0 Oct  2 13:00 cpu
-r--r--r--    1 root root 0 Oct  2 13:00 io
-r--r--r--    1 root root 0 Oct  2 13:00 memory

But it is impossible to read from them:

~ # cat /proc/pressure/io
cat: /proc/pressure/io: Operation not supported

And it causes a lot of spam in ClickHouse logs with exceptions when ClickHouse reads from these files:

2025.10.15 09:41:53.005752 [ 10180 ] {}  ReadBuffer: ReadBuffer is canceled by the exception: Code: 74. DB::ErrnoException: Cannot read from file /proc/pressure/cpu: , errno: 95, strerror: Operation not supported. (CANNOT_READ_FROM_FILE_DESCRIPTOR), Stack trace (when copying this message, always include the lines below):

0. DB::Exception::Exception(DB::Exception::MessageMasked&&, int, bool) @ 0x00000000133ac8df
1. DB::Exception::Exception(String&&, int, String, bool) @ 0x000000000c8579ce
2. DB::Exception::Exception(PreformattedMessage&&, int) @ 0x000000000c857480
3. DB::ErrnoException::ErrnoException(int, int, FormatStringHelperImpl::type>, String&&) @ 0x0000000013489d00
4. void DB::ErrnoException::throwFromPath(int, String const&, FormatStringHelperImpl::type>, String&&) @ 0x00000000134884c4
5. DB::ReadBufferFromFileDescriptor::readImpl(char*, unsigned long, unsigned long, unsigned long) const @ 0x000000001349c10f
6. DB::ReadBufferFromFileDescriptor::nextImpl() @ 0x000000001349c1a0
7. DB::ReadBuffer::next() @ 0x000000001349a2ad
8. DB::readPressureFile(std::unordered_map, std::equal_to, std::allocator>>&, String const&, DB::ReadBufferFromFilePRead&, std::unordered_map, std::equal_to, std::allocator>>&, bool) @ 0x00000000135b0074
9. DB::AsynchronousMetrics::update(std::chrono::time_point>>, bool) @ 0x00000000135a0b39
10. void std::__function::__policy_invoker::__call_impl[abi:ne190107]::ThreadFromGlobalPoolImpl(DB::AsynchronousMetrics::start()::$_0&&)::'lambda'(), void ()>>(std::__function::__policy_storage const*) @ 0x00000000135b5f3a
11. ThreadPoolImpl::ThreadFromThreadPool::worker() @ 0x00000000135077d2
12. void* std::__thread_proxy[abi:ne190107]>, void (ThreadPoolImpl::ThreadFromThreadPool::*)(), ThreadPoolImpl::ThreadFromThreadPool*>>(void*) @ 0x000000001350f29a
13. ? @ 0x0000000000094ac3
14. ? @ 0x00000000001268c0
 (version 25.8.8.26 (official build))
2025.10.15 09:41:53.005796 [ 10180 ] {}  void DB::AsynchronousMetrics::update(TimePoint, bool): Code: 74. DB::ErrnoException: Cannot read from file /proc/pressure/cpu: , errno: 95, strerror: Operation not supported. (CANNOT_READ_FROM_FILE_DESCRIPTOR), Stack trace (when copying this message, always include the lines below):

0. DB::Exception::Exception(DB::Exception::MessageMasked&&, int, bool) @ 0x00000000133ac8df
1. DB::Exception::Exception(String&&, int, String, bool) @ 0x000000000c8579ce
2. DB::Exception::Exception(PreformattedMessage&&, int) @ 0x000000000c857480
3. DB::ErrnoException::ErrnoException(int, int, FormatStringHelperImpl::type>, String&&) @ 0x0000000013489d00
4. void DB::ErrnoException::throwFromPath(int, String const&, FormatStringHelperImpl::type>, String&&) @ 0x00000000134884c4
5. DB::ReadBufferFromFileDescriptor::readImpl(char*, unsigned long, unsigned long, unsigned long) const @ 0x000000001349c10f
6. DB::ReadBufferFromFileDescriptor::nextImpl() @ 0x000000001349c1a0
7. DB::ReadBuffer::next() @ 0x000000001349a2ad
8. DB::readPressureFile(std::unordered_map, std::equal_to, std::allocator>>&, String const&, DB::ReadBufferFromFilePRead&, std::unordered_map, std::equal_to, std::allocator>>&, bool) @ 0x00000000135b0074
9. DB::AsynchronousMetrics::update(std::chrono::time_point>>, bool) @ 0x00000000135a0b39
10. void std::__function::__policy_invoker::__call_impl[abi:ne190107]::ThreadFromGlobalPoolImpl(DB::AsynchronousMetrics::start()::$_0&&)::'lambda'(), void ()>>(std::__function::__policy_storage const*) @ 0x00000000135b5f3a
11. ThreadPoolImpl::ThreadFromThreadPool::worker() @ 0x00000000135077d2
12. void* std::__thread_proxy[abi:ne190107]>, void (ThreadPoolImpl::ThreadFromThreadPool::*)(), ThreadPoolImpl::ThreadFromThreadPool*>>(void*) @ 0x000000001350f29a
13. ? @ 0x0000000000094ac3
14. ? @ 0x00000000001268c0
 (version 25.8.8.26 (official build))

This PR adds a setting to disable collection of such metrics.

[clickhouse-gh]

Workflow [PR], commit [284ae23]

Summary: ❌

job_name	test_name	status	info	comment
Stateless tests (amd_tsan, parallel, 2/2)		failure
	01443_merge_truncate_long	FAIL	cidb, issue	ISSUE CREATED
BuzzHouse (amd_debug)		failure
	Logical error: 'Inconsistent AST formatting: the query:	FAIL	cidb, issue	ISSUE EXISTS
BuzzHouse (amd_ubsan)		failure
	Let op!	ERROR	cidb, issue	ISSUE CREATED

[alesapin]

@MikhailBurdukov the only thing is that it's not a bug fix, but improvement.

[MikhailBurdukov]

@alesapin
Could you put can be tested label, pls?

Fixed changelog category, but can we try to backport the pr anyway, because the issue is pretty painful for our clusters? 🙏

[MikhailBurdukov]

@alesapin ping

[MikhailBurdukov]

@alesapin Could you put can be tested pls?

[MikhailBurdukov]

@alesapin
Can we merge the PR?

[MikhailBurdukov]

@alesapin The PR is improvement but can we try to backport to 25.8 anyway?