ref(smoketest): break out lengthy script; also remove containerdruntimeoptions if k3d

Signed-off-by: Vaughn Dice <vdice@akamai.com>

Author: vdice

.github/workflows/helm-chart-smoketest.yml

@@ -156,58 +156,26 @@ jobs:
             --set rcm.shimDownloaderConfig.content.sleepDuration=3 \
             deploy/helm
 
-      - name: verify configMap for shim-downloader is added
-        run: |
-          timeout 60s bash -c 'until [[ $(kubectl -n rcm get $(kubectl get pods -n rcm --no-headers -o name | grep install | head -n1) -o jsonpath="{.spec.initContainers[0].envFrom[0].configMapRef.name}" 2>/dev/null) == "configmap-test" ]]; do sleep 2; done'
-
       - name: apply shims
         run: |
           for shim_file in $(ls config/samples/sample_shim*); do
-            if [[ "${{ matrix.config.type }}" == "microk8s" ]]; then
-              cp $shim_file $shim_file.microk8s
-              shim_file=$shim_file.microk8s
+            if [[ "${{ matrix.config.type }}" == "microk8s" || "${{ matrix.config.type }}" == "k3d" ]]; then
+              cp $shim_file $shim_file.amended
+              shim_file=$shim_file.amended
               # update file to remove the 'containerdRuntimeOptions' field
               # as there is a known bug that MicroK8s containerd does not pass the options
+              # and k3d uses cgroupfs, not systemd
               yq -i 'del(.spec.containerdRuntimeOptions)' $shim_file
             fi
             kubectl apply -f $shim_file
           done
 
-      - name: label nodes and wait for shim to be ready
-        run: |
-          for shim_file in $(ls config/samples/sample_shim*); do
-            label="$(cat $shim_file | yq '.spec.nodeSelector' | tr -d '"' | tr -d '[:space:]' | sed s/:/=/g)"
-            kubectl label node --all $label
-
-            shim_name="$(cat $shim_file | yq '.metadata.name')"
-            # TODO: k3d can take a long round of failed install pods (exit code 6 when curling the artifact?)
-            # Once this behavior is diagnosed and resolved, we should be able to shorten this timeout substantially
-            timeout=600
-            SECONDS=0 # Reset the internal bash timer to 0
-            success=false
-
-            echo "Waiting for the $shim_name shim to be ready/installed..."
-
-            while [[ $SECONDS -lt $timeout ]]; do
-              # Fetch both nodes and nodesReady
-              read -r nodes nodesReady <<< $(kubectl get shim "$shim_name" \
-                -o jsonpath='{.status.nodes} {.status.nodesReady}' 2>/dev/null)
-
-              # Check to see if all nodes are ready
-              if [[ -n "$nodes" ]] && [[ -n "$nodesReady" ]] && [[ "$nodes" -eq "$nodesReady" ]]; then
-                echo "Success: all nodes have the $shim_name shim installed."
-                success=true
-                break
-              fi
-
-              sleep 2
-            done
+      - name: label nodes and wait for shims to be ready
+        run: ./scripts/sample-shims-label-nodes.sh
 
-            if [[ "${success}" != "true" ]]; then
-              echo "Error: Timed out after ${timeout}s waiting for the $shim_name shim to be ready."
-              exit 1
-            fi
-          done
+      - name: verify configMap for shim-downloader is added
+        run: |
+          timeout 60s bash -c 'until [[ $(kubectl -n rcm get $(kubectl get pods -n rcm --no-headers -o name | grep install | head -n1) -o jsonpath="{.spec.initContainers[0].envFrom[0].configMapRef.name}" 2>/dev/null) == "configmap-test" ]]; do sleep 2; done'
 
       # TODO: unify testdata/apps to all model the same behavor, eg simple web server, etc
       - name: run and verify spin app
@@ -228,14 +196,14 @@ jobs:
         if: failure()
         run: |
           kubectl get pods -A
-          kubectl describe shims
-          kubectl describe runtimeclasses
+          kubectl describe shims || true
+          kubectl describe runtimeclasses || true
 
           # Get install pod logs
           # Note: there may be multiple pods pending k3d fix for issue https://github.com/spinframework/runtime-class-manager/issues/393
           for shim_file in $(ls config/samples/sample_shim*); do
             shim_name="$(cat $shim_file | yq '.metadata.name')"
-            install_pods="$(kubectl get pods -n rcm --no-headers -o name | grep $shim_name-install)"
+            install_pods="$(kubectl get pods -n rcm --no-headers -o name | grep $shim_name-install || true)"
             for pod in $install_pods; do
               kubectl describe -n rcm $pod || true
               kubectl logs -n rcm -c downloader $pod || true

scripts/sample-shims-apply.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+set -euo pipefail
+
+# Currently used in .github/workflows/helm-chart-smoketest.yml
+# Iterates through all sample shims and applies their labels to all nodes
+# using the current kubernetes context.
+# Waits for all corresponding Shim resources to be ready, else fails.
+
+for shim_file in $(ls config/samples/sample_shim*); do
+  label="$(cat $shim_file | yq '.spec.nodeSelector' | tr -d '"' | tr -d '[:space:]' | sed s/:/=/g)"
+  kubectl label node --all $label
+
+  shim_name="$(cat $shim_file | yq '.metadata.name')"
+  timeout=300
+  SECONDS=0 # Reset the internal bash timer to 0
+  success=false
+
+  echo "Waiting for the $shim_name shim to be ready/installed..."
+
+  while [[ $SECONDS -lt $timeout ]]; do
+    # Fetch both nodes and nodesReady
+    read -r nodes nodesReady <<< $(kubectl get shim "$shim_name" \
+      -o jsonpath='{.status.nodes} {.status.nodesReady}' 2>/dev/null)
+
+    # Check to see if all nodes are ready
+    if [[ -n "$nodes" ]] && [[ -n "$nodesReady" ]] && [[ "$nodes" -eq "$nodesReady" ]]; then
+      echo "Success: all nodes have the $shim_name shim installed."
+      success=true
+      break
+    fi
+
+    sleep 2
+  done
+
+  if [[ "${success}" != "true" ]]; then
+    echo "Error: Timed out after ${timeout}s waiting for the $shim_name shim to be ready."
+    exit 1
+  fi
+done