A server side application to process requests to run code submitted by the user thorugh th OJ frontend.
cd OJExec-Python
source setup-worker.sh
Edit the .env file with necessary details and restart the executor worker by the following command:
sudo supervisorctl restart executor
The sandbox environment has been submoduled to the original repo so the changes immediately reflect back here. To work with submodules:
1. git submodule init
2. git submodule update
3. cd safeexec
4. cmake .
5. make
From next time onwards, we need to check if the submodules have been updated, to do that:
git pull --recurse-submodules
To setup docker swarm, first install docker to the server using sudo apt install docker.io on each worker and manager server and also install docker-compose fom here on the manager server.
Docker swarm requires an image to be deployed so we will need to keep a temporary storage to store them, which we fulfill using registry.
To run registry first we need to run the command:
sudo docker service create --name registry --publish published=5000,target=5000 registry:2
and make sure you expose port 5000 of the manager node else the worker nodes wont be able to access it.
To build the image use
sudo docker-compose -f docker-compose.prod.yml build
and to push it use sudo docker-compose -f docker-compose.prod.yml push.
On manager node:
docker swarm init
Copy the provided command and type that to worker nodes to join the swarm.
To start the service using the image we uploaded:
sudo docker stack deploy -c docker-compose.prod.yml executor
Note: The deploy configurantion contains the default number of replicase which in this case is 3 and is just taken as an example any number can be put there, but keep in mind the threshold of your server. We have assumed for safety purposes to allocate around 500 MB for each worker, so in a 2 GB RAM server you should have at max 4 workers running.
Scaling
docker service scale executor_executor=5
if you want the direct command but since we are using docker stack we can just modify the number of replicas inside the local docker-compose.prod.yml and update the service using sudo docker stack deploy -c docker-compose.prod.yml executor and check using sudo docker stack services executor.
Note: Use the service you want to scale like in our case it was executor
Updating the image
sudo docker stack deploy -c docker-compose.prod.yml executor
as stack deploy is both used to create and update a stack.
Remove a stack
sudo docker stack rm executor
Leaving the swarm(Can be run on any node)
docker swarm leave
To join the docker container to debug, first get the container id using docker ps.
sudo docker exec -it CONTAINERID /bin/bash