fix: execution info patching (#7406)

## Summary

- Fixes a bug where `runtime.execution_info` is `None` in the
distributed runtime (LangGraph Platform) because tasks are prepared by
the server and deserialized in the executor, bypassing the OSS
`_algo.py` code that normally creates `ExecutionInfo`
- Adds `_ensure_execution_info()` in `_retry.py` that lazily constructs
`ExecutionInfo` from the task config when it's missing, called in both
`run_with_retry` and `arun_with_retry`

## Why

PR #7363 introduced `ExecutionInfo` on `Runtime`, populated during task
preparation in `_algo.py`. However, on LangGraph Platform (distributed
runtime), tasks are serialized by the server and deserialized in a
separate executor process — `_algo.py` task prep never runs in that
context, so `execution_info` remains `None`. Any user code or tooling
relying on `runtime.execution_info` (e.g. for tracing, logging, or auth)
would fail with `None` fields.

The fix reconstructs `ExecutionInfo` from config keys (`checkpoint_id`,
`checkpoint_ns`, `task_id`, `thread_id`, `run_id`) that are already
present in the deserialized task config, so no server-side changes are
needed.

## Test plan

- [x] `_ensure_execution_info` is a no-op when `execution_info` already
exists
- [x] `_ensure_execution_info` creates `ExecutionInfo` from config when
missing
- [x] Falls back to `task.id` when `CONFIG_KEY_TASK_ID` is absent from
config
- [x] End-to-end `run_with_retry` test simulating the distributed
runtime scenario (runtime with `execution_info=None`)
This commit is contained in:
Sydney Runkle
2026-04-03 14:53:57 -04:00
committed by GitHub
parent 7eb2d8d6d8
commit e533d00687
2 changed files with 146 additions and 4 deletions
+32 -1
View File
@@ -9,22 +9,51 @@ from collections.abc import Awaitable, Callable, Sequence
from dataclasses import replace
from typing import Any
from langchain_core.runnables import RunnableConfig
from langgraph._internal._config import patch_configurable, recast_checkpoint_ns
from langgraph._internal._constants import (
CONF,
CONFIG_KEY_CHECKPOINT_ID,
CONFIG_KEY_CHECKPOINT_NS,
CONFIG_KEY_RESUMING,
CONFIG_KEY_RUNTIME,
CONFIG_KEY_TASK_ID,
CONFIG_KEY_THREAD_ID,
NS_SEP,
)
from langgraph.errors import GraphBubbleUp, ParentCommand
from langgraph.runtime import Runtime
from langgraph.runtime import ExecutionInfo, Runtime
from langgraph.types import Command, PregelExecutableTask, RetryPolicy
logger = logging.getLogger(__name__)
SUPPORTS_EXC_NOTES = sys.version_info >= (3, 11)
def _ensure_execution_info(
runtime: Runtime, config: RunnableConfig, task: PregelExecutableTask
) -> Runtime:
"""Ensure runtime has execution_info, creating one from config if needed.
In the distributed runtime (LangGraph Platform), tasks are prepared by the
server and deserialized in the executor, bypassing the OSS _algo.py code
that normally creates ExecutionInfo. This function fills in execution_info
from the task config when it's missing.
"""
if runtime.execution_info is not None:
return runtime
configurable = config.get(CONF, {})
return runtime.override(
execution_info=ExecutionInfo(
checkpoint_id=configurable.get(CONFIG_KEY_CHECKPOINT_ID) or "",
checkpoint_ns=configurable.get(CONFIG_KEY_CHECKPOINT_NS) or "",
task_id=configurable.get(CONFIG_KEY_TASK_ID) or task.id,
thread_id=configurable.get(CONFIG_KEY_THREAD_ID),
run_id=str(rid) if (rid := config.get("run_id")) else None,
),
)
def _checkpoint_ns_for_parent_command(ns: str) -> str:
"""Return the checkpoint namespace for the parent graph.
@@ -68,6 +97,7 @@ def run_with_retry(
config = patch_configurable(config, configurable)
runtime = config.get(CONF, {}).get(CONFIG_KEY_RUNTIME)
if isinstance(runtime, Runtime):
runtime = _ensure_execution_info(runtime, config, task)
config = patch_configurable(
config,
{
@@ -172,6 +202,7 @@ async def arun_with_retry(
config = patch_configurable(config, configurable)
runtime = config.get(CONF, {}).get(CONFIG_KEY_RUNTIME)
if isinstance(runtime, Runtime):
runtime = _ensure_execution_info(runtime, config, task)
config = patch_configurable(
config,
{
+114 -3
