Add meta ID and owner to find workspace image-build Pod

[jenting]

Description

Add meta ID (workspace ID) and owner to find the workspace image-build Pod.

The integration test TestParallelBaseImageBuild did not hit this issue on core-dev env but it always happens on preview-environment.

The integration test calls image-builder-mk3 Build twice with the same dockerfile. The image-builder-mk3 requests the ws-manager to create the image-builder Pod for building the container image. If the image-builder Pod always exists, the ws-manager returns the existed image-builder Pod to the image-builder-mk3.

In core-dev, the 2nd image-build Build call would not hit this code block However, in preview-environment, the 2nd image-build Build call would hit this code block. It seems to be a race condition.

• In core-dev, the ws-manager can receives the StartWorkspace call simultaneously.
• In preview-environment, the ws-manager receives the 2nd StartWorkspace call after the 1st image build Pod created.
  

  gitpod/components/ws-manager/pkg/manager/manager.go

  Lines 176 to 182 in 6b5f345

  	if len(wss.Status) >= 1 {
  	status := wss.Status[0]
  	return &api.StartWorkspaceResponse{
  	Url: status.Spec.Url,
  	OwnerToken: status.Metadata.Owner,
  	}, nil
  	}

In general, to find the corresponding workspace Pods, we should add the meta ID (workspace ID) as well as the owner name.
Otherwise, the 2nd image-build Build call will not launch a new Pod to execute the image build because the 1st image-build Pod matches the annotations.

Related Issue(s)

Fixes #9782

How to test

./dev/preview/install-k3s-kubeconfig.sh
cd test

go test -v ./... \
  -kubeconfig=/home/gitpod/.kube/config \
  -namespace=default \
  -run=TestParallelBaseImageBuild

Release Notes

None

Documentation

N/A

[csweichel]

Did not test, but change and explanation make perfect sense.
Did verify that the GetWorkspaces implementation respects the additional filter.