[BUG] Messages not delivered when sending via admin token - "appid is invalid" suddenly occurs

[ashin0808]

OpenIM Server Version

3.7.0

Operating System and CPU Architecture

Linux (AMD)

Deployment Method

Docker Deployment

Bug Description and Steps to Reproduce

[v3.7.0] Messages not delivered when sending via admin token - "appid is invalid" suddenly occurs

Hi OpenIM team,

We are running OpenIM v3.7.0 deployed via docker-compose in production.

Our messaging flow is:

Sender: our own backend server (not SDK) sends messages using the admin token via OpenIM API.
Receiver: clients use the official SDK to receive messages.
All clients are connected via WebSocket and sending/receiving heartbeats normally.
Initially, everything worked as expected. However, recently we suddenly observed that a large number of messages failed to reach the client.

Key points:

WebSocket connections remain active and heartbeats are normal during the issue.
Messages are sent from the server side with admin token, but fail to be delivered to clients.
Logs show "appid is invalid" in the push service and "redis: nil" in the message service.
The problem started suddenly without any changes to configs or deployment.
Restarting services temporarily restores functionality, but the issue comes back.
Logs:

push failed {"error": "code 20001, msg appid is invalid"}

GetMessagesBySeq failed {"error": "redis: nil"}

Example client connection log shows the user still online:

WebSocket connected, sending heartbeats normally

No message received during push failure period

Environment:

Version: OpenIM v3.7.0
Deployment: docker-compose (MongoDB, Redis, Kafka, etcd, MinIO)
Redis password configured and accessible by all services
Nginx used for SSL termination and WebSocket proxying, heartbeat interval < timeout
Questions:

Could "appid is invalid" occur when sending messages via admin token if some internal appid or session expires?
Why would "appid" suddenly become invalid without any deployment or configuration changes?
Can this "appid is invalid" cause Redis not to store messages ("redis: nil")?
Are there any known issues in v3.7.0 with admin token message sending that would require upgrading or fixing?
Request:

Please help:

Confirm the root cause of "appid is invalid" in this scenario.
Suggest how to configure/admin token message sending to avoid this issue.
Let us know if this is a known bug and if there is a patch or workaround available.
Thanks!

Screenshots Link

No response

[OpenIM-Robot]

Hello! Thank you for filing an issue.

If this is a bug report, please include relevant logs to help us debug the problem.

Join slack 🤖 to connect and communicate with our developers.

[ashin0808]

Here are the relevant logs from the time when the issue occurred.

You can see multiple push failed entries with "error": "code 20001, msg appid is invalid" and some GetMessagesBySeq failed {"error": "redis: nil"}.

During the same time, msggateway service logs show repeated readMessage i/o timeout, meaning the clients were still connected but possibly stalled on receiving data.

2025-09-19 01:39:55.986 [openim-push] push failed {"error": "code 20001, msg appid is invalid"}

2025-09-19 01:40:06.790 [openim-push] push failed {"error": "code 20001, msg appid is invalid"}

2025-09-19 01:40:14.988 [openim-push] push failed {"error": "code 20001, msg appid is invalid"}

2025-09-19 01:40:15.386 [openim-push] push failed {"error": "code 20001, msg appid is invalid"}

2025-09-19 01:40:31.091 [openim-push] push failed {"error": "code 20001, msg appid is invalid"}

... (many similar push failed logs) ...

2025-09-19 01:42:00.816 [openim-rpc-msg] GetMessagesBySeq failed {"error": "redis: nil"}

2025-09-19 01:42:01.122 [openim-push] push failed {"error": "code 20001, msg appid is invalid"}

At the same timestamps, msggateway service shows:

2025-09-19 01:39:51.171 [openim-msggateway] readMessage {"error": "i/o timeout"}

... repeated readMessage timeouts for different connIDs and users ...

2025-09-19 01:41:38.625 [openim-msggateway] readMessage {"error": "unexpected EOF"} followed by [rpc-user] "this user not online, error: redis: nil"

Observation:

Push service returns "appid is invalid" for all attempted server-to-server message sends during the issue window
Message RPC GetMessagesBySeq fails with redis: nil for some clients
Msggateway reports i/o timeout for many connections although clients were expected to be online
These logs align with the time when messages stopped delivering

[ashin0808]

Hello! Thank you for filing an issue.

If this is a bug report, please include relevant logs to help us debug the problem.

Join slack 🤖 to connect and communicate with our developers.

Here are the relevant logs from the time when the issue occurred.

[mo3et]

The v3.7 version has stopped receiving support. Please try to update to the latest v3.8.3-patch.