trogers1884/c77_dbh
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add several new functions for analyzing vacuums and indexes.

Browse Source
This commit is contained in:
Tom Rogers 2025-03-25 07:35:16 -05:00
parent 08750c9654
commit aa2f97bd0f
5 changed files with 549 additions and 83 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

143
README.md
View File

@ -1,3 +1,142 @@
# c77_dbh # c77_dbh - Postgres Database Health Extension
**c77_dbh** is a PostgreSQL extension designed to monitor and maintain database health, providing actionable insights for database administrators and non-technical users alike. It offers a suite of functions and tables to track vacuum performance, index efficiency, materialized view refresh health, and TOAST table overhead—all packaged as a self-contained, drop-in solution for Postgres environments.
The extension pushes logic into the database itself, minimizing reliance on external frameworks, and outputs JSON-formatted data ideal for dashboards. It dynamically adjusts thresholds based on your database’s behavior and adapts to optional Postgres extensions (e.g., `pgstattuple`, `pg_stat_statements`) for enhanced precision.
## Features
- **Vacuum Health Monitoring**: Tracks table bloat, vacuum history, and TOAST overhead with dynamic thresholds.
- **Index Health Analysis**: Monitors index usage and bloat, excluding or isolating TOAST indexes as needed.
- **Materialized View Refresh Tracking**: Measures refresh performance with event triggers, identifying bottlenecks (I/O, CPU, RAM).
- **Human-Readable Insights**: Provides status (Red/Yellow/Green), severity scores, insights, and actions for each metric.
- **Dashboard-Ready**: JSON outputs structured for easy integration into graphical interfaces.
- **Resource Awareness**: Gracefully falls back when optional extensions are unavailable.
## Installation
### Prerequisites
- PostgreSQL 13 or higher (tested up to 15 as of March 2025).
- Optional extensions for enhanced functionality:
- `pgstattuple`: Precise bloat estimates for tables and indexes.
- `pg_stat_statements`: Detailed query and refresh performance metrics.
- `pg_cron`: Automate periodic stat updates (recommended).
### Steps
1. **Clone the Repository**:
```bash
git clone https://git.jctr3.com/trogers1884/c77_dbh.git
cd c77_dbh
```
2. **Build and Install**:
```bash
make
sudo make install
```
(Note: Assumes a standard Postgres extension build process; adjust paths if needed.)
3. **Enable in Database**:
```sql
CREATE EXTENSION c77_dbh;
```
4. **Grant Permissions** (optional, adjust roles as needed):
```sql
GRANT SELECT ON ALL TABLES IN SCHEMA public TO public;
GRANT ALL ON ALL TABLES IN SCHEMA public TO your_admin_role;
```
## Usage
### Core Components
- **Tables**:
- `c77_dbh_vacuum_stats`: Stores table vacuum and bloat stats.
- `c77_dbh_index_stats`: Tracks index size, usage, and bloat.
- `c77_dbh_mv_stats`: Records materialized view refresh metrics.
- **Functions**:
- `c77_dbh_update_vacuum_stats()`: Updates vacuum stats for all tables.
