Skip to content

Self-improving AI Employees for everyone

fix(workflow): close get_node_output stale-overwrite race via setdefault

trohitg2be456395bd2
diff --git a/server/services/workflow.py b/server/services/workflow.py
index 00f3f5f6..18834694 100644
--- a/server/services/workflow.py
+++ b/server/services/workflow.py
@@ -58,7 +58,12 @@ class WorkflowService:
self.database = database
self.settings = settings
- # In-memory output storage (fast access during execution)
+ # In-memory output storage (fast access during execution).
+ # Single-threaded asyncio: synchronous dict ops between awaits
+ # are atomic, so store_node_output / clear_all_outputs need no
+ # lock. The one real concurrency hazard is the DB-fallback
+ # re-cache in get_node_output, which yields between read and
+ # write -- handled with dict.setdefault (atomic at the GIL).
self._outputs: Dict[str, Dict[str, Any]] = {}
# Initialize NodeExecutor
@@ -474,9 +479,11 @@ class WorkflowService:
) -> None:
"""Store node execution output."""
key = f"{session_id}_{node_id}"
- if key not in self._outputs:
- self._outputs[key] = {}
- self._outputs[key][output_name] = data
+ # setdefault is atomic at the GIL level; this single statement
+ # replaces the old check-then-set + indexed assignment with two
+ # ops that are individually atomic and together race-free since
+ # there is no await between them.
+ self._outputs.setdefault(key, {})[output_name] = data
logger.debug(f"[store_node_output] Stored in memory: key={key}, output_name={output_name}, _outputs keys={list(self._outputs.keys())}")
await self.database.save_node_output(node_id, session_id, output_name, data)
@@ -496,9 +503,13 @@ class WorkflowService:
output = await self.database.get_node_output(node_id, session_id, output_name)
logger.debug(f"[get_node_output] DB lookup result: {'FOUND' if output else 'NOT_FOUND'}")
if output:
- if key not in self._outputs:
- self._outputs[key] = {}
- self._outputs[key][output_name] = output
+ # The await above yielded control; another coroutine may
+ # have written a fresher value via store_node_output. Use
+ # nested setdefault so we never overwrite an existing
+ # in-memory entry with a possibly stale DB read. Then
+ # re-read the slot to return whichever value won.
+ self._outputs.setdefault(key, {}).setdefault(output_name, output)
+ output = self._outputs[key][output_name]
# Special handling for start nodes
if output is None and node_id.startswith('start-'):
@@ -528,7 +539,7 @@ class WorkflowService:
"""Clear all outputs for a session."""
keys = [k for k in self._outputs if k.startswith(f"{session_id}_")]
for k in keys:
- del self._outputs[k]
+ self._outputs.pop(k, None)
await self.database.clear_session_outputs(session_id)
# =========================================================================