e2e

End-to-end testing

The e2e tests are designed to be run in a number of ways. If you are developing CAPMVM, continue reading this doc. If you are testing the whole of Liquid Metal, check out the Liquid Metal Acceptance Tests docs.

Start flintlock

You will need a running flintlock server. This can be done locally, if you are working on Linux. Mac or windows users have the option to run flintlock on an Equinix host, but if you do not have an account there you will not be able to run these tests.

git clone https://github.com/liquidmetal-dev/flintlock
cd flintlock
sudo ./hack/scripts/provision.sh --grpc-address 0.0.0.0:9090 --dev --insecure

This will clone flintlock and bootstrap your machine to run a server. You can read the [flintlock docs][fl-docs] if you would like to set this up manually and see each individual step. Make sure you start flintlock with --grpc-address set to 0.0.0.0:9090 otherwise CAPMVM will not be able to reach it from within the kind network.

If you ran the above command, flintlock will be running as a systemd service.

DHCP

When microvms are created they will request an IP from a DHCP server. Your router should have one, but you may need to check the settings or start a new server for the purpose of the tests.

TODO: explain this

Required tools

Ensure you have the following installed:

• kind
• docker
• kubectl
• clusterctl

Run the tests

In your CAPMVM repo, save your flintlock server address to a variable and run the tests:

# your flintlock server should be bound to your machine's private address
# you can get that with hostname
FL=$(hostname -I | awk '{print $1}')
make e2e E2E_ARGS="-e2e.flintlock-hosts $FL:9090" # don't forget the port!

Note: the tests will default to using 192.168.1.15 for the workload cluster's load balancer IP. You will need to verify that this is within your network and unused, or configure the tests to use another by setting the -e2e.capmvm.vip-address flag. See "Test options" below for details.

The tests will take ~5 mins to run.

They will:

• Create a new kind cluster
• Init the cluster with CAPI providers
• Generate a CAPMVM workload cluster
• Apply the cluster
• Wait for the control plane and the worker nodes to start
• Verify that all given flintlock hosts were used (if you can start a second flintlock server on a different port, you can pass both to the tests with make e2e E2E_ARGS="-e2e.flintlock-hosts 1.2.3.4:9090,4.5.6.7:9091")
• Deploy nginx to the workload cluster
• Ensure that nginx starts successfully
• Delete the workload cluster
• Delete the kind cluster

To speed up your testing cycle, you can pass the -e2e.existing-cluster flag. See "Test options" below for details.

Test options

The following flags are available:

Usage of /home/claudia/workspace/cluster-api-provider-microvm/test/e2e/e2e.test:
  -e2e.artefact-dir string
        Location to store test yamls, logs, etc. (default "/home/claudia/workspace/cluster-api-provider-microvm/test/e2e/_artefacts")
  -e2e.capmvm.kubernetes-version string
        Version of k8s to run in the workload cluster(s) (default "1.21.8")
  -e2e.capmvm.vip-address string
        Address for the kubevip load balancer. (default "192.168.1.25")
  -e2e.config string
        Path to e2e config for this suite. (default "config/e2e_conf.yaml")
  -e2e.existing-cluster
        If true, uses the current context for the management cluster and will not create a new one.
  -e2e.flintlock-hosts value
        Comma separated list of addresses to flintlock servers. eg '1.2.3.4:9090,5.6.7.8:9090'
  -e2e.skip-cleanup
        Do not delete test-created workload clusters or the management kind cluster.

These can be passed to the make command:

make e2e E2E_ARGS="-e2e.skip-cleanup"

To use the e2e.existing-cluster boolean flag, you will need to ensure that the cluster is set as the current context.

Note: -e2e.flintlock-hosts and -e2e.artefact-dir are already passed to the tests as part of the make command.