add system.linux.memory.slab to system metrics

[rogercoll]

Fixes #531

Related to open-telemetry/opentelemetry-collector-contrib#7417

Changes

Adds the slab memory metric that is only available on Linux systems as system.linux.memory.slab.

Merge requirement checklist

• CONTRIBUTING.md guidelines followed.
• Change log entry added, according to the guidelines in When to add a changelog entry.
  • If your PR does not need a change log, start the PR title with [chore]
• schema-next.yaml updated with changes to existing conventions.

[rogercoll]

For the attributes, it references the system.memory.state which will only match the total well-known value.
reclaimable and unreclaimable are additional custom state attributes normally associated with slab memory. For example, the current (0.101.0) version of the opentelemetry-collector-contrib uses system.memory.slab_reclaimable and system.memory.slab_unreclaimable to indicate those states.

Should this metric reference the system.memory.state attributes or create a new registry (e.g system.linux.memory.slab.state) that explicitly indicates the total, reclaimable and unreclaimable states?

[ChrsMark]

Thank's @rogercoll for this PR!

Should this metric reference the system.memory.state attributes or create a new registry (e.g system.linux.memory.slab.state) that explicitly indicates the total, reclaimable and unreclaimable states?

I think it makes sense to have a specific *.state attribute which would include the reclaimable and unreclaimable.
As it was pointed out at #522 (from the usage naming convention):

Where appropriate, the sum of usage over all attribute values SHOULD be equal to the limit.

/cc @open-telemetry/semconv-system-approvers @braydonk

[mx-psi]

Agreed with @ChrsMark!

[rogercoll]

I think it makes sense to have a specific *.state attribute which would include the reclaimable and unreclaimable.

Sounds good to me. To be aligned with the system.memory model, I think the following could work:

Metric name: system.linux.memory.slab.usage       
Attributes: system.linux.memory.slab.state (reclaimable and unreclaimable)

Metric name: system.linux.memory.slab.limit 

[ChrsMark]

From SemConv perspective looks good to me.

One question: do we actually plan to implement the system.linux.memory.slab.limit as well in the Collector?

[rogercoll]

One question: do we actually plan to implement the system.linux.memory.slab.limit as well in the Collector?

I would say so, the current sum of all system.memory.usage states do not equal to the limit which is needed for v1. If this PR moves forward, I can open an issue in the collector to track the implementation.

cc @dmitryax

[dmitryax]

LGTM

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 7 days.

[lmolkova]

Left a suggestion on attribute naming, look good otherwise!

[trisch-me]

@rogercoll there are still some failing checks that prevent from merging

[trisch-me]

@joaopgrassi could you check this one as you have requested changes but I think it's resolved now

[rogercoll]

@rogercoll there are still some failing checks that prevent from merging

I think the failing checks are unrelated to the PR but to endpoints blocking the CI IP. Error message:

ERROR: 1 dead links found in ./docs/resource/k8s.md !
  [✖] https://www.itu.int/ITU-T/studygroups/com17/oid.html → Status: 0

I can reach the failed endpoint on my end.

stale