serial/

README

e2e serial testsuite

The "serial" e2e test suite holds e2e tests covering topology-aware scheduling as a whole.

For this purposes "end to end" means the cluster, taken as black box, with minimal breakages to this rule only iff and when these breakages are extremely helpful (point in case: the numacell plugin). All these cases are vetted anyway, and the goal is to gradually minimize or eliminate them all.

Because of the fact that "e2e" is the cluster here, the suite expects to run against an already configured and TAS-enabled cluster, no matter how (using the numaresources-operator proper or any other vehicle). We intentionally avoid any explicit dependency (e.g. against the "install" test suite for this reason).

All the tests in this suite create, manipulate, destroy cluster resources. They are intrinsically disruptive and want to exercise access to exclusive resources. Hence the "serial" name: the tests expect to run in isolation with any other workload, and in isolation to each other.

IOW, the suite expects to run against a pristine, unloaded cluster to validate it.

It is responsibility of the tests in this suite to clean up properly and to restore the cluster state as they found it before they run.

configuring using the environment variables

• E2E_NROP_INSTALL_SKIP_KC (accepts boolean, e.g. true) instructs the suite to NOT deploy the builtin kubeletconfig. With this enabled, the suite will expect a kubeletconfig already deployed (and consumed) in the cluster against it is running.
• E2E_NROP_MCP_UPDATE_TIMEOUT (accepts a duration, e.g. 5m) instructs the suite how much time it should wait for the MCP updates to take place before to exit with error.
• E2E_NROP_MCP_UPDATE_INTERVAL (accepts a duration, e.g. 10s) instructs the suite about how much it should wait between checks for MCP updates.
• E2E_NROP_PLATFORM (accepts a string, e.g. Kubernetes, OpenShift) instructs the suite to disable the autodetection of the platform and force it to the provided value.
• E2E_NROP_DUMP_EVENTS (accepts boolean, e.g. true) requests the suite to dump events pertaining to pods failed unexpectedly on standard output, alongside (not replacing) the logging of the said events.
• E2E_NROP_TARGET_NODE (accepts string, e.g. node-0.my-cluster.io) instructs the suite to always pick the node matching the given name when a random node is needed.
• E2E_NROP_TEST_COOLDOWN (accepts string expressing time unit, e.g. 30s) instructs the suite to wait for the specified amount of time after each spec, to give time to the cluster to settle up.
• E2E_NROP_TEST_TEARDOWN (accepts string expressing time unit, e.g. 30s) instructs the suite to wait up to the specified amount of time while tearing down the resources needed by each spec.
• E2E_NROP_DEVICE_TYPE_1 (accepts string, e.g example.com/deviceA) declares name of a device type that exists on the cluster and will be used as a resource of type device-type-1 in the tests.
• E2E_NROP_DEVICE_TYPE_2 (accepts string, e.g example.com/deviceB) declares name of a device type that exists on the cluster and will be used as a resource of type device-type-2 in the tests.
• E2E_NROP_DEVICE_TYPE_3 (accepts string, e.g example.com/deviceC) declares name of a device type that exists on the cluster and will be used as a resource of type device-type-3 in the tests.
• E2E_RTE_CI_IMAGE (accepts string, e.g quay.io/openshift-kni/resource-topology-exporter:test-ci) sets the RTE image to be used for testing purposes particularly for modifying the operator object.