Improve Pod Container Status Handling

[Bianco95]

Description

Updated handling of container statuses and introduces better support for init containers in the system.

Key Changes

1. Preserve Failed Container Status

   • Successful containers no longer overwrite the status of failed containers, ensuring accurate state reporting.

2. Improved Init Container Handling

   • If a pod includes init containers, the system now verifies their completion before proceeding to the main containers.
   • On the plugin side, init containers should be correctly handled:
     • Executed sequentially in the defined order.
     • Must complete before starting the main containers.

These changes ensure a more robust and predictable behavior for pods with complex container setups, aligning the implementation with expected Kubernetes standards.

Testing

To validate the new changes, testing was conducted using a setup that includes a Virtual Kubelet (VK) deployed in a Kubernetes (k8s) cluster created with kind, an InterLink API server running behind an OAuth proxy on an edge node alongside the Docker plugin.

1. Tests from the suite available in this repo https://github.com/interTwin-eu/vk-test-set have been successful executed.

========================================================================== test session starts ==========================================================================
platform linux -- Python 3.10.12, pytest-8.2.0, pluggy-1.5.0 -- /usr/bin/python3
cachedir: .pytest_cache
rootdir: /home/giulio/vk-setup/vk-test-set
configfile: pyproject.toml
plugins: typeguard-4.3.0
collected 10 items                                                                                                                                                      

vktestset/basic_test.py::test_namespace_exists[interlink] PASSED                                                                                                  [ 10%]
vktestset/basic_test.py::test_node_exists[virtual-node] PASSED                                                                                                    [ 20%]
vktestset/basic_test.py::test_manifest[virtual-node-000-hello-world.yaml] PASSED                                                                                  [ 30%]
vktestset/basic_test.py::test_manifest[virtual-node-010-simple-python.yaml] PASSED                                                                                [ 40%]
vktestset/basic_test.py::test_manifest[virtual-node-020-python-env.yaml] PASSED                                                                                   [ 50%]
vktestset/basic_test.py::test_manifest[virtual-node-030-simple-shared-volume.yaml] PASSED                                                                         [ 60%]
vktestset/basic_test.py::test_manifest[virtual-node-040-config-volumes.yaml] PASSED                                                                               [ 70%]
vktestset/basic_test.py::test_manifest[virtual-node-050-limits.yaml] PASSED                                                                                       [ 80%]
vktestset/basic_test.py::test_manifest[virtual-node-060-init-container.yaml] PASSED                                                                               [ 90%]
vktestset/basic_test.py::test_manifest[virtual-node-070-rclone-bind.yaml] PASSED                                                                                  [100%]

==================================================================== 10 passed in 120.25s (0:02:00) =====================================================================

2. A POD with multiple init containers and multiple containers has been tested by also varying the order of the failed and successful containers

apiVersion: v1
kind: Pod
metadata:
  name: hello-world-1
  namespace: interlink

spec:
  restartPolicy: Never
  nodeSelector:
    kubernetes.io/hostname: virtual-node

  initContainers:
  - name: init-cnt1
    image: ghcr.io/grycap/cowsay
    command: ["/bin/sh", "-c"]
    args: ["echo 'init cnt 1. now sleep'; sleep 5; exit 0"]
    imagePullPolicy: Always

  - name: init-cnt2
    image: ghcr.io/grycap/cowsay
    command: ["/bin/sh", "-c"]
    args: ["echo 'init cnt 2. now sleep'; sleep 5; exit 0"]
    imagePullPolicy: Always

  containers:
  - name: successful-container
    image: busybox
    command: ["/bin/sh", "-c"]
    args: ["exit 0"]
    imagePullPolicy: Always

  - name: hello-world
    image: ghcr.io/grycap/cowsay
    command: ["/bin/sh", "-c"]
    args: ["echo 'hello world'; sleep 5"]
    imagePullPolicy: Always

  - name: fail-container
    image: busybox
    command: ["/bin/sh", "-c"]
    args: ["exit 1"]  # FORCE FAIL EXIT CODE
    imagePullPolicy: Always

  dnsPolicy: ClusterFirst

  tolerations:
    - key: virtual-node.interlink/no-schedule
      operator: Exists
      effect: NoSchedule

[dciangot]

@Bianco95 lgtm, can you also open an issue for vktest repo noting that we need to include tests for this scenarios (your point 2)? That way we have the test enabled in the e2e PR check

[Bianco95]

Ok, an issue has been opened in the vk-test-set repo (interTwin-eu/vk-test-set#7)