Open yuanhe772 opened 1 month ago
The error alone is not enough to help resolve your issue. Different scenarios can cause new tokens to be requested regardless of caching inside MDS. Please add details describing your environment and how your application makes connections. The token caching only happens within the scope of a single process. Are you frequently spawning a new process that creates a connection? Are you connecting to a lot of different endpoints? Are you connecting with different service principal (client) IDs? How large is your connection pool? What is the token lifetime in your environment? In your environment, what does it mean (what's going on) when the load increases? What's your connection string?
The PR you cited wouldn't have improved anything for the error you are reporting. However, #2775 (in 5.2.2) could improve things, depending on your application.
@yuanhe772 Checking in, can you please share additional details to understand the issue you're facing?
Describe the bug
This is to follow up on a bug fix released with mds v5.2.2, that was supposed to fix the issue of AccessToken being refreshed to frequently which evetually caused the SQL connection to timeout when load increases.
We've bumped up our MDS to v5.2.2, but still face the same issue. Our SDK versions:
MDS 5.2.2, Azure.Identity 1.12.0
Pls help advise if the load tests before releasing confirmed the issue was fixed, and if yes, kindly share your thoughts on how we can proceed with this, thank you