He-Pin opened a new pull request, #2769: URL: https://github.com/apache/pekko/pull/2769
Fix Summary **Problem:** The test case "must update the underutilizationStreak highestUtilization if current utilization is higher" within `MetricsBasedResizerSpec` was failing intermittently on CI; the expected `highestUtilization` value was 2, but the actual value was 1. **Root Cause:** A race condition. The test sends one message each to `routee 0` and `routee 1`, followed by an `await` call; however, the `await` only guarantees that the actors have *begun* processing the messages (specifically, that `first.countDown()` has been invoked). When `reportMessageCount` checks `cell.currentMessage` and `mailbox.numberOfMessages`, a very brief time window may exist—particularly under the high load conditions of the CI environment—during which a specific routee's message count appears to be 0. This leads to the `utilized` metric being calculated as 1 instead of the correct value of 2. **Fix:** After the initial `mockSend(await = true)` calls, an additional `mockSend(await = false)` message is now sent to both `routee 0` and `routee 1`. This ensures that, even if the `currentMessage` state is momentarily indeterminate, there remain pending messages within the mailbox (i.e., `numberOfMessages > 0`), thereby guaranteeing the correct calculation of the `utilized` metric. This approach aligns with the pattern utilized by other tests within the same test file (e.g., the test for "record last totalQueueLength correctly"). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
