Maratona-Runtime

Maratona Runtime implements the core service of judging solutions for ICPC and CTF-related coding competitions.

Architecture

We are using a microservice architecture. The system, for now, is separated in the following parts:

• Verdict: Receives information regarding problem and submission details.
• Compiler: Tasked with compiling a received source code and responding with the generated binary code (except for interpreted languages, such as Python).
• Executor: Executes a binary code or a interpreted language program against a set of inputs and responds with generated outputs.
• Orchestrator: Receives challenge submissions and is responsible for triggering their evaluation.
• ORM: Responsible for administering a PostgreSQL server and saving submissions and their verdicts.

Follows a visual representation of the communication between these services:

For now, the system can be started using the docker-compose and Kubernetes.

Using docker-compose

The communication between the services is done via HTTP, using two docker networks:

• "maratona-net" supports general purpose communication between the Verdict, Compiler, Executor, Orchestrator and ORM.
• "database-net", on the other hand, only supports database-related communication, between the ORM and the Postgres database.

Using Kubernetes

If you want to use your local docker images, you might have to configure minikube to use the local Docker environment. To do that run:

eval $(minikube -p minikube docker-env)

Be sure to execute this on every shell you're using!

After this change the image names used on k8s/ files, removing the mruntime/ prefix.

You'll need to rebuild any pre-existing images to make them accessible from the cluster after this step. To build the images, run docker-compose build.

Another possibility is pulling the docker images from our DockerHub registry.

Then, to deploy the project run:

minikube start --cpus 4 --memory=8192 --vm=true
minikube addons enable ingress
kubectl apply -f k8s/
kubectl get ingress 

Add on your /etc/hosts the ip provided on the last command above for the route mart-route.

Troubleshooting

Error from server (InternalError): error when creating "k8s/ingress.yml": Internal error occurred: failed calling webhook "validate.nginx.ingress.kubernetes.io": an error on the server ("") has prevented the request from succeeding

Run:

kubectl delete validatingwebhookconfigurations ingress-nginx-admission

Submissions Database

We are using a PostgreSQL database to store challenges and submissions details.

Testing

All tests were developed using Go's "testing" package.

To run all tests on the project:

go test ./...

For service-specific tests, search for "*_test.go" files, such as: "verdict_test.go". One can execute those by running go test on the folder where they are located.

Coverage

To keep track of coverage information this project uses codecov.io.

To check the test coverage locally one can use:

export CGO_ENABLED=0
go test ./... -coverprofile=coverage.txt -covermode=atomic
bash <(curl -s https://codecov.io/bash) -t $CODECOV_TOKEN

These commands should output a codecov URL with the coverage report.

Travis

All configuration is on the project's travis yaml file.

Travis currently tests:

• If the project files are styled respecting official golang style guidelines (by using gofmt).
• If the project tests are running successfully

Travis also uploads the updated code coverage report to codecov.io.

Tasks and Organization

The developers organized the development tasks on a Trello Kanban (in portuguese).