langchain-ai/langgraph
1
0
Fork 0
mirror of https://github.com/langchain-ai/langgraph.git synced 2026-07-15 12:35:43 -04:00
Code Issues 314 Packages Projects Releases 17 Wiki Activity

Duplicate subgraph task execution with identical run_id/request_id creates forked checkpoints #1130

New Issue
Closed
opened 2026-02-20 17:43:11 -05:00 by yindo · 1 comment
yindo commented 2026-02-20 17:43:11 -05:00
Owner
Copy Link
Copy Source

Originally created by @MarkoApostoloski on GitHub (Jan 29, 2026).

Checked other resources

  • This is a bug, not a usage question. For questions, please use the LangChain Forum (https://forum.langchain.com/).
  • I added a clear and detailed title that summarizes the issue.
  • I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example).
  • I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue.

Example Code

recognition_graph = recognition_graph.compile(checkpointer=True)
# ... other subgraphs similarly compiled

primary_graph.add_node("recognition_agent", recognition_graph)
# ... other subgraph nodes

graph = primary_graph.compile()  # Parent graph without explicit checkpointer

Error Message and Stack Trace (if applicable)

Description

When using a multi-agent architecture with subgraphs compiled with checkpointer=True, a single user request is being processed multiple times, creating duplicate checkpoints that fork from the same parent checkpoint.

Reproduction:

  1. Send a single message to the graph via LangGraph Studio or API
  2. Observe checkpoint history - multiple checkpoints created from the same parent

Evidence:
Example 1: Subgraph Agent Fork
Two checkpoints created from identical parent, same run_id and request_id:
Both forks generate separate LLM responses, resulting in duplicate API calls and inconsistent state.

Example 2: Fork at Step 0 (Before Subgraph Entry)
Forking also occurs at the initial node before any subgraph is entered:
In this case, even the message IDs differ, suggesting the request itself is being processed twice.

Example 3: Schema Mismatch Causes KeyError in One Fork
When routing to a subgraph, one fork succeeds while the other fails with KeyError('{agent_name}') due to state schema differences between parent and subgraph.

Expected Behavior
A single request with one run_id and request_id should create exactly one execution path and one set of checkpoints.

Actual Behavior
The same request creates multiple parallel execution paths with:

  • Different task IDs
  • Different checkpoint IDs
  • Different checkpoint namespaces
  • Potential state inconsistencies

Additional Context

  • Only one LangGraph API container is running
  • No load balancer or proxy in front of the service
  • Issue occurs both via LangGraph Studio and direct API calls
  • Issue persists across different subgraphs (recognition, dev_plan, etc.)

State Dumps

Fork 1:

{
"values": {
"messages": [
{
"content": "hello",
"type": "human",
"id": "7ed36b66-64df-4382-9d73-ba4484d17ddb"
}
],
"dialog_state": ["recognition_agent"]
},
"next": ["recognition_agent"],
"tasks": [
{
"id": "20351ca0-5254-b48b-309f-f148705b4ae0",
"name": "recognition_agent",
"checkpoint": {
"thread_id": "f6fa5556-3910-434d-9fc5-30c76c742d72",
"checkpoint_ns": "recognition_agent:20351ca0-5254-b48b-309f-f148705b4ae0"
}
}
],
"metadata": {
"step": 1,
"run_id": "019bac31-ef16-77fb-9840-c55a65bd526d",
"langgraph_request_id": "1ed0253d-8ce4-45f4-9bb0-4649c2fb2c75",
"langgraph_version": "1.0.2"
},
"created_at": "2026-01-11T08:35:13.548498+00:00",
"checkpoint_id": "1f0eec87-2ce6-6921-8001-302cc3716cd4",
"parent_checkpoint_id": "1f0eec87-25fb-6322-8000-84326049837c"
}

Fork 2:

{
"values": {
"messages": [
{
"content": "hello",
"type": "human",
"id": "7ed36b66-64df-4382-9d73-ba4484d17ddb"
}
],
"dialog_state": ["recognition_agent"]
},
"next": ["recognition_agent"],
"tasks": [
{
"id": "e1c3460e-b7b9-00f6-79f0-8060350c0bc4",
"name": "recognition_agent",
"checkpoint": {
"thread_id": "f6fa5556-3910-434d-9fc5-30c76c742d72",
"checkpoint_ns": "recognition_agent:e1c3460e-b7b9-00f6-79f0-8060350c0bc4"
}
}
],
"metadata": {
"step": 1,
"run_id": "019bac31-ef16-77fb-9840-c55a65bd526d",
"langgraph_request_id": "1ed0253d-8ce4-45f4-9bb0-4649c2fb2c75",
"langgraph_version": "1.0.2"
},
"created_at": "2026-01-11T08:35:15.466804+00:00",
"checkpoint_id": "1f0eec87-3f31-6f06-8001-c91e87c87951",
"parent_checkpoint_id": "1f0eec87-25fb-6322-8000-84326049837c"
}

