What problem does it solve?
Investigating slow queries, OOMs, and timeouts across PostHog's production ClickHouse clusters (US and EU) requires knowing the right data source, attribution columns, and report structure; this Skill provides the complete methodology so the analysis is consistent and unbiased.
Core Features & Use Cases
- Slow-query analysis over query_log_archive: Query the multi-week
query_log_archive table with typed lc_* columns instead of the hours-only system.query_log, with ready-to-run SQL for headline summaries, daily distributions, and category breakdowns.
- Attribution and root-cause workflows: Rank offenders by team, API key, and product using cluster-hours, OOM counts, bytes, and CPU-seconds, then form hypotheses about causes like unmaterialized JSONExtract or high-cardinality breakdowns.
- Materialization candidate analysis: Identify JSON-extracted event and person properties worth materializing and materialized columns safe to drop, across both regions.
- Use Case: When asked for a slow query report over the last 14 days, run the step-by-step queries via the query-clickhouse-via-metabase skill, write a structured markdown report with linked example queries, and diff it against the previous report.
Quick Start
Generate a ClickHouse slow query performance report for the US cluster covering the last 14 days, including OOM attribution and materialization candidates.
