[Bug] [llm-worker] model create with dashboard unhealty when llm-worker restart

[zxlzhd]

Search before asking

• I had searched in the issues and found no similar issues.

Operating system information

Linux

Python version information

=3.11

DB-GPT version

main

Related scenes

• Chat Data
• Chat Excel
• Chat DB
• Chat Knowledge
• Model Management
• Dashboard
• Plugins

Installation Information

• Installation From Source

• Docker Installation

• Docker Compose Installation

• Cluster Installation

• AutoDL Image

• Other

Device information

CPU

Models information

LLM:chatgpt_proxyllm

What happened

I create a model "4o-last"with chatgt in the dashboard,but when i restart the llm-worker,the new model changed unhealthy,
I don't know which config can fix it

after llm-worker restart

(base) ubuntu@new-test:/gt/DB-GPT$ docker compose exec controller dbgpt model list
WARN[0000] /home/ubuntu/gt/DB-GPT/docker-compose.yml: version is obsolete
+------------------+------------+------------+------+---------+---------+-----------------+---------------------+
| Model Name | Model Type | Host | Port | Healthy | Enabled | Prompt Template | Last Heartbeat |
+------------------+------------+------------+------+---------+---------+-----------------+---------------------+
| chatgpt_proxyllm | llm | llm-worker | 8001 | True | True | | 2025-01-03T08:52:56 |
| WorkerManager | service | llm-worker | 8001 | True | True | | 2025-01-03T08:53:07 |
| 4o-last | llm | llm-worker | 8001 | True | True | | 2025-01-03T08:53:18 |
+------------------+------------+------------+------+---------+---------+-----------------+---------------------+
(base) ubuntu@new-test:/gt/DB-GPT$ docker compose stop llm-worker
WARN[0000] /home/ubuntu/gt/DB-GPT/docker-compose.yml: version is obsolete
[+] Stopping 1/1
✔ Container db-gpt-llm-worker-1 Stopped 2.0s
(base) ubuntu@new-test:/gt/DB-GPT$ docker compose start llm-worker
WARN[0000] /home/ubuntu/gt/DB-GPT/docker-compose.yml: version is obsolete
[+] Running 1/1
✔ Container db-gpt-llm-worker-1 Started 0.3s
(base) ubuntu@new-test:/gt/DB-GPT$ docker compose exec controller dbgpt model list
WARN[0000] /home/ubuntu/gt/DB-GPT/docker-compose.yml: version is obsolete
+------------------+------------+------------+------+---------+---------+-----------------+---------------------+
| Model Name | Model Type | Host | Port | Healthy | Enabled | Prompt Template | Last Heartbeat |
+------------------+------------+------------+------+---------+---------+-----------------+---------------------+
| chatgpt_proxyllm | llm | llm-worker | 8001 | True | True | | 2025-01-03T08:53:45 |
| WorkerManager | service | llm-worker | 8001 | True | True | | 2025-01-03T08:53:45 |
| 4o-last | llm | llm-worker | 8001 | False | True | | 2025-01-03T08:53:18 |
+------------------+------------+------------+------+---------+---------+-----------------+---------------------+

What you expected to happen

the llm-worker logs only register chatgpt_proxyllm@llm,it's not get all register model from controller

How to reproduce

...

Additional context

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[zxlzhd]

it's my controller and llm-worker config，
i want know how to register the model in table dbgpt_cluster_registry_instance
controller:
image: eosphorosai/dbgpt:latest
command: dbgpt start controller --host 0.0.0.0 --port 8000
ports:
- "18000:8000" # 映射到18000
environment:
- DBGPT_LOG_LEVEL=DEBUG
# MySQL注册中心配置
- REGISTRY_TYPE=database
- REGISTRY_DB_TYPE=mysql
- REGISTRY_DB_HOST=db
- REGISTRY_DB_PORT=3306
- REGISTRY_DB_USER=root
- REGISTRY_DB_PASSWORD=aa123456
- REGISTRY_DB_NAME=dbgpt
depends_on:
- db
networks:
- dbgptnet
api-server:
image: eosphorosai/dbgpt:latest
command: dbgpt start apiserver --controller_addr http://controller:8000
restart: unless-stopped
depends_on:
- controller
networks:
- dbgptnet
ports:
- 18100:8100/tcp
llm-worker:
image: eosphorosai/dbgpt:latest
command: >
dbgpt start worker
--model_name chatgpt_proxyllm
--model_path chatgpt_proxyllm
--worker_type llm
--proxy_server_url
--proxy_api_key
--controller_addr http://controller:8000
--worker_register_host llm-worker
--port 8001
--heartbeat_interval 30
hostname: llm-worker
ports:
- "18001:8001" # 映射到18001
environment:
- DBGPT_LOG_LEVEL=DEBUG
- OPENAI_API_KEY=***
- OPENAI_API_BASE=***
- WORKER_REGISTER=true
- WORKER_SEND_HEARTBEAT=true
- WORKER_HEARTBEAT_INTERVAL=30
healthcheck:
test: ["CMD", "curl", "-f", "http://localhost:8001/health"]
interval: 30s
timeout: 10s
retries: 3
volumes:
- ./data:/data
# Please modify it to your own model directory
- ./data/models:/app/models
depends_on:
- controller
networks:
- dbgptnet

[Aries-ckt]

@fangyinc please take a look.