Process
The image worker crashes and takes down the whole site. Put that work in a separate process and verify the web service stays available.
The image worker crashes and takes down the whole site. Put that work in a separate process and verify the web service stays available.
One unhandled crash exits the whole runtime.
Problem and key mechanism: When web requests and image work share one process, an image-worker crash can stop the site; send image jobs to a separate process through a queue or pipe so the OS manages resources and exit states separately and the failure returns to job status.
Verify the failure boundary: Crash the image process and confirm the web process still responds while the job fails or retries. Communication and durable state still need design.
Code editors also separate processes: A code editor can run its language analyzer in a separate process. If the analyzer crashes, the editor can still save files and restart only the analyzer.
Move image processing into a separate process that receives work through a queue or pipe. Record the web and image PIDs, memory, and exit states. Crash the image process and verify web requests still work while the image job becomes visibly failed or retryable.