Load Testing helper scripts for JupyterHubs

We use helm to spawn the tests!

1. Make sure you have helm installed and configured.

2. Prepare a configuration file. We use YAML, and you can find all the possible options in loadtest/values.yaml. At a minimum, you require:

   hub:
     url: <full-url-to-your-hub>

3. Install the chart:

   helm upgrade --install --wait --namespace=<test-run-name> <test-run-name> loadtest -f config.yaml

   Where <test-run-name> is just a unique name you can use to identify this particular run. Make sure you use a unique namespace and release name for each run!

4. You can look at the events being generated in many ways:

   a. Tailing the logs of the individual load test pods with kubectl logs. This is just the stderr and stdout of the process, which is mingled JSON events + logs. b. Tail the logs of the collector, which is aggregating just events. This is in the format of fluent-bit's out_file. Some logging is also co-mingled here. c. Copy the aggregated logs out of the collector pod. You can do this with a kubectl cp command, like:

   kubectl --namespace=<test-run-name> cp  $(kubectl --namespace=<test-run-name> get pod -l component=collector -o name | sed 's:pods/::'):/srv/events.log events.log

   This should copy just the events into events.log in your current directory. They're in the same out_file format, from which JSON can be easily extracted. There should be no other logs co-mingled here, so this is the best method for further analysis.

You can delete the helm release easily with:

helm delete --purge <test-run-name>

Make sure to delete all the pods spawned by the hub before starting another test!