Reserved concurrency limits: does the account pool auto‑scale?
24K reputation · 28 May 2024, 02:41 UTC
Goal
Determine how AWS Lambda behaves when a function’s reserved concurrency setting exceeds the account‑level concurrency limit.
Constraints
- Default account limit is 1,000 concurrent executions (can be increased via support).
- Maximum reserved concurrency per function is 1,000.
- Reserved concurrency is persistent across deployments and does not trigger automatic account limit adjustments.
Unresolved Behavior
When a function’s reserved concurrency is set higher than the current account limit, the function appears to retain its reserved value, but it may be throttled because the effective concurrency is capped at the account limit. It is unclear whether AWS automatically scales the account pool to accommodate the reservation or whether the user must manually request a limit increase.
Questions
- When a function’s reserved concurrency exceeds the account limit, does AWS automatically throttle the function to the account limit or does the reservation remain unchanged?
- Is there any documented mechanism that adjusts the account concurrency limit in response to reserved concurrency settings, or is a support request required each time?
- How does this behavior interact with provisioned concurrency when both are enabled simultaneously?