菜单
Pattern for Prometheus point-in-time historical queries using the instant query endpoint with ?time=<epoch>. Use when comparing metrics now vs N days ago, or when avg_over_time / subquery approaches fail.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Audit CloudNativePG (CNPG) clusters for backup health, S3 storage bloat, and disk-space risk — schedule misconfiguration, WAL/archive_timeout churn, orphaned backups, compression, and PVC resize status. Use when investigating CNPG backup storage growth, a "storage near empty" alert on the backup target (e.g. QNAP), a crash-looping/"Not enough disk space" CNPG cluster, or doing a periodic CNPG health check.
Push an updated Coder workspace Terraform template and redeploy running workspaces onto it. Use when workspace-template/main.tf changes (resources, volumes, etc.) need to reach live Coder workspaces.
Diagnose and fix ONLYOFFICE Document Server showing "An error has occurred while opening the file" with no failed requests visible in the browser. Root cause is usually an orphaned task_result row in the docserver's Postgres DB, left over from a conversion interrupted by a pod crash/OOM/restart.
Task commands, bootstrap sequence, CI/CD, branch/PR workflow, linting, and devcontainer setup for the home-ops repository.
Use when auditing disk health, filesystem usage, or I/O on cluster nodes — especially when filesystem queries return incomplete results or a node's second drive is missing from metrics.
Use when removing, replacing, or decommissioning a Talos node — covers the correct talosctl reset workflow
| name | prometheus-historical-queries |
| description | Pattern for Prometheus point-in-time historical queries using the instant query endpoint with ?time=<epoch>. Use when comparing metrics now vs N days ago, or when avg_over_time / subquery approaches fail. |
| when_to_use | Trigger phrases: "prometheus", "metrics comparison", "7 days ago", "historical query", "now vs last week", "promql over time". |
Extracted: 2026-03-23 Context: Comparing current metrics to a past snapshot in Prometheus
avg_over_time(metric[7d]) often returns errors or misleading values
(e.g., when exporters have gaps, counters reset, or subquery syntax is rejected
by the Prometheus version in use).
Use the instant query endpoint with a ?time=<epoch> parameter to snapshot
metric values at a specific past moment:
WEEK_AGO=$(date -d '7 days ago' +%s)
curl -s "http://localhost:9090/api/v1/query?time=${WEEK_AGO}" \
--data-urlencode 'query=<your_promql>'
Then compare that snapshot against the current instant query (no ?time=).
Any time you need "now vs N days ago" comparisons in Prometheus. Prefer this over range functions when the metric has resets or sparse coverage.