Codecov Report

Merging #3799 into master will decrease coverage by 0.04%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master    #3799      +/-   ##
==========================================
- Coverage   42.33%   42.29%   -0.05%     
==========================================
  Files         130      130              
  Lines       14678    14688      +10     
==========================================
- Hits         6214     6212       -2     
- Misses       7539     7550      +11     
- Partials      925      926       +1

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f01665a...b02e20f. Read the comment docs.

@dmcgowan @crosbymichael @kolyshkin ptal

I know about it, its not ready yet from the cgroup lib POV

I will update PR when containerd/cgroups reach feature-complete for memory stats

Added PID metrics.

I intentionally omitted other metrics so that multiple contributors can work on them in parallel as we have done in containerd/cgroups#104. (Thanks @Zyqsempai and Michael)

How to test this PR

$ mkdir /sys/fs/cgroup/default
$ echo +pids > /sys/fs/cgroup/default/cgroup.subtree_control
$ ctr run -t --runc-binary crun --rm docker.io/library/alpine:latest foo
$ ctr t metrics foo
ID     TIMESTAMP                                  
foo    2019-11-16 08:53:45.740847757 +0000 UTC    

METRIC          VALUE                   
pids.current    1                       
pids.limit      18446744073709551615

• Currently subtree_control needs to be toggled outside containerd. We may want to change this behavior later.
• --runc-binary crun is not necessary but recommended until runc gets matured.

@AkihiroSuda Can you please create todo list as well, for better control of what and when should be done, I would love to work here as well.

checklist: #3726

Shall we merge it? to continue working on other metrics!

Let me test this on my setup

Do you know the kernel configs that I need for the ebpf devices?

Trying this out I get:

ctr: OCI runtime create failed: container_linux.go:346: starting container process caused "process_linux.go:449: container init caused \"process_linux.go:415: setting cgroup config for procHooks process caused \\\"failed to load program: function not implemented\\\"\"": unknown

I dont' like how the stats are handled. We should output the prometheus metrics without any changes. People shouldn't have to have all new metric keys because we start using cgroups v2. We need to make the v2 code to the existing stats structs within containerd

I just tested this and got containers running under v2 with runc. It's looking good on the runtime side, we just need to map the stats structures so that this isn't a huge burden for users. We also need to figureout what we want to do with the subtree control file whenever we create a directory for a namespace.

Do you know the kernel configs that I need for the ebpf devices?

kernel >= 4.15 with CONFIG_CGROUP_DEVICE and CONFIG_CGROUP_BPF is required.
I'm using kernel 5.3 shipped with Ubuntu 19.10.

I got it to work. you need a couple others like BPF_syscall

If separating stat structs are problematic for downstream projects, maybe we can have some translator library in contrib directory to help them.

also added commit for enabling controllers automatically

need to rebase 😂

rebased

Build succeeded.

• containerd-build-arm64 : SUCCESS in 4m 26s (non-voting)

LGTM