Fix the failing nightly test (#3803)

The test, as preparations, installs the latest version of OLM,
then it installs the cert-manager.

But the latest version of OLM, rleased 2 days ago, requires
cert-manager to be already installed.

This PR changes the order so the test will first install cert-manager,
and only then install OLM.

Also, stop using the deploy/cert-manager.yaml file. Instead, use the
file from the cert-manager source, directly.

Signed-off-by: Nahshon Unna-Tsameret <[email protected]>

Author: nunnatsa

automation/nightly/test-nightly-build.sh

@@ -135,7 +135,7 @@ chmod +x operator-sdk
 OLM_VERSION=$(curl https://api.github.com/repos/operator-framework/operator-lifecycle-manager/releases/latest | jq -r .name)
 
 # start K8s cluster
-export KUBEVIRT_PROVIDER=k8s-1.32
+export KUBEVIRT_PROVIDER=k8s-1.34
 export KUBEVIRT_MEMORY_SIZE=12G
 export KUBEVIRT_NUM_NODES=4
 # cluster/kubevirtci_tag.txt is auto updated by hack/bump-kubevirtci.sh
@@ -145,13 +145,11 @@ make cluster-up
 export KUBECONFIG=$(_kubevirtci/cluster-up/kubeconfig.sh)
 export KUBECTL=$(pwd)/_kubevirtci/cluster-up/kubectl.sh
 
+CMD=${KUBECTL} ./hack/deploy-cert-manager.sh
+
 # install OLM on the cluster
 ./operator-sdk olm install --version "${OLM_VERSION}"
 
-# Deploy cert-manager for webhooks
-$KUBECTL apply -f deploy/cert-manager.yaml
-$KUBECTL -n cert-manager wait deployment/cert-manager-webhook --for=condition=Available --timeout="300s"
-
 trap "dump" INT TERM EXIT
 
 # install HCO on the cluster

deploy/deploy.sh

@@ -4,6 +4,8 @@ set -ex
 
 hco_namespace=kubevirt-hyperconverged
 
+CERT_MANAGER_VERSION=${CERT_MANAGER_VERSION:-"v1.18.2"}
+
 IS_OPENSHIFT=${IS_OPENSHIFT:-false}
 if kubectl api-resources |grep clusterversions |grep config.openshift.io; then
   IS_OPENSHIFT="true"
@@ -36,7 +38,7 @@ kubectl apply ${LABEL_SELECTOR_ARG} -f https://raw.githubusercontent.com/kubevir
 kubectl apply ${LABEL_SELECTOR_ARG} -f https://raw.githubusercontent.com/kubevirt/hyperconverged-cluster-operator/main/deploy/crds/application-aware-quota00.crd.yaml
 
 # Deploy cert-manager for webhook certificates
-kubectl apply ${LABEL_SELECTOR_ARG} -f https://raw.githubusercontent.com/kubevirt/hyperconverged-cluster-operator/main/deploy/cert-manager.yaml
+kubectl apply -f "https://github.com/cert-manager/cert-manager/releases/download/${CERT_MANAGER_VERSION}/cert-manager.yaml"
 kubectl -n cert-manager wait deployment/cert-manager --for=condition=Available --timeout="300s"
 kubectl -n cert-manager wait deployment/cert-manager-webhook --for=condition=Available --timeout="300s"
 

hack/clean.sh

@@ -52,7 +52,7 @@ source hack/common.sh
 "${CMD}" delete -f _out/service_account.yaml --ignore-not-found || true
 
 # Remove cert-manager
-"${CMD}" delete -f _out/cert-manager.yaml --ignore-not-found || true
+"${CMD}" delete -f "https://github.com/cert-manager/cert-manager/releases/download/${CERT_MANAGER_VERSION}/cert-manager.yaml" --ignore-not-found || true
 "${CMD}" delete validatingwebhookconfigurations cert-manager-webhook --ignore-not-found || true
 "${CMD}" delete mutatingwebhookconfigurations cert-manager-webhook --ignore-not-found || true
 "${CMD}" delete secrets -n kubevirt-hyperconverged -l controller.cert-manager.io/fao=true --ignore-not-found || true

hack/common.sh

@@ -39,6 +39,8 @@ KUBECTL=${KUBECTL:-}
 TEST_OUT_PATH=${TEST_OUT_PATH:-"tests/func-tests/_out"}
 JOB_TYPE=${JOB_TYPE:-}
 
+CERT_MANAGER_VERSION=${CERT_MANAGER_VERSION:-"v1.18.2"}
+
 KUBECTL=$(which kubectl 2> /dev/null) || true
 
 if [ -z "${CMD}" ]; then

hack/deploy-cert-manager.sh

@@ -0,0 +1,17 @@
+#!/usr/bin/env bash
+
+source hack/common.sh
+
+set -ex
+
+CERT_MANAGER_TIMEOUT=${CERT_MANAGER_TIMEOUT:-"120s"}
+
+echo "Installing cert-manager ${CERT_MANAGER_VERSION}"
+${CMD} apply -f https://github.com/cert-manager/cert-manager/releases/download/${CERT_MANAGER_VERSION}/cert-manager.yaml
+
+echo "Waiting for cert-manager to be ready..."
+${CMD} wait --for=condition=Available --namespace=cert-manager deployment/cert-manager --timeout=${CERT_MANAGER_TIMEOUT}
+${CMD} wait --for=condition=Available --namespace=cert-manager deployment/cert-manager-cainjector --timeout=${CERT_MANAGER_TIMEOUT}
+${CMD} wait --for=condition=Available --namespace=cert-manager deployment/cert-manager-webhook --timeout=${CERT_MANAGER_TIMEOUT}
+
+echo "cert-manager ${CERT_MANAGER_VERSION} installed successfully"

hack/deploy.sh

@@ -114,9 +114,7 @@ if [ "$IS_OPENSHIFT" != "true" ]; then
     LABEL_SELECTOR_ARG="-l name!=ssp-operator,name!=hyperconverged-cluster-cli-download"
 fi
 
-# Deploy cert-manager for webhooks
-"${CMD}" apply -f _out/cert-manager.yaml
-"${CMD}" -n cert-manager wait deployment/cert-manager-webhook --for=condition=Available --timeout="300s"
+hack/deploy-cert-manager.sh
 
 # Deploy local manifests
 "${CMD}" apply $LABEL_SELECTOR_ARG -f _out/cluster_role.yaml