Repository navigation
[Bug] Name of metrics dynamically change #744
Description
Search before asking
- I searched in the issues and found nothing similar.
Description
Currently, when I start a Flink Agent job and check the metrics on Prometheus, I can see that the name of the metric is as follows:
flink_taskmanager_job_task_operator_chat_model_action_serviceType_qwen3_6_plus_totalTokens
In this metric group, "chat_model_action" and "qwen3_6_plus" are the name of the action and the model. It means this metric name would alter according to different models and actions, which makes it hard to track Flink Agent metrics
How to reproduce
Configure a Prometheus metric server for a Flink Agent job, and check the reported metrics.
Version and environment
Flink 1.20, currently latest Flink Agent code (1ca367d).
Are you willing to submit a PR?
- I'm willing to submit a PR!
Activity