Rev 36363 |
Rev 36369 |
Go to most recent revision |
Last modification |
Compare with Previous |
View Log
| RSS feed
Last modification
- Rev 36367 2026-04-23 17:07:24
- Author: amit
- Log message:
- HDFC webhook: delegate writes to HdfcProcessingHelper; always respond OK
Fixes the concurrent-duplicate deadlock observed on hadb1 2026-04-23 12:54:11
(two HDFC retries inserting same utr='308490624083' → InnoDB deadlock →
LockAcquisitionException → 500 → HDFC retries again → amplification).
- addPayment: the HDFC row build + persist + wallet/sidbi is now a single call
to hdfcProcessingHelper.processPayment() which runs in REQUIRES_NEW. The
helper catches DataIntegrityViolationException and logs silently; our tx
is unaffected.
- The else branch (pre-select found an existing row) now returns the same
code=100 status=Success as the main path instead of code=200 Duplicate.
Per product direction: HDFC should see 'OK' regardless of whether the row
already existed from a prior webhook, manual entry, or reconciliation —
retries stop cleanly, no 'Duplicate' surface distinction.
Did NOT add @Transactional(readOnly=true) on the outer method despite its
read-heavy shape — checkManualPayments() at line 277 has a write side-effect
(rejecting pending manual wallet requests for the same utr) which would be
silently dropped in a read-only session. Kept class-level writable tx for
the outer; only HDFC writes moved to REQUIRES_NEW helper.