We are experiencing persistent healthcheck failures with our FastAPI application on Railway. Despite multiple attempts to fix the issue, the healthcheck continues to fail, preventing successful deployment of the application.

Configuration Details

1. Application: FastAPI API (Python)

2. Railway Configuration:

   CopyInsert

   [deploy] startCommand = "uvicorn api.main:app --host 0.0.0.0 --port $PORT" healthcheckPath = "/railway-health" healthcheckTimeout = 300 healthcheckInterval = 15 healthcheckGracePeriod = 60

3. Healthcheck Endpoint:

   CopyInsert

   @app.get("/railway-health") def railway_healthcheck(): """Extremely simple endpoint for Railway healthcheck""" return {"status": "ok"}

Solution Attempts

1. Dedicated healthcheck endpoint implementation:

   • Created a simple

     /health

     endpoint that always returns a 200 OK status

   • Later created an even simpler

     /railway-health

     endpoint specifically for Railway

   • Both endpoints are independent of any environment variables or external services

2. Application resilience improvements:

   • Modified the

     get_supabase_client()

     function to return

     None

     instead of throwing exceptions when credentials are unavailable

   • Updated endpoints to explicitly check if the Supabase client is available

   • Implemented proper error handling with 503 status when services are unavailable

3. Railway configuration adjustments:

   • Updated railway.toml to use the dedicated healthcheck endpoint

   • Increased healthcheck timeout to 300 seconds

   • Added a 15-second interval between checks

   • Configured a 60-second grace period for initialization

4. Local testing:

   • Tested the healthcheck endpoints locally and confirmed they are working correctly

   • Verified that the application starts without errors even without Supabase environment variables
     
     LOG ERROR:
     ====================

     Jun 10 18:07:55

     Starting Healthcheck

     Jun 10 18:07:55

     ====================

     Jun 10 18:07:55

     Jun 10 18:07:55

     Path: /health

     Jun 10 18:07:55

     Retry window: 1m40s

     Jun 10 18:07:55

     Jun 10 18:08:06

     Attempt #1 failed with service unavailable. Continuing to retry for 1m29s

     Jun 10 18:08:17

     Attempt #2 failed with service unavailable. Continuing to retry for 1m18s

     Jun 10 18:08:19

     Attempt #3 failed with service unavailable. Continuing to retry for 1m16s

     Jun 10 18:08:24

     Attempt #4 failed with service unavailable. Continuing to retry for 1m11s

     Jun 10 18:08:32

     Attempt #5 failed with service unavailable. Continuing to retry for 1m3s

     Jun 10 18:08:48

     Attempt #6 failed with service unavailable. Continuing to retry for 47s

     Jun 10 18:09:18

     Attempt #7 failed with service unavailable. Continuing to retry for 17s

     Jun 10 18:09:18

     Jun 10 18:09:18

     1/1 replicas never became healthy!

     Jun 10 18:09:18

     Healthcheck failed!