CI scripts improvements

Updates to OCP: * add a retry to the policy create to fix OCP CI flake * add more ram to the OCP master single node * add a wait for registry to be available * update the ocp version to latest 4.14 rc release Updates for K8s: * switch to not use dockerhub for the registry image to overcome the pull limit * add alias for golang to the dockerhub image github actions: * always remove the cluster Signed-off-by: Sebastian Sch <[email protected]>
k8snetworkplumbingwg · Oct 22, 2023 · 6f54e23 · 6f54e23
1 parent bf595e4
commit 6f54e23
Show file tree

Hide file tree

Showing 4 changed files with 32 additions and 10 deletions.
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
@@ -119,6 +119,12 @@ jobs:
           name: k8s-artifact
           path: ./k8s-artifacts
 
+      - name: remove virtual cluster
+        if: always()
+        run: |
+          kcli delete cluster $CLUSTER_NAME -y
+          kcli delete network $CLUSTER_NAME -y
+
   virtual-ocp:
     name: ocp
     needs: [
@@ -145,3 +151,9 @@ jobs:
         with:
           name: ocp-artifact
           path: ./ocp-artifacts
+
+      - name: remove virtual cluster
+        if: always()
+        run: |
+          kcli delete cluster $CLUSTER_NAME -y
+          kcli delete network $CLUSTER_NAME -y
diff --git a/hack/run-e2e-conformance-virtual-cluster.sh b/hack/run-e2e-conformance-virtual-cluster.sh
@@ -120,6 +120,9 @@ controller_ip=`kubectl get node -o wide | grep ctlp | awk '{print $6}'`
 insecure_registry="[[registry]]
 location = \"$controller_ip:5000\"
 insecure = true
+
+[aliases]
+\"golang\" = \"docker.io/library/golang\"
 "
 
 cat << EOF > /etc/containers/registries.conf.d/003-${cluster_name}.conf
@@ -223,7 +226,7 @@ spec:
         - effect: NoSchedule
           key: node-role.kubernetes.io/control-plane
       containers:
-      - image: docker.io/registry:latest
+      - image: quay.io/libpod/registry:2.8.2
         imagePullPolicy: Always
         name: registry
         volumeMounts:

diff --git a/hack/run-e2e-conformance-virtual-ocp.sh b/hack/run-e2e-conformance-virtual-ocp.sh
@@ -41,8 +41,8 @@ kcli create network -c 192.168.123.0/24 ocp
 kcli create network -c 192.168.${virtual_router_id}.0/24 --nodhcp -i $cluster_name
 
 cat <<EOF > ./${cluster_name}-plan.yaml
-tag: 4.14.0-rc.1
-ctlplane_memory: 24576
+tag: 4.14.0-rc.6
+ctlplane_memory: 32768
 worker_memory: 8192
 pool: default
 disk_size: 50
@@ -191,13 +191,17 @@ podman build -t "${SRIOV_NETWORK_CONFIG_DAEMON_IMAGE}" -f "${root}/Dockerfile.sr
 echo "## build webhook image"
 podman build -t "${SRIOV_NETWORK_WEBHOOK_IMAGE}" -f "${root}/Dockerfile.webhook" "${root}"
 
+echo "## wait for registry to be available"
+kubectl wait configs.imageregistry.operator.openshift.io/cluster --for=condition=Available --timeout=120s
+
 dockercgf=`kubectl -n ${NAMESPACE} get sa builder -oyaml | grep imagePullSecrets -A 1 | grep -o "builder-.*"`
 auth=`kubectl -n ${NAMESPACE} get secret ${dockercgf} -ojson | jq '.data.".dockercfg"'`
 auth="${auth:1:-1}"
 auth=`echo ${auth} | base64 -d`
 echo ${auth} > registry-login.conf
 
-pass=$( jq .\"$registry\".password registry-login.conf )
+internal_registry="image-registry.openshift-image-registry.svc:5000"
+pass=$( jq .\"$internal_registry\".password registry-login.conf )
 podman login -u serviceaccount -p ${pass:1:-1} $registry --tls-verify=false
 
 podman push --tls-verify=false "${SRIOV_NETWORK_OPERATOR_IMAGE}"

diff --git a/test/conformance/tests/test_sriov_operator.go b/test/conformance/tests/test_sriov_operator.go
@@ -143,8 +143,9 @@ var _ = Describe("[sriov] operator", func() {
 				cfg.Spec.ConfigDaemonNodeSelector = map[string]string{
 					"sriovenabled": "true",
 				}
-				err = clients.Update(context.TODO(), &cfg)
-				Expect(err).ToNot(HaveOccurred())
+				Eventually(func() error {
+					return clients.Update(context.TODO(), &cfg)
+				}, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred())
 
 				By("Checking that a daemon is scheduled only on selected node")
 				Eventually(func() bool {
@@ -159,8 +160,9 @@ var _ = Describe("[sriov] operator", func() {
 				}, &cfg)
 				Expect(err).ToNot(HaveOccurred())
 				cfg.Spec.ConfigDaemonNodeSelector = map[string]string{}
-				err = clients.Update(context.TODO(), &cfg)
-				Expect(err).ToNot(HaveOccurred())
+				Eventually(func() error {
+					return clients.Update(context.TODO(), &cfg)
+				}, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred())
 
 				By("Checking that a daemon is scheduled on each worker node")
 				Eventually(func() bool {
@@ -2316,8 +2318,9 @@ func createVanillaNetworkPolicy(node string, sriovInfos *cluster.EnabledNodes, n
 			DeviceType: "netdevice",
 		},
 	}
-	err = clients.Create(context.Background(), config)
-	Expect(err).ToNot(HaveOccurred())
+	Eventually(func() error {
+		return clients.Create(context.Background(), config)
+	}, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred())
 
 	Eventually(func() sriovv1.Interfaces {
 		nodeState, err := clients.SriovNetworkNodeStates(operatorNamespace).Get(context.Background(), node, metav1.GetOptions{})