System Info

Environment
langgraph: 1.0.2
langgraph-checkpoint: 3.0.1
langgraph-checkpoint-postgres: 3.0.1
langgraph-api: 0.5.8
Deployment: Self-hosted LangGraph Server (Docker, single container)
Database: PostgreSQL (Azure)

Originally created by @MarkoApostoloski on GitHub (Jan 29, 2026). ### Checked other resources - [x] This is a bug, not a usage question. For questions, please use the LangChain Forum (https://forum.langchain.com/). - [x] I added a clear and detailed title that summarizes the issue. - [x] I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example). - [x] I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue. ### Example Code ```python recognition_graph = recognition_graph.compile(checkpointer=True) # ... other subgraphs similarly compiled primary_graph.add_node("recognition_agent", recognition_graph) # ... other subgraph nodes graph = primary_graph.compile() # Parent graph without explicit checkpointer ``` ### Error Message and Stack Trace (if applicable) ```shell ``` ### Description When using a multi-agent architecture with subgraphs compiled with checkpointer=True, a single user request is being processed multiple times, creating duplicate checkpoints that fork from the same parent checkpoint. __Reproduction__: 1. Send a single message to the graph via LangGraph Studio or API 2. Observe checkpoint history - multiple checkpoints created from the same parent __Evidence__: Example 1: Subgraph Agent Fork Two checkpoints created from identical parent, same run_id and request_id: Both forks generate separate LLM responses, resulting in duplicate API calls and inconsistent state. Example 2: Fork at Step 0 (Before Subgraph Entry) Forking also occurs at the initial node before any subgraph is entered: In this case, even the message IDs differ, suggesting the request itself is being processed twice. Example 3: Schema Mismatch Causes KeyError in One Fork When routing to a subgraph, one fork succeeds while the other fails with KeyError('{agent_name}') due to state schema differences between parent and subgraph. __Expected Behavior__ A single request with one run_id and request_id should create exactly one execution path and one set of checkpoints. __Actual Behavior__ The same request creates multiple parallel execution paths with: - Different task IDs - Different checkpoint IDs - Different checkpoint namespaces - Potential state inconsistencies __Additional Context__ - Only one LangGraph API container is running - No load balancer or proxy in front of the service - Issue occurs both via LangGraph Studio and direct API calls - Issue persists across different subgraphs (recognition, dev_plan, etc.) State Dumps Fork 1: { "values": { "messages": [ { "content": "hello", "type": "human", "id": "7ed36b66-64df-4382-9d73-ba4484d17ddb" } ], "dialog_state": ["recognition_agent"] }, "next": ["recognition_agent"], "tasks": [ { "id": "20351ca0-5254-b48b-309f-f148705b4ae0", "name": "recognition_agent", "checkpoint": { "thread_id": "f6fa5556-3910-434d-9fc5-30c76c742d72", "checkpoint_ns": "recognition_agent:20351ca0-5254-b48b-309f-f148705b4ae0" } } ], "metadata": { "step": 1, "run_id": "019bac31-ef16-77fb-9840-c55a65bd526d", "langgraph_request_id": "1ed0253d-8ce4-45f4-9bb0-4649c2fb2c75", "langgraph_version": "1.0.2" }, "created_at": "2026-01-11T08:35:13.548498+00:00", "checkpoint_id": "1f0eec87-2ce6-6921-8001-302cc3716cd4", "parent_checkpoint_id": "1f0eec87-25fb-6322-8000-84326049837c" } Fork 2: { "values": { "messages": [ { "content": "hello", "type": "human", "id": "7ed36b66-64df-4382-9d73-ba4484d17ddb" } ], "dialog_state": ["recognition_agent"] }, "next": ["recognition_agent"], "tasks": [ { "id": "e1c3460e-b7b9-00f6-79f0-8060350c0bc4", "name": "recognition_agent", "checkpoint": { "thread_id": "f6fa5556-3910-434d-9fc5-30c76c742d72", "checkpoint_ns": "recognition_agent:e1c3460e-b7b9-00f6-79f0-8060350c0bc4" } } ], "metadata": { "step": 1, "run_id": "019bac31-ef16-77fb-9840-c55a65bd526d", "langgraph_request_id": "1ed0253d-8ce4-45f4-9bb0-4649c2fb2c75", "langgraph_version": "1.0.2" }, "created_at": "2026-01-11T08:35:15.466804+00:00", "checkpoint_id": "1f0eec87-3f31-6f06-8001-c91e87c87951", "parent_checkpoint_id": "1f0eec87-25fb-6322-8000-84326049837c" } ### System Info Environment langgraph: 1.0.2 langgraph-checkpoint: 3.0.1 langgraph-checkpoint-postgres: 3.0.1 langgraph-api: 0.5.8 Deployment: Self-hosted LangGraph Server (Docker, single container) Database: PostgreSQL (Azure)
yindo added the bugpending labels 2026-02-20 17:43:11 -05:00
yindo closed this issue 2026-02-20 17:43:11 -05:00
yindo commented 2026-02-20 17:43:11 -05:00
Author
Owner
Copy Link
Copy Source