- `c77_dbh_update_index_stats()`: Updates index stats.
- `c77_dbh_get_vacuum_health()`: Returns JSON with table health status.
- `c77_dbh_get_index_health()`: JSON for non-TOAST index health.
- `c77_dbh_get_toast_health()`: JSON for TOAST index health.
- `c77_dbh_get_mv_health()`: JSON for materialized view refresh health.
- `c77_dbh_get_mv_details(mv_name text)`: Detailed JSON for a specific MV.
- **Event Triggers**: Automatically track materialized view creation, alteration, and refreshes.
### Example Workflow
1. **Initialize Stats**:
```sql
SELECT c77_dbh_update_vacuum_stats();
SELECT c77_dbh_update_index_stats();
```
2. **Query Health**:
```sql
SELECT * FROM c77_dbh_get_vacuum_health();
SELECT * FROM c77_dbh_get_index_health();
SELECT * FROM c77_dbh_get_mv_health();
```
3. **Automate Updates** (with `pg_cron`):
```sql
SELECT cron.schedule('update_vacuum_stats', '0 * * * *', $$SELECT c77_dbh_update_vacuum_stats()$$);
SELECT cron.schedule('update_index_stats', '0 * * * *', $$SELECT c77_dbh_update_index_stats()$$);
```
4. **Sample Output** (from `c77_dbh_get_vacuum_health()`):
```json
[
{
"metric": "Table Vacuum Health",
"table_name": "sales.orders",
"status": "Red",
"severity": 5,
"insight": "Table bloat is critically high; TOAST overhead significant.",
"action": "Run VACUUM FULL; review large fields.",
"table_size": "128 MB",
"bloat_estimate": "65%",
"details": {"toast_bloat": "40%", "median_bloat": "20%"}
}
]
```
### Dashboard Integration
- Use the JSON outputs to populate a dashboard:
- **Vacuum Health**: Pie chart of statuses, table of critical issues.
- **Index Health**: Bar chart of usage categories, recommendations.
- **MV Refresh**: Progress bars per view, bottleneck alerts.
- Link to detail pages using `details` fields or separate functions like `c77_dbh_get_toast_health()`.
## Configuration
- **Frequency**: Adjust update schedules based on database activity (hourly for high-write systems, daily for stable ones).
- **Permissions**: Default setup grants `SELECT` to `public` and full access to an admin role (e.g., `homestead`). Modify as needed.
- **Extensions**: Install `pgstattuple` and `pg_stat_statements` for best results; fallbacks work without them.
## Limitations
- TOAST bloat estimates are crude without `pgstattuple`.
- Materialized view bottleneck analysis is less precise without `pg_stat_statements`.
- Dynamic thresholds rely on sufficient historical data; new databases may need time to stabilize.
## Future Development
- Query performance monitoring with `pg_stat_statements`.
- Automated maintenance actions (e.g., trigger `VACUUM` on Red status).
- Integration with external frameworks (e.g., Laravel package).
## Contributing
Contributions are welcome! Fork the repo, submit pull requests, or open issues at `https://git.jctr3.com/trogers1884/c77_dbh`.
## License
MIT License – see `LICENSE` file for details.
## Credits
Developed by [Your Name] with assistance from xAI’s Grok. Inspired by Dr. W. Edwards Deming’s statistical quality control principles.
Postgres Database Health

