No available node types

[Zhaozeyong]

单卡H20运行时碰到了以下问题，请问该如何解决呢，可以配置为单卡GPU，GPU数量的配置是在哪呢
[2025-11-12 11:00:08,331 E 354661 355602] core_worker_process.cc:825: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(autoscaler +1m53s) Error: No available node types can fulfill resource request defaultdict(<class 'float'>, {'GPU': 8.0, 'CPU': 8.0}). Add suitable node types to this cluster to resolve this issue.

[steveepreston]

Facing this too:

core_worker_process.cc:825: Failed to establish connection to the metrics exporter agent.

[SixPlusSeven67]

You can try explicitly setting the following in your training script:

trainer.n_gpus_per_node=1

This forces the trainer to use a single GPU and prevents it from requesting 8 GPUs automatically.

For more configuration options, please refer to the default Trainer config:
https://github.com/WooooDyy/AgentGym-RL/blob/main/AgentGym-RL/verl/agent_trainer/config/ppo_trainer.yaml

Regarding the message:

core_worker_process.cc:825: Failed to establish connection to the metrics exporter agent.

It only means that the Ray metrics exporter (used for monitoring) failed to initialize. Training and core functionality are not affected.