feat: add advanced datapath observability config option

[TheKangaroo]

fixes #1755

I'm not sure about the variable names though. Any ideas for shorter names?

[TheKangaroo]

Hi @ericyz , WDYT about the variable names? I shortened them to

monitoring_enable_observability_metrics = false
monitoring_observability_metrics_relay_mode = null

The default for the mode variable is INTERNAL_VPC_LB in my private cluster, but I suspect this is different for different cluster types. I cannot verify this though. Anyway I think a default of null makes sense?

[TheKangaroo]

@ericyz @apeabody could you have a look at this?

[apeabody]

/gcbrun

[TheKangaroo]

I had to merge the latest master.
However, the pipeline status was green before.
I don't think I can trigger /gcbrun myself, can I?

[apeabody]

/gcbrun

[TheKangaroo]

@apeabody I have rebeased merged master again. Last time the pipeline status was red. Did the PR break something?

[apeabody]

/gcbrun

[apeabody]

@apeabody I have rebeased merged master again. Last time the pipeline status was red. Did the PR break something?

Hi @TheKangaroo - For terraform-google-kubernetes-engine-int-trigger a maintainer needs to comment gcbrun after the latest commits, otherwise the check won't actually run in CloudBuild and will (confusingly) be marked as failed in GitHub.

I just triggered it now. If the previous result was red from an actual run, it's very possible that result was a flake.

[TheKangaroo]

Thanks for the explanation @apeabody.
Looks good now, doesn't it?

[apeabody]

Thanks for the contribution @TheKangaroo!

[Markieta]

Trying to disable these does not seem to take effect:

  monitoring_enable_observability_metrics = false
  monitoring_observability_metrics_relay_mode = "DISABLED"

Not sure if the problems lies in this module or the underlying google_container_cluster resource.

[TheKangaroo]

@Markieta
Hmm, thats strange, I run all my clusters with:

  monitoring_enable_observability_metrics     = true
  monitoring_observability_metrics_relay_mode = "INTERNAL_VPC_LB"

but changed one to test this to

  monitoring_enable_observability_metrics     = false
  monitoring_observability_metrics_relay_mode = "INTERNAL_VPC_LB"

and I think everyting works as expected.

Terraform needed like half an hour to complete, but after that, everything seems to be fine.

The only thing that was a bit suprising to me ist that the cluster state show true if enalbed and nothing if disabled:
Enabled (before terraform change):

> gcloud container clusters describe <name> | grep monitoringConfig -A 3
monitoringConfig:
  advancedDatapathObservabilityConfig:
    enableMetrics: true
    relayMode: INTERNAL_VPC_LB

Disabled (after terraform change):

> gcloud container clusters describe <name> | grep monitoringConfig -A 3
monitoringConfig:
  advancedDatapathObservabilityConfig:
    relayMode: INTERNAL_VPC_LB
  componentConfig: {}

I think it would be better if you opened a new issue describing exactly what is happening.

[Markieta]

Thanks, perhaps it's only an issue for existing cluster (modules) that were created before adding these inputs? I've created an issue.