[receiver/k8scluster] Invalid image name and tag attributes when containing a digest

[rogercoll]

Component(s)

receiver/k8scluster

What happened?

Description

If a container that is being instrumented using the k8scluster receiver has been deployed using a container image that contains a cryptographic digest, the receiver produces invalid attributes. The regex used to parse the container does not apply to images with a digest.

Steps to Reproduce

1. Deploy a collector in K8s with the k8scluster receiver enabled
2. Deploy a Pod (as a daemonset or any other deployment strategy) with a container's image that contains a digest. For example (simplified output of a pod description):

Containers:
  opentelemetry-collector:
    Container ID:  containerd://6a8fb288d22908242a8508d8625e5a328307a09218a2e659cc9457a0c7a5fe28
    Image:         docker.elastic.co/beats/elastic-agent@sha256:52eeed9e66facc651d4344ef7d9ce912fccff5bb3969e745eed3ab953f309534
    Image ID:      docker.elastic.co/beats/elastic-agent@sha256:52eeed9e66facc651d4344ef7d9ce912fccff5bb3969e745eed3ab953f309534

Expected Result

Resource attributes:
     -> container.image.name: Str(docker.elastic.co/beats/elastic-agent)

Actual Result

Resource attributes:
     -> container.image.name: Str(sha256)
     -> container.image.tag: Str(a048efd6da2a01c20fb2342b624d740e44a7f651183b017c1962835d9e05186a)

Collector version

v0.112.0

Environment information

Environment

OS: (e.g., "Ubuntu 20.04")
Compiler(if manually compiled): (e.g., "go 14.2")

OpenTelemetry Collector configuration

receivers:
   k8s_cluster:
     auth_type: serviceAccount
    metrics:
      k8s.pod.status_reason:
        enabled: true

Log output

No response

Additional context

No response

[github-actions]

Pinging code owners:

• receiver/k8scluster: @dmitryax @TylerHelmuth @povilasv

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[rogercoll]

The following commit makes the receiver's tests fail: e47aa1a

What do you think if instead of using our custom image parser we use the K8s one? Similar to what I am proposing in this draft PR: https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/36281/files

[bacherfl]

I also just had a look and it seems like the current regex expects the sha256 digest to consist of digits only.

I tried to adapt the existing regex to handle this, as well as images containing no tags, as included in the draft PR, so something like the following may be worth a try to use instead of the previous regex:

^(?P<repository>(?:[^/\s]+\/)*[\w\.-]+)(?::(?P<tag>[\w\.-]+))?(?:@sha256:(?P<sha256>[A-Fa-f0-9]+))?$

This may be an alternative to consider if the code owners would like to avoid adding a new dependency for this. However I do see the value of having an established library doing the parsing, as it may be easy to miss certain cases with an own regex.

Pinging the code owners (@dmitryax @TylerHelmuth @povilasv) - which approach would you say is the best in this case?

[ChrsMark]

There is a related PR fixing the k8sattributes processor: #36145

Could we apply a similar logic here?

[rogercoll]

There is a related PR fixing the k8sattributes processor: #36145

Could we apply a similar logic here?

Is there any reasoning behind a new function to parse the image in the k8sattributes processor? I would rather fix the internal docker.ParseImageName function to parse images with digest and use it across all components.