[CI] Add the integration tests to Gitlab CI (#29489)

.gitlab/integration_test/linux.yml

@@ -0,0 +1,25 @@
+.integration_tests_deb:
+  stage: integration_test
+  needs: ["go_deps", "go_tools_deps"]
+  image: 486234852809.dkr.ecr.us-east-1.amazonaws.com/ci/datadog-agent-buildimages/deb_x64$DATADOG_AGENT_BUILDIMAGES_SUFFIX:$DATADOG_AGENT_BUILDIMAGES
+  tags: ["docker-in-docker:amd64"]
+  before_script:
+    - !reference [.retrieve_linux_go_deps]
+    - !reference [.retrieve_linux_go_tools_deps]
+
+agent_integration_tests:
+  extends: .integration_tests_deb
+  # We have capacity issues on the docker-in-docker:amd64 runners for now.
+  # Setting allow_failure to true to avoid blocking the pipelines.
+  # Setting the timeout to 30 min and retry to 2 to avoid the job from pending for too long.
+  allow_failure: true
+  timeout: 30m
+  retry: 2
+  script:
+    - inv -e integration-tests --race --remote-docker
+
+docker_integration_tests:
+  extends: .integration_tests_deb
+  script:
+    - inv -e docker.test
+    - inv -e docker.integration-tests

tasks/agent.py

@@ -17,7 +17,6 @@
 from tasks.build_tags import filter_incompatible_tags, get_build_tags, get_default_build_tags
 from tasks.devcontainer import run_on_devcontainer
 from tasks.flavor import AgentFlavor
-from tasks.go import deps
 from tasks.libs.common.utils import (
     REPO_PATH,
     bin_name,
@@ -593,29 +592,28 @@ def hacky_dev_image_build(
 
 
 @task
-def integration_tests(ctx, install_deps=False, race=False, remote_docker=False, go_mod="mod"):
+def integration_tests(ctx, race=False, remote_docker=False, go_mod="mod", timeout=""):
     """
     Run integration tests for the Agent
     """
-    if install_deps:
-        deps(ctx)
 
     if sys.platform == 'win32':
-        return _windows_integration_tests(ctx, race=race, go_mod=go_mod)
+        return _windows_integration_tests(ctx, race=race, go_mod=go_mod, timeout=timeout)
     else:
         # TODO: See if these will function on Windows
-        return _linux_integration_tests(ctx, race=race, remote_docker=remote_docker, go_mod=go_mod)
+        return _linux_integration_tests(ctx, race=race, remote_docker=remote_docker, go_mod=go_mod, timeout=timeout)
 
 
-def _windows_integration_tests(ctx, race=False, go_mod="mod"):
+def _windows_integration_tests(ctx, race=False, go_mod="mod", timeout=""):
     test_args = {
         "go_mod": go_mod,
         "go_build_tags": " ".join(get_default_build_tags(build="test")),
         "race_opt": "-race" if race else "",
         "exec_opts": "",
+        "timeout_opt": f"-timeout {timeout}" if timeout else "",
     }
 
-    go_cmd = 'go test -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'.format(**test_args)  # noqa: FS002
+    go_cmd = 'go test {timeout_opt} -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'.format(**test_args)  # noqa: FS002
 
     tests = [
         {
@@ -644,12 +642,13 @@ def _windows_integration_tests(ctx, race=False, go_mod="mod"):
             ctx.run(f"{go_cmd} {test['prefix']} {test['extra_args']}")
 
 
-def _linux_integration_tests(ctx, race=False, remote_docker=False, go_mod="mod"):
+def _linux_integration_tests(ctx, race=False, remote_docker=False, go_mod="mod", timeout=""):
     test_args = {
         "go_mod": go_mod,
         "go_build_tags": " ".join(get_default_build_tags(build="test")),
         "race_opt": "-race" if race else "",
         "exec_opts": "",
+        "timeout_opt": f"-timeout {timeout}" if timeout else "",
     }
 
     # since Go 1.13, the -exec flag of go test could add some parameters such as -test.timeout
@@ -659,7 +658,7 @@ def _linux_integration_tests(ctx, race=False, remote_docker=False, go_mod="mod")
     if remote_docker:
         test_args["exec_opts"] = f"-exec \"{os.getcwd()}/test/integration/dockerize_tests.sh\""
 
-    go_cmd = 'go test -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'.format(**test_args)  # noqa: FS002
+    go_cmd = 'go test {timeout_opt} -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'.format(**test_args)  # noqa: FS002
 
     prefixes = [
         "./test/integration/config_providers/...",

tasks/cluster_agent.py

@@ -15,7 +15,6 @@
 from tasks.build_tags import get_build_tags, get_default_build_tags
 from tasks.cluster_agent_helpers import build_common, clean_common, refresh_assets_common, version_common
 from tasks.cws_instrumentation import BIN_PATH as CWS_INSTRUMENTATION_BIN_PATH
-from tasks.go import deps
 from tasks.libs.releasing.version import load_release_versions
 
 # constants
@@ -88,22 +87,20 @@ def clean(ctx):
 
 
 @task
-def integration_tests(ctx, install_deps=False, race=False, remote_docker=False, go_mod="mod"):
+def integration_tests(ctx, race=False, remote_docker=False, go_mod="mod", timeout=""):
     """
     Run integration tests for cluster-agent
     """
     if sys.platform == 'win32':
         raise Exit(message='cluster-agent integration tests are not supported on Windows', code=0)
 
