scenario.md

Scenarios

The scenario runner - perturb.sh

• Cloud provider - is hardcoded to work with digital ocean and hobbykube
• Node detection - Expects exactly 3 nodes - 1 master and 2 slaves
  • --auto-populate uses doctl and a regexp to find the droplets

Phases

• handle_arguments - Argument parsing and vaildation
• run_scenario - Setup the scenario
• run_test_loop - Run Test loop

Scenario setup

Validation validate_instructions

Before doing anything perturb.sh checks that it can only find bash scripts with well-known names.

Scenario scripts are found by looking for files with a .sh extension under the $SCENARIO_DIR

• worker-any.sh - runs on a randomly chosen slave
• worker-1.sh - runs on slave 1
• worker-2.sh - runs on slave 2
• workers-every.sh - runs on slaves 1 & 2
• nodes-every.sh - runs on master and slaves 1 and 2
• master.sh - runs on master
• test.sh - Ignored for setup

Any other .sh scripts will cause an error

Configure Kubernetes run_kubectl_yaml

1. Loop over subdirectories in $SCENARIO_DIR. Subdirectories must be named after the command to be run by kubectl. This is currently always only apply
2. Concatenate all the *.ya?ml files together into a string
3. Run the concatenated string through ssh/kubectl on the master

Run the validated shell scripts run_scenario

Runs each shell script on the appropriate host (see above)

Cleanup

Special scenarios

• cleanup
• noop

Scripts run always

• no-cleanup.do If this is present scenario/cover-tracks.sh will not run
• reboot-all.do If this is present scenario/reboot.sh will run on all nodes
• reboot-workers.do If this is present scenario/reboot.sh will run on master node
• reboot-master.do If this is present scenario/reboot.sh will run on worker nodes
• temporary script to copy base64 encoded$SCENARIO_DIR/flag.txt to /root/flag.txt if that file exists in $SCENARIO_DIR

Scripts only run for "normal scenarios

• temporary script to copy $SCENARIO_DIR/challenge.txt to /opt/challenge.txt