Add minimal integration tests (#3)

The pipeline currently fails as there's no integration test defined.

We'll add a minimal test that deploys the vsphere chart. The actual
k8s deployment is expcted to enter a crash loop as it expects
an actual vsphere environment, for which reason the assertions
are commented out for the time being.

tests/integration/test_csi_driver.py

@@ -0,0 +1,69 @@
+#
+# Copyright 2024 Canonical, Ltd.
+# See LICENSE file for licensing details
+#
+
+import uuid
+
+import pytest
+from k8s_test_harness import harness
+from k8s_test_harness.util import constants, env_util, k8s_util
+from k8s_test_harness.util.k8s_util import HelmImage
+
+IMG_PLATFORM = "amd64"
+INSTALL_NAME = "vsphere-csi-driver"
+
+
+def _get_rock_image(name: str, version: str):
+    rock = env_util.get_build_meta_info_for_rock_version(name, version, IMG_PLATFORM)
+    return rock.image
+
+
+@pytest.mark.parametrize("version", ["v3.3.1"])
+def test_csi_driver(function_instance: harness.Instance, version: str):
+    rock_image = _get_rock_image("vsphere-csi-driver", version)
+
+    # This helm chart requires the registry to be separated from the image.
+    registry = "docker.io"
+    parts = rock_image.split("/")
+    if len(parts) > 1:
+        registry = parts[0]
+        rock_image = "/".join(parts[1:])
+
+    images = [
+        HelmImage(uri=rock_image, prefix="controller"),
+        HelmImage(uri=rock_image, prefix="node"),
+    ]
+
+    cluster_id = "test-cluster-%s" % uuid.uuid4().fields[0]
+    set_configs = [
+        f"global.config.global.cluster-id={cluster_id}",
+    ]
+    for image in images:
+        set_configs.append(
+            f"{image.prefix}.image.registry={registry}",
+        )
+
+    # Specifying the registry and chart version didn't work, as a workaround
+    # we're passing the chart archive url.
+    chart = "vsphere-csi-3.6.0"
+    helm_command = k8s_util.get_helm_install_command(
+        name=INSTALL_NAME,
+        chart_name=f"https://github.com/vsphere-tmm/helm-charts/releases/download/{chart}/{chart}.tgz",
+        images=images,
+        namespace=constants.K8S_NS_KUBE_SYSTEM,
+        set_configs=set_configs,
+    )
+    function_instance.exec(helm_command)
+
+    # The CSI deployment is expected to enter a crash loop since it requires an
+    # actual vSphere environment. For the time being, the following assertions
+    # are commented out.
+
+    # k8s_util.wait_for_daemonset(
+    #     function_instance, "vsphere-csi-driver", constants.K8S_NS_KUBE_SYSTEM
+    # )
+
+    # k8s_util.wait_for_deployment(
+    #     function_instance, "vsphere-csi-controller", constants.K8S_NS_KUBE_SYSTEM
+    # )