Deleting a live flow ends the calls it has out on workers and drops what a removed node still returns, so a node that never returns gives its slot and process back and a late answer no longer recreates the flow's values; a cancel goes through the call's own inbox and is sent after the call's frame, so a service stopped right after starting no longer cancels before its serve

This commit is contained in:
2026-11-11 09:40:59 +01:00
parent 138c632576
commit a3d532f3a5
8 changed files with 187 additions and 12 deletions
+16 -1
View File
@@ -483,8 +483,16 @@ class Pipeline:
self._gate_revision += 1
def remove_flow(self, flow: str) -> None:
"""Take a deleted flow out of the graph."""
"""Take a deleted flow out of the graph.
Its nodes are unbound as well. One still running has nowhere left to
put what it returns, and writing it anyway would bring the deleted
flow's values back.
"""
gone = self.flow_nodes(flow)
self.replace_flow(flow, [])
for node in gone:
node.bind(None)
def _seed(self, initial_values: dict[str, Any] | None) -> None:
"""Give messages a starting value, without overwriting one already there.
@@ -1119,6 +1127,9 @@ class Pipeline:
result = node.execute(inputs)
finally:
ARRIVED.reset(token)
if node._pipeline is not self:
# Its flow was deleted while it ran (`remove_flow`).
return None
self.publish_log(node, collected, "")
if (
@@ -1180,6 +1191,10 @@ class Pipeline:
)
return result or {}
except Exception as exc:
if node._pipeline is not self:
# Deleted while it ran, and most likely failing because the
# call was ended for that: nobody is left to tell.
return None
# One failing node must not take the rest of the graph down.
error = self.publish_error(node, exc, collected, entry_id)
if self.observer is not None: