Split the DB health page into independent panes so fast pg_catalog-based
diagnostics render without waiting for the slower job table scans, and
enrich the slow queries view with server-side sort, reset, and better
setup guidance.
Backend:
- Split /api/db_health into two endpoints: fast panes (database_size,
connection_pool, table_maintenance, slow_queries, datatables) and
/jobs (job_retention, large_results with scan_limit).
- Add GET /api/db_health/slow_queries?sort=total|mean|calls for
server-side sorting of pg_stat_statements queries (sort whitelisted
via enum, SQL-injection safe).
- Add POST /api/db_health/slow_queries/reset to call
pg_stat_statements_reset().
- Return stats_reset timestamp from pg_stat_statements_info (PG 14+).
- Bump slow queries to top 50 sorted by total_exec_time (was top 10 by
mean_exec_time, which misses high-cumulative-load queries).
- Truncate slow queries to 500 chars (was 200).
- Filter table_maintenance to tables with >= 1000 total tuples.
Frontend (DbHealth.svelte):
- Two tabs (Overview / Jobs) with auto-refresh on selection.
- Refresh buttons right-aligned in both tabs; Jobs tab keeps the
scan_limit selector on the left.
- Job Retention & Large Results always render, with "Click Refresh to
load" placeholders when no data yet.
- Slow queries table: clickable column headers for server-side sort,
click a row to toggle the full query text.
- Reset stats button with confirmation dialog, displays "Stats since"
timestamp for before/after comparison workflow.
- When pg_stat_statements is not installed, show numbered setup
instructions with copyable SQL snippets.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
01e39d9cd1