feat(net): optimize transaction rate limiting#6714
feat(net): optimize transaction rate limiting#6714xxo1shine wants to merge 1 commit intotronprotocol:developfrom
Conversation
…e check 1. Add Manager.getCachedTransactionSize() = pushTransactionQueue + pendingTransactions + rePushTransactions to expose the true cached transaction count across all three queues. 2. Fix isTooManyPending() to include pushTransactionQueue, which was previously omitted, causing the pending threshold to be underestimated. 3. Update TransactionsMsgHandler.isBusy() to factor in the Manager cache size via TronNetDelegate.getCachedTransactionSize(), so the node stops accepting TRX INV messages when the full pipeline is busy. 4. Make the busy threshold configurable via node.maxTrxCacheSize (default: 50000), replacing the hardcoded MAX_TRX_SIZE constant. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
| receiveTcpMinDataLength = 2048 | ||
| maxTransactionPendingSize = 2000 | ||
| pendingTransactionTimeout = 60000 | ||
| maxTrxCacheSize = 50000 // total cached trx across handler queues + pending + rePush |
| TransactionsMsgHandler handler = new TransactionsMsgHandler(); | ||
|
|
||
| TronNetDelegate tronNetDelegateMock = Mockito.mock(TronNetDelegate.class); | ||
| Mockito.when(tronNetDelegateMock.getCachedTransactionSize()).thenReturn(50_001); |
There was a problem hiding this comment.
Good targeted test for the new busy logic — mocking TronNetDelegate to control the cached size is exactly the right approach 👍
Tiny nit: the magic number 50_001 couples the test to the default maxTrxCacheSize. If someone tweaks the default later, this test silently keeps passing only because 50_001 > 50_000 happens to hold. Using Args.getInstance().getMaxTrxCacheSize() + 1 would be future-proof, and adding a boundary case (e.g. exact-threshold returns false) would catch off-by-one regressions. Totally optional polish.
| public boolean isBusy() { | ||
| return queue.size() + smartContractQueue.size() > MAX_TRX_SIZE; | ||
| return queue.size() + smartContractQueue.size() | ||
| + tronNetDelegate.getCachedTransactionSize() > Args.getInstance().getMaxTrxCacheSize(); |
There was a problem hiding this comment.
[NIT] May be specify what happens when maxrxCacheSize <=0 .
| public long pendingTransactionTimeout; | ||
| @Getter | ||
| @Setter | ||
| public int maxTrxCacheSize; // clearParam: 50000 |
There was a problem hiding this comment.
[SHOULD] Remove this comment; otherwise, you’ll need to keep two places in the code consistent.
What does this PR do?
Optimize transaction rate limiting with accurate cache size check.
Add
Manager.getCachedTransactionSize() = pushTransactionQueue + pendingTransactions + rePushTransactionsto expose the true cached transaction count across all three queues.Fix
isTooManyPending()to include pushTransactionQueue, which was previously omitted, causing the pending threshold to be underestimated.Update
TransactionsMsgHandler.isBusy()to factor in the Manager cache size viaTronNetDelegate.getCachedTransactionSize(), so the node stops accepting TRX INV messages when the full pipeline is busy.Make the busy threshold configurable via
node.maxTrxCacheSize (default: 50000), replacing the hardcodedMAX_TRX_SIZEconstant.Fixes #6684
Why are these changes required?
This PR has been tested by:
Follow up
Extra details