Add env files for Kind-in-Prow integration test jobs

The existing `test/e2e-tests-kind.env` is specifically for the `PipelineRun` approach. These new files are for running the e2e tests, via `kind`, in Prow.

There are four new env files - one for just the go e2e tests each for `stable` and `alpha`, and one for just the yaml tests each for `stable` and `alpha`.

Additionally, `examples/v1beta1/taskruns/git-volume.yaml` is moved to `examples/v1beta1/taskruns/no-ci/git-volume.yaml`. This is because Kind nodes don't have `git` installed, which is necessary for git volumes to work. Also, `--ignore-path=/product_uuid` has been added to the Kaniko args in `examples/v1beta1/pipelineruns/pipelinerun.yaml` and `test/yamls/v1beta1/pipelineruns/pipelinerun.yaml` to work around an issue with Kaniko multi-stage builds on Kind (GoogleContainerTools/kaniko#2164).

Signed-off-by: Andrew Bayer <andrew.bayer@gmail.com>

examples/v1beta1/pipelineruns/pipelinerun.yaml

@@ -142,7 +142,7 @@ spec:
     default: ""
   - name: BUILDER_IMAGE
     description: The image on which builds will run
-    default: gcr.io/kaniko-project/executor:v1.3.0
+    default: gcr.io/kaniko-project/executor:v1.8.1
   - name: baseImage
     description: Base image for GoogleContainerTools/skaffold microservice apps
     default: BASE=alpine:3.9
@@ -166,6 +166,7 @@ spec:
     - --destination=$(params.IMAGE)
     - --oci-layout-path=$(workspaces.source.path)/$(params.CONTEXT)/image-digest
     - --build-arg=$(inputs.params.baseImage)
+    - --ignore-path=/product_uuid # TODO(abayer): Work around Kaniko multi-stage build issues on Kind: https://github.com/GoogleContainerTools/kaniko/issues/2164
     # kaniko assumes it is running as root, which means this example fails on platforms
     # that default to run containers as random uid (like OpenShift). Adding this securityContext
     # makes it explicit that it needs to run as root.

test/e2e-common.sh

@@ -20,8 +20,9 @@ source $(git rev-parse --show-toplevel)/vendor/github.com/tektoncd/plumbing/scri
 
 function install_pipeline_crd() {
   echo ">> Deploying Tekton Pipelines"
-  ko resolve -R -f config/ \
-      | sed -e 's%"level": "info"%"level": "debug"%' \
+  local ko_target="$(mktemp)"
+  ko resolve -R -f config/ > "${ko_target}" || fail_test "Pipeline image resolve failed"
+  cat "${ko_target}" | sed -e 's%"level": "info"%"level": "debug"%' \
       | sed -e 's%loglevel.controller: "info"%loglevel.controller: "debug"%' \
       | sed -e 's%loglevel.webhook: "info"%loglevel.webhook: "debug"%' \
       | kubectl apply -R -f - || fail_test "Build pipeline installation failed"

test/e2e-tests-kind-prow-alpha.env

@@ -0,0 +1,6 @@
+SKIP_INITIALIZE=true
+PIPELINE_FEATURE_GATE=alpha
+EMBEDDED_STATUS_GATE=minimal
+RUN_YAML_TESTS=true
+KO_DOCKER_REPO=registry.local:5000
+E2E_GO_TEST_TIMEOUT=40m

test/e2e-tests-kind-prow.env

@@ -0,0 +1,5 @@
+SKIP_INITIALIZE=true
+PIPELINE_FEATURE_GATE=stable
+RUN_YAML_TESTS=true
+KO_DOCKER_REPO=registry.local:5000
+E2E_GO_TEST_TIMEOUT=40m

test/e2e-tests.sh

@@ -25,6 +25,8 @@ PIPELINE_FEATURE_GATE=${PIPELINE_FEATURE_GATE:-stable}
 EMBEDDED_STATUS_GATE=${EMBEDDED_STATUS_GATE:-full}
 SKIP_INITIALIZE=${SKIP_INITIALIZE:="false"}
 RUN_YAML_TESTS=${RUN_YAML_TESTS:="true"}
+SKIP_GO_E2E_TESTS=${SKIP_GO_E2E_TESTS:="false"}
+E2E_GO_TEST_TIMEOUT=${E2E_GO_TEST_TIMEOUT:="20m"}
 failed=0
 
 # Script entry point.
@@ -66,13 +68,16 @@ function set_embedded_status() {
 function run_e2e() {
   # Run the integration tests
   header "Running Go e2e tests"
-  go_test_e2e -timeout=20m ./test/... || failed=1
+  # Skip ./test/*.go tests if SKIP_GO_E2E_TESTS == true
+  if [ "${SKIP_GO_E2E_TESTS}" != "true" ]; then
+    go_test_e2e -timeout=${E2E_GO_TEST_TIMEOUT} ./test/... || failed=1
+  fi
 
   # Run these _after_ the integration tests b/c they don't quite work all the way
   # and they cause a lot of noise in the logs, making it harder to debug integration
   # test failures.
   if [ "${RUN_YAML_TESTS}" == "true" ]; then
-    go_test_e2e -parallel=4 -mod=readonly -tags=examples -timeout=20m ./test/ || failed=1
+    go_test_e2e -mod=readonly -tags=examples -timeout=${E2E_GO_TEST_TIMEOUT} ./test/ || failed=1
   fi
 }
 

test/yamls/v1beta1/pipelineruns/pipelinerun.yaml

@@ -142,7 +142,7 @@ spec:
     default: ""
   - name: BUILDER_IMAGE
     description: The image on which builds will run
-    default: gcr.io/kaniko-project/executor:v1.3.0
+    default: gcr.io/kaniko-project/executor:v1.8.1
   - name: baseImage
     description: Base image for GoogleContainerTools/skaffold microservice apps
     default: BASE=alpine:3.9
@@ -166,13 +166,16 @@ spec:
     - --destination=$(params.IMAGE)
     - --oci-layout-path=$(workspaces.source.path)/$(params.CONTEXT)/image-digest
     - --build-arg=$(inputs.params.baseImage)
+    - --ignore-path=/product_uuid # TODO(abayer): Work around Kaniko multi-stage build issues on Kind: https://github.com/GoogleContainerTools/kaniko/issues/2164
     # kaniko assumes it is running as root, which means this example fails on platforms
     # that default to run containers as random uid (like OpenShift). Adding this securityContext
     # makes it explicit that it needs to run as root.
     securityContext:
       runAsUser: 0
   - name: write-digest
     workingDir: $(workspaces.source.path)
+    # TODO(abayer): Using ko:// means we end up rebuilding the image, which can be annoying. Consider alternatives while ensuring
+    # we're always replacing the ko:// images with ones already built from our source...
     image: ko://github.com/tektoncd/pipeline/cmd/imagedigestexporter
     # output of imagedigestexport [{"name":"image","digest":"sha256:eed29..660"}]
     command: ["/ko-app/imagedigestexporter"]