[Bug] doris_be_compaction_task_state_total emitted twice in /metrics endpoint on BE nodes (compute-storage decoupled mode)

[sonu1311]

Search before asking

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

Version

Environment

• Doris Version: 4.0.2
• Deployment Mode: Compute-storage decoupled
• Operator: Doris Disaggregated Operator on Kubernetes (EKS)
• Compute Groups: Multiple (confirmed on all BE pods across all compute groups)
• Prometheus Version: [3.9.1]

What's Wrong?

The BE /metrics endpoint emits doris_be_compaction_task_state_total twice
for each label combination (type="base" and type="cumulative") in a single
scrape, resulting in 4 duplicate samples per scrape cycle.

This causes Prometheus to log the following warning on every scrape interval:

level=WARN source=scrape.go:1949 
msg="Error on ingesting samples with different value but same timestamp" 
component="scrape manager" 
num_dropped=4

What You Expected?

Expected output:

# TYPE doris_be_compaction_task_state_total counter
doris_be_compaction_task_state_total{type="cumulative"} 0
doris_be_compaction_task_state_total{type="base"} 0

### How to Reproduce?

Curl any BE pod metrics endpoint directly:
```bash
curl -s http://<be-pod-ip>:8040/metrics | grep compaction_task_state_total

Actual output (all BE pods):

# TYPE doris_be_compaction_task_state_total counter
doris_be_compaction_task_state_total{type="cumulative"} 0
doris_be_compaction_task_state_total{type="cumulative"} 0
doris_be_compaction_task_state_total{type="base"} 0
doris_be_compaction_task_state_total{type="base"} 0

Anything Else?

Additional Notes

The metric appears to be registered twice in the BE metrics registry.
Likely caused by duplicate registration of the same counter in the compaction
subsystem. Worth checking whether CompactionTaskState is registered in
multiple places during BE initialization.

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct