html_url,issue_url,id,node_id,user,user_label,created_at,updated_at,author_association,body,reactions,issue,issue_label,performed_via_github_app
https://github.com/simonw/datasette/issues/1522#issuecomment-976023405,https://api.github.com/repos/simonw/datasette/issues/1522,976023405,IC_kwDOBm6k_c46LO9t,360895,steren,2021-11-23T00:08:07Z,2021-11-23T00:08:07Z,NONE,"If you suspect that Cloud Run throttled CPU could be the cause, you can request to have CPU always allocated with `gcloud beta run deploy --no-cpu-throttling` ([read more](https://cloud.google.com/blog/products/serverless/cloud-run-gets-always-on-cpu-allocation))

It could also be the Cloud Run sandbox that somehow gets in the way here, in which case I recommend testing with the second generation execution environment: `gcloud beta run deploy --execution-environment gen2`","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1058896236,Deploy a live instance of demos/apache-proxy,
https://github.com/simonw/datasette/issues/1522#issuecomment-976117989,https://api.github.com/repos/simonw/datasette/issues/1522,976117989,IC_kwDOBm6k_c46LmDl,813732,glasnt,2021-11-23T03:00:34Z,2021-11-23T03:00:34Z,CONTRIBUTOR,"I tried deploying the most recent version of the Dockerfile in this thread ([link to comment](https://github.com/simonw/datasette/issues/1522#issuecomment-974605128)), and after trying a few different different combinations, I was only successful when I used `--no-cpu-throttling` (""CPU Is always allocated"" in the UI)

Using this method, I got a very similar issue to you: The first time I'd load the site I'd get a 503. But after that first load, I didn't get the issue again. It would re-occur if the service started from cold boot. 

I suspect this is a race condition in the supervisord configuration. The errors I got were the same `Connection refused: AH00957: http: attempt to connect to 127.0.0.1:8001 (127.0.0.1) failed`, and that seems to indicate that `datasette` hadn't yet started. 

Looking at the order of logs getting back, the processes reported successfully completing loading after the first 503 was returned, so that makes me think race condition. 

I can replicate this locally, if I `docker run` and request `localhost:5000/prefix` _before_ I get the `datasette entered RUNNING state` message. Cloud Run wakes up when requests are received, so this test would semi-replicate that, but local docker would be the equivalent of a persistent process, hence it doesn't normally exhibit the same issues.

Unfortunately supervisor/supervisor issue 122 (not linking as to prevent cross-project link spam) seems to say that dependency chaining is a feature that's been asked for for a long time, but hasn't been implemented. You could try some suggestions in that thread. ","{""total_count"": 1, ""+1"": 1, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1058896236,Deploy a live instance of demos/apache-proxy,