@MarkoApostoloski commented on GitHub (Jan 29, 2026):

Additional Critical Evidence: Fork at Step -1 (__start__ Node)

Finding

Two checkpoints are created at step -1 (the __start__ node) with parent_checkpoint: null for both forks. This proves the request is being duplicated before any graph code executes.

Fork 1 (Step -1):

{
  "values": {},
  "next": ["__start__"],
  "tasks": [{"id": "5e14aa9c-8cdf-f98a-7df6-0a20bd395e4a", "name": "__start__"}],
  "metadata": {
    "step": -1,
    "run_id": "019bac23-c7f9-73ae-be8b-98d6cc6a9fb7",
    "langgraph_request_id": "f9d65e63-c1f6-450f-8d41-55913fd0839d"
  },
  "parent_checkpoint": null
}

Fork 2 (Step -1):

{
  "values": {},
  "next": ["__start__"],
  "tasks": [{"id": "da0f4c14-6f56-3ed7-bc36-2fd75dd8f3b9", "name": "__start__"}],
  "metadata": {
    "step": -1,
    "run_id": "019bac23-c7f9-73ae-be8b-98d6cc6a9fb7",
    "langgraph_request_id": "f9d65e63-c1f6-450f-8d41-55913fd0839d"
  },
  "parent_checkpoint": null
}
@MarkoApostoloski commented on GitHub (Jan 29, 2026): ## Additional Critical Evidence: Fork at Step -1 (`__start__` Node) ### Finding Two checkpoints are created at **step -1** (the `__start__` node) with **`parent_checkpoint: null`** for both forks. This proves the request is being duplicated **before any graph code executes**. ### Fork 1 (Step -1): ```json { "values": {}, "next": ["__start__"], "tasks": [{"id": "5e14aa9c-8cdf-f98a-7df6-0a20bd395e4a", "name": "__start__"}], "metadata": { "step": -1, "run_id": "019bac23-c7f9-73ae-be8b-98d6cc6a9fb7", "langgraph_request_id": "f9d65e63-c1f6-450f-8d41-55913fd0839d" }, "parent_checkpoint": null } Fork 2 (Step -1): { "values": {}, "next": ["__start__"], "tasks": [{"id": "da0f4c14-6f56-3ed7-bc36-2fd75dd8f3b9", "name": "__start__"}], "metadata": { "step": -1, "run_id": "019bac23-c7f9-73ae-be8b-98d6cc6a9fb7", "langgraph_request_id": "f9d65e63-c1f6-450f-8d41-55913fd0839d" }, "parent_checkpoint": null }
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
agents
awaiting author revision
bug
change
chat models
checkpointer
ci
code quality
code quality
dependencies
documentation
drawing
duplicate
enhancement
good first issue
help wanted
invalid
maintainer
meta
missing information
not maintained by langchain
open-swe-auto
open-swe-max
open-swe-max-auto
pending
performance
pull-request
Mirrored from GitHub Pull Request
 question
requires design
sdk
server
stale
studio
tests
type checking
usage question
v0.5
v1.1
No Label bug pending
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/langgraph#1130
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?