cgroup support for physical_processor_count #1035
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
eregon
reviewed
Jan 22, 2024
Comment on lines
39
to
40
| if Dir.exist?("/sys/fs/cgroup/cpu,cpuacct") && (cfs_quota_us = IO.read("/sys/fs/cgroup/cpu,cpuacct/cpu.cfs_quota_us").to_i) > 0 | ||
| (cfs_quota_us / IO.read("/sys/fs/cgroup/cpu,cpuacct/cpu.cfs_period_us").to_i.to_f).ceil |
There was a problem hiding this comment.
Is there some documentation about these cgroups files?
It'd be nice to add a link as a code comment here.
Comment on lines
39
to
40
| if Dir.exist?("/sys/fs/cgroup/cpu,cpuacct") && (cfs_quota_us = IO.read("/sys/fs/cgroup/cpu,cpuacct/cpu.cfs_quota_us").to_i) > 0 | ||
| (cfs_quota_us / IO.read("/sys/fs/cgroup/cpu,cpuacct/cpu.cfs_period_us").to_i.to_f).ceil |
There was a problem hiding this comment.
.to_i.to_f can probably be just .to_f
| elsif ln.start_with?("core") | ||
| cid = phy + ":" + ln[/\d+/] | ||
| cores[cid] = true if not cores[cid] | ||
| if Dir.exist?("/sys/fs/cgroup/cpu,cpuacct") && (cfs_quota_us = IO.read("/sys/fs/cgroup/cpu,cpuacct/cpu.cfs_quota_us").to_i) > 0 |
There was a problem hiding this comment.
Could you use File.read instead of IO.read? It seems clearer (same on the next line).
|
It works as expected inside a container limited to 3 cores running in a host machine that has over 10: |
byroot
reviewed
Jan 29, 2024
| cid = phy + ":" + ln[/\d+/] | ||
| cores[cid] = true if not cores[cid] | ||
| # https://kernel.googlesource.com/pub/scm/linux/kernel/git/glommer/memcg/+/cpu_stat/Documentation/cgroups/cpu.txt | ||
| if Dir.exist?("/sys/fs/cgroup/cpu,cpuacct") && (cfs_quota_us = File.read("/sys/fs/cgroup/cpu,cpuacct/cpu.cfs_quota_us").to_i) > 0 |
There was a problem hiding this comment.
IMO we shouldn't change the behavior of physical_processor_count. We should instead have a cpu_quota method that returns a float, and then eventually a usable_processor_count which returns cpu_quota * processor_count.
This way it's up to the caller to decide if they want to floor/round/ceil
There was a problem hiding this comment.
Also I think what you implemented here is cgroups V1? Not sure how much it's used these days, in cgroups V2 the info is in /sys/fs/cgroup/cpu.max.
casperisfine
pushed a commit
to casperisfine/concurrent-ruby
that referenced
this pull request
Jan 29, 2024
Closes: ruby-concurrency#1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
|
I took the liberty to push this a bit further in #1038 with cgroups v2 support etc. |
casperisfine
pushed a commit
to casperisfine/concurrent-ruby
that referenced
this pull request
Jan 29, 2024
Closes: ruby-concurrency#1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
casperisfine
pushed a commit
to casperisfine/concurrent-ruby
that referenced
this pull request
Jan 29, 2024
Closes: ruby-concurrency#1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
casperisfine
pushed a commit
to casperisfine/concurrent-ruby
that referenced
this pull request
Jan 29, 2024
Closes: ruby-concurrency#1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
|
Thank you for the PR, let's continue this on #1038 |
casperisfine
pushed a commit
to casperisfine/concurrent-ruby
that referenced
this pull request
Feb 1, 2024
Closes: ruby-concurrency#1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
casperisfine
pushed a commit
to casperisfine/concurrent-ruby
that referenced
this pull request
Feb 1, 2024
Closes: ruby-concurrency#1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
eregon
pushed a commit
that referenced
this pull request
Feb 1, 2024
Closes: #1035 A running gag since the introduction of containerization is software that starts one process per logical or physical core while running inside a container with a restricted CPU quota and totally blowing up memory usage in containerized environments. The proper question to ask is how many CPU cores are usable, not how many the machine has. To do that we have to read the cgroup info from `/sys`. There is two way of doing it depending on the version of cgroups used. Co-Authored-By: usiegl00 <50933431+usiegl00@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
As rails depends on
physical_processor_countto spin up the right number of puma workers, rails will fall over in a containerized environment where cgroups are used to limit cpu time. This pr is an attempt to accurately reflect the available cpu cores when the ruby process is running inside a cgroup.