Fix #123

Björn Vonheiden authored 4 years ago and Sören Henning committed 4 years ago

The lag analysis and trend slope computer may fail to various reasons.
But the benchmark should continoue, because this may be only temporary errors.

Check that all the pods for load generators and uc gets deleted.
Otherwise they interfere with kafka and create topics again.

Remove the commit interval option as it is espacially for kafka
streams. Instead use the configurations option, which takes key
value pairs and set these as environment variables for the UC.

Pass more configratioin from the theodolite scrip to run uc script
instead of using defaults.
Use the result path in the analyis.

Make the functions parameterized in order to call the main method
from another python class. Before the methods used a global
variable to access the values.

Use the parameters -res and -reso to reset the kubernetes cluster.
With -reso only the cluster is resetted in nothing else is executed.
With -res the cluster gets resetted before execution of the usecase.

The order of the environment variables is atm crucial for the
run uc py.

Run uc py still used statefullSet instead of deployment

Use the API instead of the kubectl program from the host machine.
This enables better portability of the script to other machines.

Apply the DRY principle and create method to load the methods with
a given path.
Further add comments to methods for better understanding and enable
the logging.

Use the kubernetes api in python to run the kubernetes creation
operations instead of the default platform commands.