-    if install_deps:
-        deps(ctx)
-
     # We need docker for the kubeapiserver integration tests
     tags = get_default_build_tags(build="cluster-agent") + ["docker", "test"]
 
     go_build_tags = " ".join(get_build_tags(tags, []))
     race_opt = "-race" if race else ""
     exec_opts = ""
+    timeout_opt = f"-timeout {timeout}" if timeout else ""
 
     # since Go 1.13, the -exec flag of go test could add some parameters such as -test.timeout
     # to the call, we don't want them because while calling invoke below, invoke
@@ -112,7 +109,7 @@ def integration_tests(ctx, install_deps=False, race=False, remote_docker=False,
     if remote_docker:
         exec_opts = f"-exec \"{os.getcwd()}/test/integration/dockerize_tests.sh\""
 
-    go_cmd = f'go test -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'
+    go_cmd = f'go test {timeout_opt} -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'
 
     prefixes = [
         "./test/integration/util/kube_apiserver",

tasks/cluster_agent_cloudfoundry.py

@@ -44,7 +44,7 @@ def refresh_assets(ctx, development=True):
 
 
 @task
-def integration_tests(ctx, install_deps=False, race=False, remote_docker=False):  # noqa: U100
+def integration_tests(ctx, race=False, remote_docker=False):  # noqa: U100
     """
     Run integration tests for cluster-agent-cloudfoundry
     """

tasks/dogstatsd.py

@@ -11,7 +11,6 @@
 
 from tasks.build_tags import filter_incompatible_tags, get_build_tags, get_default_build_tags
 from tasks.flavor import AgentFlavor
-from tasks.go import deps
 from tasks.libs.common.utils import REPO_PATH, bin_name, get_build_flags, get_root
 from tasks.windows_resources import build_messagetable, build_rc, versioninfo_vars
 
@@ -171,19 +170,17 @@ def size_test(ctx, skip_build=False):
 
 
 @task
-def integration_tests(ctx, install_deps=False, race=False, remote_docker=False, go_mod="mod"):
+def integration_tests(ctx, race=False, remote_docker=False, go_mod="mod", timeout=""):
     """
     Run integration tests for dogstatsd
     """
     if sys.platform == 'win32':
         raise Exit(message='dogstatsd integration tests are not supported on Windows', code=0)
 
-    if install_deps:
-        deps(ctx)
-
     go_build_tags = " ".join(get_default_build_tags(build="test"))
     race_opt = "-race" if race else ""
     exec_opts = ""
+    timeout_opt = f"-timeout {timeout}" if timeout else ""
 
     # since Go 1.13, the -exec flag of go test could add some parameters such as -test.timeout
     # to the call, we don't want them because while calling invoke below, invoke
@@ -192,7 +189,7 @@ def integration_tests(ctx, install_deps=False, race=False, remote_docker=False,
     if remote_docker:
         exec_opts = f"-exec \"{os.getcwd()}/test/integration/dockerize_tests.sh\""
 
-    go_cmd = f'go test -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'
+    go_cmd = f'go test {timeout_opt} -mod={go_mod} {race_opt} -tags "{go_build_tags}" {exec_opts}'
 
     prefixes = [
         "./test/integration/dogstatsd/...",

tasks/gotest.py

@@ -395,15 +395,15 @@ def test(
 
 
 @task
-def integration_tests(ctx, install_deps=False, race=False, remote_docker=False, debug=False):
+def integration_tests(ctx, race=False, remote_docker=False, debug=False, timeout=""):
     """
     Run all the available integration tests
     """
     tests = [
-        lambda: agent_integration_tests(ctx, install_deps, race, remote_docker),
-        lambda: dsd_integration_tests(ctx, install_deps, race, remote_docker),
-        lambda: dca_integration_tests(ctx, install_deps, race, remote_docker),
-        lambda: trace_integration_tests(ctx, install_deps, race),
+        lambda: agent_integration_tests(ctx, race=race, remote_docker=remote_docker, timeout=timeout),
+        lambda: dsd_integration_tests(ctx, race=race, remote_docker=remote_docker, timeout=timeout),
+        lambda: dca_integration_tests(ctx, race=race, remote_docker=remote_docker, timeout=timeout),
+        lambda: trace_integration_tests(ctx, race=race, timeout=timeout),
     ]
     for t in tests:
         try:

tasks/trace_agent.py

@@ -6,7 +6,6 @@
 from tasks.agent import build as agent_build
 from tasks.build_tags import filter_incompatible_tags, get_build_tags, get_default_build_tags
 from tasks.flavor import AgentFlavor
-from tasks.go import deps
 from tasks.libs.common.utils import REPO_PATH, bin_name, get_build_flags
 from tasks.windows_resources import build_messagetable, build_rc, versioninfo_vars
 
@@ -92,20 +91,18 @@ def build(
 
 
 @task
-def integration_tests(ctx, install_deps=False, race=False, go_mod="mod"):
+def integration_tests(ctx, race=False, go_mod="mod", timeout="10m"):
     """
     Run integration tests for trace agent
     """
     if sys.platform == 'win32':
         raise Exit(message='trace-agent integration tests are not supported on Windows', code=0)
 
-    if install_deps:
-        deps(ctx)
-
     go_build_tags = " ".join(get_default_build_tags(build="test"))
     race_opt = "-race" if race else ""
+    timeout_opt = f"-timeout {timeout}" if timeout else ""
 
-    go_cmd = f'go test -mod={go_mod} {race_opt} -v -tags "{go_build_tags}"'
+    go_cmd = f'go test {timeout_opt} -mod={go_mod} {race_opt} -v -tags "{go_build_tags}"'
     ctx.run(f"{go_cmd} ./cmd/trace-agent/test/testsuite/...", env={"INTEGRATION": "yes"})