Merge pull request #15276 from justinsb/simple_scalability_test_2

Update scalability test so it passes

tests/e2e/scenarios/lib/common.sh

@@ -31,11 +31,6 @@ if [[ -z "${WORKSPACE-}" ]]; then
     WORKSPACE=$(mktemp -dt kops.XXXXXXXXX)
 fi
 
-if [[ -z "${WORKSPACE-}" ]]; then
-    export WORKSPACE
-    WORKSPACE=$(mktemp -dt kops.XXXXXXXXX)
-fi
-
 if [[ -z "${NETWORKING-}" ]]; then
     export NETWORKING="calico"
 fi

tests/e2e/scenarios/scalability/run-test.sh

@@ -21,17 +21,23 @@ REPO_ROOT=$(git rev-parse --show-toplevel)
 cd "${REPO_ROOT}"
 cd ..
 WORKSPACE=$(pwd)
+cd "${WORKSPACE}/kops"
 
 # Create bindir
-BINDIR=${WORKSPACE}/bin
+BINDIR="${WORKSPACE}/bin"
-export PATH=${BINDIR}:${PATH}
+export PATH="${BINDIR}:${PATH}"
 mkdir -p "${BINDIR}"
 
+# Build kubetest-2 kOps support
+pushd "${WORKSPACE}/kops"
+GOBIN=${BINDIR} make test-e2e-install
+popd
+
 # Setup our cleanup function; as we allocate resources we set a variable to indicate they should be cleaned up
 function cleanup {
   # shellcheck disable=SC2153
   if [[ "${DELETE_CLUSTER:-}" == "true" ]]; then
-      kubetest2 kops "${KUBETEST2_ARGS}" --down || echo "kubetest2 down failed"
+      kubetest2 kops "${KUBETEST2_ARGS[@]}" --down || echo "kubetest2 down failed"
   fi
 }
 trap cleanup EXIT
@@ -43,8 +49,6 @@ if [[ -z "${CLUSTER_NAME:-}" ]]; then
 fi
 echo "CLUSTER_NAME=${CLUSTER_NAME}"
 
-exit 0
-
 if [[ -z "${K8S_VERSION:-}" ]]; then
   K8S_VERSION="$(curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt)"
 fi
@@ -80,8 +84,12 @@ fi
 echo "ADMIN_ACCESS=${ADMIN_ACCESS}"
 
 # cilium does not yet pass conformance tests (shared hostport test)
-create_args="--networking cilium"
+#create_args="--networking cilium"
-create_args="${create_args} --node-size=c6g.medium --master-size=c6g.xlarge --node-count=100"
+create_args="--networking calico"
+# TODO: Use the newer non-DNS mode, more scalable than gossip and generally recommended
+# However, it currently fails two tests (HostPort & OIDC) so need to track that down
+#create_args="--dns none"
+create_args="${create_args} --node-size=c6g.medium --master-size=c6g.xlarge --node-count=101"
 if [[ -n "${ZONES:-}" ]]; then
     create_args="${create_args} --zones=${ZONES}"
 fi
@@ -93,25 +101,29 @@ echo "KOPS_FEATURE_FLAGS=${KOPS_FEATURE_FLAGS}"
 
