There are a few limitations that are important to understand.

In the latest version:

  • Runs on localhost are limited to 5 minutes.
  • On long-running servers (not serverless) Runs can be retried erroneously.
  • Compute intensive jobs are not well supported.

Runs on localhost are limited to 5 minutes

When developing locally the CLI dev command uses ngrok so messages can be sent to your machine.

Ngrok has a timeout of 5 minutes on a Request/Response cycle. so, if a localhost Run takes longer than 5 minutes to complete, the Run will fail.

This limitation will be removed in the future by adding an alternative run strategy that works well on localhost and long-running servers. This won’t use the request/response cycle.

On long-running servers (not serverless) Runs can be retried erroneously

Currently the only way that Runs are performed is by a Request/Response cycle when run is called on a Job. This is optimized for serverless functions (where you have to use a Request/Response cycle), but not for long-running servers.

This limitation will be removed in the future by adding an alternative mode so Jobs works well on localhost and long-running servers. This won’t use the request/response cycle.

Compute intensive jobs are not well supported

Currently the only way that Runs are performed is inside a Request/Response cycle when run is called on a Job. This is not a good way to perform compute intensive jobs.

In the future we will add good support for compute intensive jobs.