Experiment execution page

How to view and manage experiment runs

When you access a project and select an experiment definition, you can list all the runs carried out on that definition and the status of those in progress. The page lists the executions in chronological order (from the most recent to the oldest), as shown in the figure below:

Currently, each run allows the following actions:

• Monitor in real time (only available for experiment executions in progress)

• Status monitoring: A progress bar allows the experimenter to track the current state of the run and estimate the remaining time based on the total duration of the experiment.

• Currently, The last 10% of the experiment's execution time is allocated to saving data. The time required for this step can vary depending on the amount of data being saved.

• Delete runs: This action removes the run from the table. While it is not guaranteed, the results may still be saved in the system and can potentially be retrieved by contacting system administrators.

• Download logs: This action downloads a .tar.gz file containing logs and information related to the experiment, as well as the persistent results defined in the experiment definitions via the persistent_volume_path key for each set of simulated nodes. Among the standard information found in the file are:

• experiment_logs_X.tar: File containing the logs of the containers run on the Kubernetes cluster, which include the STDOUT and STDERR of each container. Within this file you can obtain a txt file with the log of each container.

• MENTORED_IP_LIST.[json|yaml]: File that lists the IP assigned to each node in the experiment at the start of the run.

• MENTORED_READY.txt: File used to manage the experiment's WARMUP and which contains the timestamp at which the experiment was actually started.

• <Node-Actor>/<Node-Actor>-X_container.tar: Directory containing the files saved for each container in the experiment that has had its persistent_volume_path defined.

Example of the structure of an experiment's logs:

Monitoring experiments in real time

To perform real-time monitoring, the user can monitor the attack in real time using the Kubectl web tool, with more detailed monitoring available in Real-Time Monitoring.

Checking experiment status

Additionally, you can check the status of the experiment by referring to the colors displayed in the status column:

• Gray indicates that the experiment has not yet been run or encountered an error before starting.
• Yellow indicates that the experiment is currently running.
• Red indicates that the experiment was run, but an error occurred during execution.
• Green indicates that the experiment was completed successfully, without errors.