 
 # Note that these arguments for kubetest2
-KUBETEST2_ARGS=""
+KUBETEST2_ARGS=()
-KUBETEST2_ARGS="${KUBETEST2_ARGS} -v=2 --cloud-provider=${CLOUD_PROVIDER}"
+KUBETEST2_ARGS+=("-v=2")
-KUBETEST2_ARGS="${KUBETEST2_ARGS} --cluster-name=${CLUSTER_NAME:-}"
+KUBETEST2_ARGS+=("--cloud-provider=${CLOUD_PROVIDER}")
-KUBETEST2_ARGS="${KUBETEST2_ARGS} --kops-binary-path=${KOPS_BIN}"
+KUBETEST2_ARGS+=("--cluster-name=${CLUSTER_NAME:-}")
-KUBETEST2_ARGS="${KUBETEST2_ARGS} --admin-access=${ADMIN_ACCESS:-}"
+KUBETEST2_ARGS+=("--kops-binary-path=${KOPS_BIN}")
-KUBETEST2_ARGS="${KUBETEST2_ARGS} --env=KOPS_FEATURE_FLAGS=${KOPS_FEATURE_FLAGS}"
+KUBETEST2_ARGS+=("--admin-access=${ADMIN_ACCESS:-}")
+KUBETEST2_ARGS+=("--env=KOPS_FEATURE_FLAGS=${KOPS_FEATURE_FLAGS}")
+
+# More time for bigger clusters
+KUBETEST2_ARGS+=("--validation-wait=30m")
 
 # The caller can set DELETE_CLUSTER=false to stop us deleting the cluster
 if [[ -z "${DELETE_CLUSTER:-}" ]]; then
   DELETE_CLUSTER="true"
 fi
 
-kubetest2 kops "${KUBETEST2_ARGS}" \
+kubetest2 kops "${KUBETEST2_ARGS[@]}" \
   --up \
   --kubernetes-version="${K8S_VERSION}" \
   --create-args="${create_args}" \
   --control-plane-size="${KOPS_CONTROL_PLANE_SIZE:-1}"
 
-kubetest2 kops "${KUBETEST2_ARGS}" \
+kubetest2 kops "${KUBETEST2_ARGS[@]}" \
   --test=kops \
   --kubernetes-version="${K8S_VERSION}" \
   -- \
@@ -121,6 +133,6 @@ kubetest2 kops "${KUBETEST2_ARGS}" \
   --focus-regex="\[Conformance\]"
 
 if [[ "${DELETE_CLUSTER:-}" == "true" ]]; then
-  kubetest2 kops "${KUBETEST2_ARGS}" --down
+  kubetest2 kops "${KUBETEST2_ARGS[@]}" --down
   DELETE_CLUSTER=false # Don't delete again in trap
 fi

upup/pkg/fi/cloudup/awsup/aws_cloud.go

@@ -2317,13 +2317,16 @@ func (c *awsCloudImplementation) DefaultInstanceType(cluster *kops.Cluster, ig *
 	igZonesSet := sets.NewString(igZones...)
 
 	// TODO: Validate that instance type exists in all AZs, but skip AZs that don't support any VPC stuff
+	var reasons []string
 	for _, instanceType := range candidates {
 		if strings.HasPrefix(instanceType, "t4g") {
 			if imageArch != "arm64" {
+				reasons = append(reasons, fmt.Sprintf("instance type %q does not match image architecture %q", instanceType, imageArch))
 				continue
 			}
 		} else {
 			if imageArch == "arm64" {
+				reasons = append(reasons, fmt.Sprintf("instance type %q does not match image architecture %q", instanceType, imageArch))
 				continue
 			}
 		}
@@ -2335,10 +2338,16 @@ func (c *awsCloudImplementation) DefaultInstanceType(cluster *kops.Cluster, ig *
 		if zones.IsSuperset(igZonesSet) {
 			return instanceType, nil
 		} else {
+			reasons = append(reasons, fmt.Sprintf("instance type %q is not available in all zones (available in zones %v, need %v)", instanceType, zones, igZones))
 			klog.V(2).Infof("can't use instance type %q, available in zones %v but need %v", instanceType, zones, igZones)
 		}
 	}
 
+	// Log the detailed reasons why we can't find an instance type
+	klog.Warning("cannot find suitable instance type")
+	for _, reason := range reasons {
+		klog.Warning("  *  " + reason)
+	}
 	return "", fmt.Errorf("could not find a suitable supported instance type for the instance group %q (type %q) in region %q", ig.Name, ig.Spec.Role, c.region)
 }