WIP: [Slice] Test E2E quick rebuild. (#593)

[Slice] Test E2E quick rebuild.

Author: mbobrovskyi

slice/Makefile

@@ -64,6 +64,7 @@ E2E_K8S_FULL_VERSION ?= $(filter $(E2E_K8S_VERSION).%,$(E2E_K8S_VERSIONS))
 E2E_K8S_FULL_VERSION := $(or $(E2E_K8S_FULL_VERSION),$(E2E_K8S_VERSION).0)
 E2E_KIND_VERSION ?= kindest/node:v$(E2E_K8S_FULL_VERSION)
 E2E_RUN_ONLY_ENV ?= false
+E2E_RUN_ONLY_SLICE ?= false
 
 .PHONY: all
 all: build
@@ -122,7 +123,7 @@ run-test-e2e-%:
 	E2E_KIND_VERSION="kindest/node:v$(K8S_VERSION)" KIND_CLUSTER_NAME=$(KIND_CLUSTER_NAME) CREATE_KIND_CLUSTER=$(CREATE_KIND_CLUSTER) \
 		EXTERNAL_CRDS_DIR=$(EXTERNAL_CRDS_DIR) ARTIFACTS="$(ARTIFACTS)/$@" IMAGE_TAG=$(IMAGE_TAG) GINKGO_ARGS="$(GINKGO_ARGS)" \
 		KUEUE_IMAGE=$(KUEUE_IMAGE) JOBSET_IMAGE=$(JOBSET_IMAGE) \
-		TEST_LOG_LEVEL=$(TEST_LOG_LEVEL) E2E_RUN_ONLY_ENV=$(E2E_RUN_ONLY_ENV) \
+		TEST_LOG_LEVEL=$(TEST_LOG_LEVEL) E2E_RUN_ONLY_ENV=$(E2E_RUN_ONLY_ENV) E2E_RUN_ONLY_SLICE=$(E2E_RUN_ONLY_SLICE) \
 		./hack/e2e-test.sh
 
 .PHONY: lint

slice/README.md

@@ -117,6 +117,29 @@ is manually re-applied afterwards.
 
 More information can be found via the [Kubebuilder Documentation](https://book.kubebuilder.io/introduction.html)
 
+## Running e2e tests using custom build
+```shell
+make kind-image-build test-e2e
+```
+
+## Attaching e2e tests to an existing kind cluster
+You can use the following approach to start up a kind cluster and then run e2e tests from commandline or VSCode,
+attaching them to the existing cluster. For example, suppose you want to test some of the multikueue-e2e tests.
+
+Run `make kind-image-build test-e2e E2E_RUN_ONLY_ENV=true` and wait for the `Do you want to cleanup? [Y/n] ` to appear.
+
+The cluster is ready, and now you can run tests from another terminal:
+```shell
+./bin/ginkgo -focus "JobSet (when )?Creating a JobSet it should create Slice based on created Workload with TPU topology 4x4x4, TPU topology 4 and parallelism 16" -v ./test/e2e/...
+```
+or from VSCode.
+
+## Quick rebuild Kueue in e2e tests
+If the `E2E_RUN_ONLY_ENV` variable is set, you can rebuild **only the Kueue image** without needing to rebuild the entire cluster.
+```shell
+make kind-image-build test-e2e E2E_RUN_ONLY_ENV=true E2E_RUN_ONLY_SLICE=true
+```
+
 ## License
 
 Copyright The Kubernetes Authors.

slice/hack/e2e-common.sh

@@ -111,6 +111,10 @@ function cluster_slice_deploy {
     local build_output
     build_output=$($KUSTOMIZE build "${ROOT_DIR}/config/dev")
     build_output="${build_output//$DEFAULT_SLICE_NAMESPACE/$SLICE_NAMESPACE}"
+    # Add rollout-timestamp annotation to the Deployment's pod template
+    local timestamp
+    timestamp=$(date +%s) # Unix timestamp for uniqueness
+    build_output=$(yq eval "select(.kind == \"Deployment\").spec.template.metadata.annotations += {\"kubectl.kubernetes.io/timestamp\": \"$timestamp\"}" - <<< "$build_output")
     echo "$build_output" | kubectl apply --kubeconfig="$1" --server-side -f -
 
     (cd "${ROOT_DIR}/config/manager" && $KUSTOMIZE edit set image controller="$initial_image")

slice/hack/e2e-test.sh

@@ -44,21 +44,34 @@ function startup {
     fi
 }
 
-trap cleanup EXIT
-startup
-prepare_docker_images
+if [[ "${E2E_RUN_ONLY_SLICE}" != 'true' ]]; then
+  trap cleanup EXIT
+  startup
+  prepare_docker_images
+fi
+
 kind_load "$KIND_CLUSTER_NAME" ""
 
-# We need to wait for Kueue to become available;
-# otherwise, we encounter the error: "no endpoints available for service 'kueue-webhook-service'"
-echo "Waiting for Kueue to become available..."
-kubectl wait deploy/kueue-controller-manager -nkueue-system --for=condition=available --timeout=5m
+if [[ "${E2E_RUN_ONLY_SLICE}" != 'true' ]]; then
+  # We need to wait for Kueue to become available;
+  # otherwise, we encounter the error: "no endpoints available for service 'kueue-webhook-service'"
+  echo "Waiting for Kueue to become available..."
+  kubectl wait deploy/kueue-controller-manager -nkueue-system --for=condition=available --timeout=5m
+fi
 
 cluster_slice_deploy ""
 
-if [ "$E2E_RUN_ONLY_ENV" == 'true' ]; then
-  read -rp "Press Enter to cleanup."
-else
-  # shellcheck disable=SC2086
-  $GINKGO $GINKGO_ARGS --junit-report=junit.xml --json-report=e2e.json --output-dir="$ARTIFACTS" -v ./test/e2e/...
+if [ "$E2E_RUN_ONLY_ENV" = "true" ]; then
+  if [[ "${E2E_RUN_ONLY_SLICE}" != 'true' ]]; then
+    read -rp "Do you want to cleanup? [Y/n] " reply
+    if [[ "$reply" =~ ^[nN]$ ]]; then
+      trap - EXIT
+      echo "Skipping cleanup for kind cluster."
+      echo -e "\nKind cluster cleanup:\n  kind delete cluster --name $KIND_CLUSTER_NAME"
+    fi
+  fi
+  exit 0
 fi
+
+# shellcheck disable=SC2086
+$GINKGO $GINKGO_ARGS --junit-report=junit.xml --json-report=e2e.json --output-dir="$ARTIFACTS" -v ./test/e2e/...