Session storage error

[mmarzol-hatchworks]

Hi everyone,

I have seen this error entry in my Cloud Run logs:

File "/app/.venv/lib/python3.12/site-packages/google/adk/sessions/database_session_service.py", line 437, in append_event
raise ValueError(
ValueError: The last_update_time provided in the session object '2026-03-20 14:59:14' is earlier than the update_time in the storage_session '2026-03-20 14:59:30'. Please check if it is a stale session.

We are using PostgreSQL as the database for the ADK database.

Any help would be really appreciated.
Thank you
Mauricio

[adk-bot]

Response from ADK Answering Agent (experimental, answer may be inaccurate)

TLDR: This error is a safeguard to prevent stale session data from overwriting newer data in the database. This can happen in environments with multiple application instances, like Cloud Run. To fix this, you should refetch the session from the DatabaseSessionService at the beginning of each request that modifies the session, ensuring you are always working with the most up-to-date version.

---

Hi Mauricio,

This is a great question. The ValueError you're seeing is an important feature of the ADK's session management. It's designed to prevent race conditions, which are common in distributed environments like Cloud Run where multiple instances of your application can run simultaneously.

Why is this happening?

The error message "The last_update_time provided in the session object ... is earlier than the update_time in the storage_session" indicates that the session object your code is trying to use is "stale." Here’s a likely scenario:

1. An instance of your app on Cloud Run loads a session from the PostgreSQL database. This session object exists in the memory of that instance.
2. Before this instance does anything with it, another app instance loads the same session, modifies it, and successfully saves it back to the database. The session record in the database now has a newer update_time.
3. Your first instance, still holding the original, now-outdated session object, attempts to append an event.
4. The DatabaseSessionService detects that the last_update_time of the session object in memory is older than the update_time of the session record in the database. To prevent overwriting the newer data with the old, it raises the ValueError you observed.

How to fix it

To prevent this, you should not hold onto session objects for long periods or across different requests. Instead, adopt a pattern of fetching the latest session state from the database at the beginning of any operation that might modify it.

You can do this by calling session_service.get_session() before you interact with the session. This ensures you are always working with the most recent data.

Here is an example of how you can get the latest session before using it:

# Before performing an action, always get the latest session state
session = await session_service.get_session(
    app_name=APP_NAME, 
    user_id=user_id, 
    session_id=session_id
)

# Now you can safely append an event or perform other actions
# with the up-to-date session object.

By following this pattern, you ensure that your application is resilient to the dynamic scaling and concurrent processing inherent to Cloud Run.

For more details on session management, you can refer to the ADK documentation. [1][2]

[1] https://google.github.io/adk-docs/streaming/dev-guide/part1/
[2] https://google.github.io/adk-docs/sessions/session/

[KOKOSde]

This is a stale session write, not a Postgres bug. On Cloud Run it usually means two requests updated the same session close together, so the fix is to reload the session before each write and serialize writes per session id.