huawei-cloud-cce-capacity-trend-forecaster

Run the huawei_analyze_cce_capacity_trend action with region and cluster_id, optionally setting hours from 1 to 744. It collects node metrics, HPA status, and autoscaler status, then writes JSON, Markdown, and HTML reports with trend charts.

The analyzer computes slope and trend direction per resource and predicts hours until the configured bottleneck threshold (default 80 percent) is reached. Rising trends with high p95 are flagged as capacity risks in the report.

No. Analysis and simulation are read-only, and HPA configuration uses a two-step workflow: call huawei_configure_cce_hpa without confirm to preview, then add confirm=true only after explicit user approval. Node pool changes always require authorization.

Set HUAWEI_AK and HUAWEI_SK environment variables; project ID is auto-fetched via IAM. Required permissions include cce:cluster:list, cce:node:list, cce:nodepool:list, cce:deployment:list, cce:hpa:list, and aom:metric:get.

A single window can miss burst patterns, so the skill requires at least two comparable history records before lowering baseline capacity. If p95 or max utilization is near the bottleneck threshold, it recommends keeping headroom instead of reducing nodes.

The skill reports the gap in the data_gaps output field and does not fabricate trend conclusions. You should treat the affected resources as unknown and rerun analysis once metric collection is restored.