Clickhouse system queries

ClickHouse System Queries Skill

Use this skill when the user asks for operational inspection on ClickHouse system.* tables.

Current table coverage:

• system.query_log via references/system-query-log.md

Relationship to sql-expert:

• sql-expert handles general SQL generation and user/business tables.
• This skill handles system-table operational patterns and routing to table-specific references.

System Metrics and ProfileEvents

• Confirm column shape from schema/reference before writing predicates.
• If the user named an exact metric, pass it in the columns list via explore_schema instead of loading the full table schema.
• If ProfileEvents is a Map, access entries as ProfileEvents['Name']. If flattened, use ProfileEvent_Name.

Example — map vs flattened access:

-- Map access
SELECT ProfileEvents['DistributedConnectionFailTry'] AS fails
FROM system.query_log WHERE event_date = today();

-- Flattened column access
SELECT ProfileEvent_DistributedConnectionFailTry AS fails
FROM system.query_log WHERE event_date = today();

Workflow

1. Resolve target — identify system table and intent. Inherit the most recent time window from conversation, or default to last 60 minutes.

2. Load reference — for system.query_log, call skill_resource to load references/system-query-log.md before writing any SQL. For unsupported tables, fall back to sql-expert.

3. Execute — choose the right tool:

   • search_query_log for standard ranked searches and filtered lookups
   • execute_sql for visualization, time-bucketed aggregation, trends, or histograms

   -- search_query_log: standard lookup
   -- finds top 10 slowest queries in the last hour
   
   -- execute_sql: time-bucketed visualization
   SELECT toStartOfFiveMinutes(event_time) AS bucket,
          count() AS queries,
          avg(query_duration_ms) AS avg_ms
   FROM system.query_log
   WHERE event_date = today() AND event_time > now() - INTERVAL 1 HOUR
   GROUP BY bucket ORDER BY bucket
   

   Default to LIMIT 50 unless the user specifies otherwise.

4. Summarize with concise findings and next actions.

Guardrails

• Always apply time bounds for log-like system tables
• Always use the table-specific reference when available
• Never generate system.query_log SQL until references/system-query-log.md is loaded in the current turn
• Never use search_query_log for chart-oriented requests
• Never omit LIMIT in exploratory queries