Support monitoring AWS Cloud EKS

[pg-yang]

• If this pull request closes/resolves/fixes an existing issue, replace the issue number. Closes #.
• Update the CHANGES log.

Related issue #9883

1. UI repo has synced
2. I copied otlp proto to java-test-service for mocking EKS e2e test

UI screenshot

Node metrics

Service metrics

[wankai123]

Can you confirm why these metrics use downsampling(SUM), I can see these units are Count/Second from the
origin OTEL metrics. SUM would plus the value of the metrics each time, should use the default downsampling
AVG?

[wankai123]

please ignore the approval for now... It's a mistake.

[wu-sheng]

Sum may be not correct, especially at high dimensiona(hour/day)

[wu-sheng]

@pg-yang Could you check here? Network TX Error Count (per second) is the description, why it could be downsampling(SUM)?

[pg-yang]

Yep, I will confirm these issues by the end of this week at the latest.

[wu-sheng]

Yes, it is.

[pg-yang]

I couldn't find any documentation that describes the calculation, but I found the source code.
In a word, the calculation logic is (preValue-currentValue) / (preTimeInSecond-currentTimeInSecond).

1. value (preValue or currentValue) is cumulative count of transmit errors encountered. It's defined in https://github.com/google/cadvisor/blob/master/info/v1/container.go#L434
2. currentValue, currentTimeInSecond will be stored to cache as preValue, preTimeInSecond for the next calculation

I put the source code here:

1. Calculate method: https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/951245d4273845de59db3b00f5445bcad9ce4372/internal/aws/metrics/metric_calculator.go#L67
2. The calculateFunc method that is called by Calculate method: https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/951245d4273845de59db3b00f5445bcad9ce4372/receiver/awscontainerinsightreceiver/internal/cadvisor/extractors/extractor.go#L127

BTW

  - name: net_tx_error
    exp: node_network_tx_errors.downsampling(SUM)

The above metric is organized from node level to cluster level,so I think latest should be inappropriate.Also,if we use avg, the meaning will be net errors per second, per node.

[wu-sheng]

From your description, I would say avg makes more sense.
Sum for day and hour dimension would be a disaster, right?
We have to keep per node, I am afraid, there is no way we could sum first and then avg, because from data process perspective, there is no difference between two node and two period report of one node, right?

[pg-yang]

Agree.

[pg-yang]

dismiss

[wu-sheng]

LGTM. Pending on more confirmations from @kezhenxu94 @wankai123

[kezhenxu94]

This sample yaml is way too long that it doesn't highlight the important part, please remove unrelated content and replace with their official documentation link, an example would be

# ... other resources definitions
# OTEL configuration that specify OAP address("oap-service:11800") as otlp exporter address
apiVersion: v1
kind: ConfigMap
metadata:
  name: otel-agent-conf
  namespace: aws-otel-eks
  labels:
    app: opentelemetry
    component: otel-agent-conf
data:
  otel-agent-config: |
    # ...
    receivers:
      awscontainerinsightreceiver:

    # Make sure to add a job-name attribute
    processors:
      resource/job-name:
        attributes:
        - key: job_name   
          value: aws-cloud-eks-monitoring
          action: insert     

    exporters:
      otlp:
        endpoint: oap-service:11800
        tls:
          insecure: true
      logging:
          loglevel: debug          

    service:
      pipelines:
        metrics:
          receivers: [awscontainerinsightreceiver]
          processors: [resource/job-name]
          exporters: [otlp,logging]

# ... other resources

[kezhenxu94]

If you want to give a complete sample, let's create a dedicated yaml file and link to that file, so users can simply use kubectl apply -f sample.yaml to set up the sample, without having to copy the large chunk of content.

[wu-sheng]

I think a complete OTEL YAML file would be good to host in the doc. Or we could link the file to the incoming file in the showcase.

[pg-yang]

I think a complete OTEL YAML file would be good to host in the doc

Agree with it, we don't have showcase for now.

[wu-sheng]

@pg-yang Are you going to move this to a separate file?

[pg-yang]

I only provide a OTEL configuration sample in the doc

[kezhenxu94]

A nit, otherwise LGTM

[wu-sheng]

@pg-yang Let's fix the left issues and merge this :)

[wu-sheng]

LGTM

[wankai123]

LGTM