View File
@@ -1,13 +1,27 @@
from collections import deque
from unittest.mock import Mock, patch
import pytest
from langgraph.checkpoint.memory import MemorySaver
from typing_extensions import TypedDict
from langgraph._internal._constants import (
CONF,
CONFIG_KEY_CHECKPOINT_ID,
CONFIG_KEY_CHECKPOINT_NS,
CONFIG_KEY_RUNTIME,
CONFIG_KEY_TASK_ID,
CONFIG_KEY_THREAD_ID,
)
from langgraph.graph import START, StateGraph
from langgraph.pregel._retry import _checkpoint_ns_for_parent_command, _should_retry_on
from langgraph.runtime import Runtime
from langgraph.types import RetryPolicy
from langgraph.pregel._retry import (
_checkpoint_ns_for_parent_command,
_ensure_execution_info,
_should_retry_on,
run_with_retry,
)
from langgraph.runtime import DEFAULT_RUNTIME, ExecutionInfo, Runtime
from langgraph.types import PregelExecutableTask, RetryPolicy
def test_should_retry_on_single_exception():
@@ -456,3 +470,100 @@ def test_execution_info_identity_fields_populated_on_retry():
# node_attempt should increment
assert captured_infos[0]["node_attempt"] == 1
assert captured_infos[1]["node_attempt"] == 2
def test_ensure_execution_info_noop_when_already_set():
"""Test that _ensure_execution_info is a no-op when execution_info exists."""
existing_info = ExecutionInfo(
checkpoint_id="cp-1", checkpoint_ns="ns-1", task_id="task-1"
)
runtime = DEFAULT_RUNTIME.override(execution_info=existing_info)
config = {CONF: {CONFIG_KEY_THREAD_ID: "thread-1"}}
task = Mock(id="task-2")
result = _ensure_execution_info(runtime, config, task)
assert result is runtime
assert result.execution_info is existing_info
def test_ensure_execution_info_creates_from_config():
"""Test that _ensure_execution_info creates ExecutionInfo from config when missing."""
runtime = DEFAULT_RUNTIME.override(execution_info=None)
config = {
"run_id": "run-123",
CONF: {
CONFIG_KEY_CHECKPOINT_ID: "cp-42",
CONFIG_KEY_CHECKPOINT_NS: "ns-42",
CONFIG_KEY_TASK_ID: "task-42",
CONFIG_KEY_THREAD_ID: "thread-42",
},
}
task = Mock(id="fallback-task-id")
result = _ensure_execution_info(runtime, config, task)
assert result.execution_info is not None
assert result.execution_info.checkpoint_id == "cp-42"
assert result.execution_info.checkpoint_ns == "ns-42"
assert result.execution_info.task_id == "task-42"
assert result.execution_info.thread_id == "thread-42"
assert result.execution_info.run_id == "run-123"
def test_ensure_execution_info_falls_back_to_task_id():
"""Test that _ensure_execution_info uses task.id when CONFIG_KEY_TASK_ID is missing."""
runtime = DEFAULT_RUNTIME.override(execution_info=None)
config = {CONF: {}}
task = Mock(id="fallback-task-id")
result = _ensure_execution_info(runtime, config, task)
assert result.execution_info.task_id == "fallback-task-id"
assert result.execution_info.checkpoint_id == ""
assert result.execution_info.checkpoint_ns == ""
def test_run_with_retry_creates_execution_info_when_missing():
"""Test that run_with_retry works when runtime has no execution_info (distributed runtime scenario)."""
captured_infos: list[ExecutionInfo] = []
class FakeProc:
def invoke(self, input, config):
runtime = config[CONF][CONFIG_KEY_RUNTIME]
captured_infos.append(runtime.execution_info)
return input
runtime = DEFAULT_RUNTIME.override(execution_info=None)
config = {
"run_id": "run-abc",
CONF: {
CONFIG_KEY_RUNTIME: runtime,
CONFIG_KEY_CHECKPOINT_ID: "cp-99",
CONFIG_KEY_CHECKPOINT_NS: "__start__:task123",
CONFIG_KEY_TASK_ID: "task123",
CONFIG_KEY_THREAD_ID: "thread-xyz",
},
}
task = PregelExecutableTask(
name="__start__",
input={"messages": []},
proc=FakeProc(),
writes=deque(),
config=config,
triggers=["__start__"],
retry_policy=[],
cache_key=None,
id="task123",
path=("__pregel_pull", "__start__"),
)
run_with_retry(task, retry_policy=None)
assert len(captured_infos) == 1
info = captured_infos[0]
assert info.checkpoint_id == "cp-99"
assert info.checkpoint_ns == "__start__:task123"
assert info.task_id == "task123"
assert info.thread_id == "thread-xyz"
assert info.run_id == "run-abc"
assert info.node_attempt == 1
assert info.node_first_attempt_time is not None