-
Notifications
You must be signed in to change notification settings - Fork 475
Open
Description
Investigative information
- Timestamp: 2025-12-16T05:52:49.3675888Z
- Function App version:
- Function App name:
- Function name(s) (as appropriate):
- Invocation ID: 2721b80b-8fe3-41fb-a917-84fc63bdfed3
- ec6fc964-7894-4491-babe-669a14946f84@2025-12-16T05:52:49.0296607Z for a working request
- 57e4009a-381d-4153-8c7b-1b49ecec9175@2025-12-16T05:45:56.3561633Z for a slow request with no client timeout
- Region: Australia East
Repro steps
- Deploy a Flex Consumption app with a single HTTP trigger
- Send 10+ HTTP requests with short timeouts. In my case, clients have a 1.5s timeout. I've reproed with
curl - Review logs - I've seen anywhere up to 10s if I remove the timeout
Expected behavior
Host always starts immediately and function is executed afterwards. Host usually takes <10ms and my function takes ~100ms.
Actual behavior
Intermittently, host initialization (Initializing Warmup Extension. log) starts 400ms-10s after a request is sent. This occurs despite 10 always-ready instances and very low RPS, eg 1 per minute.
In my case, clients are POSTing with a 1.5s timeout, causing 499 errors and lost data. Over 12 hours, I receive ~500 requests and ~50 of those fail.
Known workarounds
I really don't want to swap SKUs... Increasing the client timeout is not possible.
Related information
- Programming language used: Python 3.12
- Bindings used: Microsoft.Azure.Functions.ExtensionBundle.Experimental