Hi, I’m using EC2 to host a single agent in the hybrid setup, and I’d appreciate any feedback on my setup.
1. Running a t3a.small instance where the agent and user code run as separate docker containers
2. The user code container only makes http requests and accesses a postgres db
3. When a run starts, a larger container instance is created (via Fargate). All heavy-lifting work, which requires more memory and CPU, is done here. The result is either saved to db or S3
4. When a run finishes or errors out, the large instance is stopped.
What do you think? Are there any potential issues with this setup? At which point should I start exploring multiple agents (for fault tolerance and scalability)?