Function on flex consumption cancelling <100ms under heavy load

[jeremy-skippen-jbhifi]

I'm seeing some behavior in a test app I've stood up where a small percentage of functions are being cancelled when under heavy load.
The app is a .NET 9 isolated function with a storage queue trigger.

Example:

• Timestamp: 2024-12-12T03:33:23.2017486Z
• Function App version: azurefunctions-netiso: 2.0.0+d8b5fe998a8c92819b8ee41d2569d252541
• Invocation ID: 61472a11-657d-4bd9-9fa7-4430ff62de9a
• Host Instance ID: 498be43b-39ce-4b92-94f5-9944e9e3f43e
• Region: Australia East

Repro steps

My test function is triggered off a storage queue.
It does the following.

QueueMessage -> Get Blob -> Update Blob -> Call API

For my test I'm spamming the queue with messages and getting anywhere from 50k - 120k requests per minute.
A small percentage of these are failing with a task cancelled exception:

The cancellation is happening at all points in my test app; some during the Get Blob, some during the Update Blob, and some during the API step. 1 seems to have failed before the worker was invoked.

[jeremy-skippen-jbhifi]

I've checked on a linux consumption plan and can see the same behavior, e.g.:

• Timestamp: 2024-12-17T03:32:28.837434Z
• Function App version: azurefunctions-netiso: 2.0.0+d8b5fe998a8c92819b8ee41d2569d252541
• Invocation ID: b72b1602-73c1-4a59-81dc-5f0165cd2c32
• Host Instance ID: 256a5709-c776-4aab-b097-4d3752a2840d
• Region: Australia East

The consumption plan performed significantly worse and I started to get ~2.4k of these messages a minute:

[HostMonitor] Host CPU threshold exceeded (82 >= 80)

[jeremy-skippen-jbhifi]

I've run the same load on the same build on an elastic premium plan and did not get exceptions

[JAdluri]

Hello @jeremy-skippen-jbhifi Thank you reporting the issue will check and let you know further steps