fix: improve backoff retry mechanism #859
Open
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This PR fixes issue #599 by improving the backoff retry mechanism in the TAP agent. The issue described that when a sender is denied, it aggregates too slowly since RAV requests are only triggered by
UpdateReceiptFeesmessages after a fixedretry_interval(~30 secs), making the aggregation process too slow when the aggregator is UP.Problem
As described in issue #599:
UpdateReceiptFeesmessagesretry_interval(~30 secs)Solution
Use backoff information to retry aggregating instead of a fixed retry_interval (as suggested in the issue):
Backoff System Improvements
BackoffInfo: Replacedin_backoff()withremaining()method for more precise timing controlmin_remaining_backoff(): New method inSenderFeeTrackerto get the minimum remaining backoff across all allocationsschedule_retry()method that considers both global and allocation-specific backoff statesretry_intervalfromSenderAccountArgsin favor of dynamic backoff-based schedulingTesting & Quality
retry_helpers_testsmodule with tests for:BackoffInfo::remaining()reset behavior after successful operationsSenderFeeTracker::min_remaining_backoff()roundtrip functionalitytracker.rsTechnical Details
The retry mechanism now works as follows:
BackoffInfo) and allocation-specific (SenderFeeTracker) backoff states are updatednext_retry_delay()method calculates the maximum of global and allocation-specific remaining backoff timesTesting
The new functionality has been tested with standalone tests that verify:
Files Modified
crates/tap-agent/src/agent/sender_account.rs- Main retry logic improvementscrates/tap-agent/src/agent/sender_accounts_manager.rs- Removed hardcoded retry intervalcrates/tap-agent/src/backoff.rs- Enhanced backoff information trackingcrates/tap-agent/src/test.rs- Added comprehensive unit testscrates/tap-agent/src/tracker.rs- Added min_remaining_backoff method and fixed syntaxFixes
Closes #599