# 502 error during cold boot? | Railway | Page 1

karmic cliff Feb 21, 2025, 4:34 PM

#

Project ID: 483187d7-a9d4-4961-b10f-f2804b57d52b

The "serverless" (formerly "sleep" right?) is active in my fastAPI application in Railway. There have been a few occasions this week when a request was not received (or least I didn't see a trace of it in the logs). The container started and then closed, but the request was not queued as far as I can see.

I would prefer not to disable the sleep/serverless feature, so I am considering a recursive call to a dummy /awake endpoint that would take place prior to the actual request to "awaken" the app, and would iterate a predetermined numbered of times (say, 5) while it gets status code 502. Would you say this is a good approach to avoid the problem?

Or are you aware of any issues on your end that might be related?
Thanks, Manuel

#1006629907067064482

fallow craterBOT Feb 21, 2025, 4:34 PM

#

Project ID: 483187d7-a9d4-4961-b10f-f2804b57d52b

queen venture Feb 21, 2025, 4:36 PM

#

how long does your app take to boot? if it takes longer than 10 seconds we will cancel the incoming request and return our 502 page instead

karmic cliff Feb 21, 2025, 5:10 PM

#

Yes, that's what I have seen, it's possible that it takes longer than 10 seconds.

queen venture Feb 21, 2025, 5:11 PM

#

then yeah theres nothing we can do about that from our side, your app would have to start faster

#

are you doing something at boot that takes a long time?

karmic cliff Feb 21, 2025, 5:16 PM

#

Every time the app is (re) deployed, there are a few dependencies that need to be downloaded and installed, but I don't consider that as part of the booting.

#

Although if by "boot" you mean start after sleep (you probably do), I don't think there's anything in particular in the booting that takes long. The action of the app take long but I don't consider them as part of the booting.

#

Could there be something incorrect in the app's settings?

queen venture Feb 21, 2025, 5:16 PM

#

yeah im only talking about during runtime

karmic cliff Feb 21, 2025, 5:16 PM

#

(rather than in the code of the application)

queen venture Feb 21, 2025, 5:18 PM

#

yeah i can see from the logs its taking a long time for your application to start

karmic cliff Feb 21, 2025, 5:19 PM

#

do you see why that is?

#

I know the processing may take up to 1 min (from the moment the API gets the request until it sends a response), but I'm not aware why it could take long to start

queen venture Feb 21, 2025, 5:20 PM

#

no im sorry but i have no way to answer that, we do not have any observability into what your code is / is not doing

karmic cliff Feb 21, 2025, 5:21 PM

#

so you are saying that the fact it takes long to start comes from the code itself, right? (not the settings of the app)

queen venture Feb 21, 2025, 5:26 PM

#

correct, this is an issue at the application level, and not at the platform level

karmic cliff Feb 21, 2025, 5:30 PM

#

thanks, @queen venture

#

what do you think about my hacky workaround (having the client send recursive call to that /awake)

#

(the client is another API deployed in Railway)

queen venture Feb 21, 2025, 5:31 PM

#

i think it would be best to work towards having your app start faster

karmic cliff Feb 21, 2025, 5:32 PM

#

thanks, we'll try to achieve that

#

Thanks a lot for your help

#

If you have any other tips about how I can track what is causing the problem, they would be appreciated

round galleon Feb 21, 2025, 5:35 PM

#

with a quick google search u can see that fastapi has some troubles with cold starts so you aren't the only one with these issues

#

a reddit thread I saw recommended adjusting some configs

#

other said to optimize dockerfile

#

https://www.reddit.com/r/googlecloud/comments/1d4hwo5/cloud_run_fastapi_slow_cold_starts/

karmic cliff Feb 21, 2025, 5:42 PM

#

thanks, @round galleon

queen venture Feb 21, 2025, 8:40 PM

#

!s

## 502 error during cold boot?