@adonis-agora/durable 0.46.0 → 0.47.0

package/CHANGELOG.md

@@ -1,5 +1,20 @@
 # @adonis-agora/durable
 
+## 0.47.0
+
+### Minor Changes
+
+- [#327](https://github.com/DavideCarvalho/adonis-agora-durable/pull/327) [`64efe2e`](https://github.com/DavideCarvalho/adonis-agora-durable/commit/64efe2e4cd7419600ed60e4d65aca0e15591636a) Thanks [@DavideCarvalho](https://github.com/DavideCarvalho)! - A step result is no longer overwritten back to `pending`. A turn that stamps a pending step's redispatch deadline (or re-dispatches it) wrote the row it had read a moment earlier. If the worker's result landed in between, the step went back to `pending` with no result, the run parked until the deadline, and the step ran a second time.
+  
+  - `StateStore.saveCheckpointIfPending` (optional, new) writes only while the stored row is still `pending` and reports whether it wrote. The engine uses it for the redispatch stamp, the redispatch itself, and for settling a step from its result. A result that loses the race is treated like a redelivery (the resume is re-driven, the step is not settled twice). The Lucid, in-memory and codec stores implement it, and the shared store contract covers it. A store without it falls back to `saveCheckpoint`.
+  - BullMQ result jobs are retried (5 attempts, exponential backoff) when the engine's handler throws, failed ones are kept for 24h instead of deleted, and a result given up on is logged. Before, one store error at peak load dropped the result silently.
+
+## 0.46.1
+
+### Patch Changes
+
+- fix(deps): update dependency @adonis-agora/filter to v0.11.0 ([#326](https://github.com/DavideCarvalho/adonis-agora-durable/issues/326))
+
 ## 0.46.0
 
 ### Minor Changes