About this skill

Problem

When debugging Tencent Cloud CLS, the issue is rarely just missing logs. It may be a wrong topic ID, an unrebuilt index, an offline machine, a collection rule not bound to the right machine group, an alarm shield, or a bad PromQL query. Manual API calls are easy to get wrong: SearchLog uses millisecond timestamps while QueryRangeMetric uses second-level timestamps, and SQL analysis cannot rely on Limit or Offset. This skill turns those checks into an API workflow around scripts/cls_api.py.

How It Works

Log path: use DescribeTopics to find the topic, SearchLog for CQL/SQL search, DescribeLogContext for surrounding logs, and DescribeLogHistogram for time distribution.
Collection path: use DescribeMachineGroups, DescribeMachines, DescribeConfigs, and DescribeConfigMachineGroups to verify machines, collection rules, and bindings.
Metrics and alarms: use QueryRangeMetric for PromQL, and DescribeAlarms, DescribeAlertRecordHistory, GetAlarmLog, and DescribeAlarmShields to trace missing or false alarms.

Boundaries

Credentials must come from environment variables; do not embed TENCENTCLOUD_SECRET_ID or TENCENTCLOUD_SECRET_KEY in commands.
Time precision varies: most log APIs use milliseconds, most metric APIs use seconds, and DescribeLogContext.BTime uses YYYY-mm-dd HH:MM:SS.FFF.
DescribeLogContext has regional limitations; in unsupported regions, simulate context with SearchLog over a time range.

Screenshots

Use Cases

When production errors appear, search a CLS topic by time range with CQL and aggregate error codes.
When log collection is missing, check machine groups, machine status, collection rules, and bindings.
When an alarm did not fire, inspect policies, history, execution logs, and shield rules to find the gap.
When checking host or cluster metrics, run PromQL range queries and verify topic, labels, and configs.

Best For

SREs handling production incidents: search logs with CQL/SQL and inspect context and trends.
Ops engineers maintaining log collection: verify machine groups, machine status, rules, and bindings.
Platform engineers managing alerts: trace policies, history, execution details, and shields.
Backend engineers using CLS metrics: query PromQL and verify labels and collection configs.