122
dasboardexample.html Normal file
View File

@ -0,0 +1,122 @@
<!DOCTYPE html>
<html>
<head>
<title>Database Health Dashboard</title>
<style>
.dashboard { font-family: Arial, sans-serif; max-width: 1200px; margin: 20px auto; }
.header { text-align: center; padding: 10px; background: #f0f0f0; }
.tabs { display: flex; justify-content: space-around; margin: 10px 0; }
.tab { padding: 10px; cursor: pointer; background: #ddd; }
.tab.active { background: #007bff; color: white; }
.section { border: 1px solid #ccc; padding: 20px; margin: 10px 0; }
.chart { width: 50%; float: left; }
.summary { width: 45%; float: right; }
.status-red { background: #dc3545; color: white; padding: 5px; border-radius: 3px; }
.status-yellow { background: #ffc107; color: black; padding: 5px; border-radius: 3px; }
.status-green { background: #28a745; color: white; padding: 5px; border-radius: 3px; }
.recommendation { font-style: italic; color: #555; }
.details-link { display: block; margin-top: 10px; color: #007bff; text-decoration: none; }
</style>
</head>
<body>
<div class="dashboard">
<div class="header">
<h1>Database Health Dashboard</h1>
</div>
<div class="tabs">
<div class="tab active">Vacuum Health</div>
<div class="tab">Index Health</div>
<div class="tab">MV Refresh</div>
</div>
<!-- Vacuum Health Section -->
<div class="section">
<h2>Vacuum Health Overview</h2>
<div class="chart">
<!-- Pie Chart: 60% Green, 30% Yellow, 10% Red -->
<canvas id="vacuumChart"></canvas>
<script>
// Placeholder for Chart.js or similar
const ctx = document.getElementById('vacuumChart').getContext('2d');
new Chart(ctx, {
type: 'pie',
data: {
labels: ['Healthy', 'Warning', 'Critical'],
datasets: [{ data: [60, 30, 10], backgroundColor: ['#28a745', '#ffc107', '#dc3545'] }]
}
});
</script>
</div>
<div class="summary">
<p><strong>Tables Scanned:</strong> 25</p>
<p><strong>Critical Issues:</strong> <span class="status-red">2 tables</span></p>
<p><strong>Warnings:</strong> <span class="status-yellow">5 tables</span></p>
<p><strong>Healthy:</strong> <span class="status-green">18 tables</span></p>
</div>
<div style="clear: both;"></div>
<p class="recommendation">
Recommendation: Run VACUUM FULL on 2 tables with critical bloat; TOAST overhead notable in 1 case.
</p>
<a href="/vacuum-details" class="details-link">View Detailed Vacuum Stats</a>
</div>
<!-- Index Health Section -->
<div class="section">
<h2>Index Health Overview</h2>
<div class="chart">
<!-- Bar Chart: Index Usage -->
<canvas id="indexChart"></canvas>
<script>
new Chart(document.getElementById('indexChart').getContext('2d'), {
type: 'bar',
data: {
labels: ['Healthy', 'Low Usage', 'Unused'],
datasets: [{ data: [15, 3, 1], backgroundColor: '#28a745' }]
},
options: { scales: { y: { beginAtZero: true } } }
});
</script>
</div>
<div class="summary">
<p><strong>Indexes Monitored:</strong> 19</p>
<p><strong>Critical Issues:</strong> <span class="status-red">1 unused</span></p>
<p><strong>Warnings:</strong> <span class="status-yellow">3 low usage</span></p>
<p><strong>Healthy:</strong> <span class="status-green">15 indexes</span></p>
</div>
<div style="clear: both;"></div>
<p class="recommendation">
Recommendation: Consider dropping 1 unused index; review query patterns for 3 others.
</p>
<a href="/index-details" class="details-link">View Detailed Index Stats</a>
<a href="/toast-details" class="details-link">View TOAST Index Stats</a>
</div>
<!-- MV Refresh Section -->
<div class="section">
<h2>Materialized View Refresh Overview</h2>
<div class="chart">
<!-- Progress Bars: Refresh Performance -->
<div style="margin-bottom: 10px;">
<label>MV1: <span class="status-green">Green</span></label>
<progress value="20" max="100" style="width: 100%;"></progress>
</div>
<div>
<label>MV2: <span class="status-yellow">Yellow</span></label>
<progress value="60" max="100" style="width: 100%;"></progress>
</div>
</div>
<div class="summary">
<p><strong>MVs Monitored:</strong> 2</p>
<p><strong>Critical Issues:</strong> <span class="status-red">0 MVs</span></p>
<p><strong>Warnings:</strong> <span class="status-yellow">1 MV</span></p>
<p><strong>Healthy:</strong> <span class="status-green">1 MV</span></p>
</div>
<div style="clear: both;"></div>
<p class="recommendation">
Recommendation: Monitor MV2 refresh; consider scheduling during off-peak hours.
</p>
<a href="/mv-details" class="details-link">View Detailed MV Stats</a>
</div>
</div>
</body>
</html>

104
health_functions.sql
View File

@ -1,110 +1,95 @@
CREATE OR REPLACE FUNCTION public.c77_dbh_get_mv_health() CREATE OR REPLACE FUNCTION public.c77_dbh_get_mv_health()
RETURNS JSON AS $$ RETURNS JSON AS $$
DECLARE DECLARE
v_has_pg_stat_statements boolean; -- Check if extension is available v_has_pg_stat_statements boolean;
v_stats_reset timestamp; -- Time of last stats reset v_stats_reset timestamp;
v_base_io_load numeric; -- Fallback I/O load v_base_io_load numeric;
v_base_mem_hit_ratio numeric; -- Fallback memory hit ratio v_base_mem_hit_ratio numeric;
v_avg_refresh_time interval; -- Average refresh time across all MVs
BEGIN BEGIN
-- Check if pg_stat_statements is installed
SELECT EXISTS ( SELECT EXISTS (
SELECT 1 FROM pg_extension WHERE extname = 'pg_stat_statements' SELECT 1 FROM pg_extension WHERE extname = 'pg_stat_statements'
) INTO v_has_pg_stat_statements; ) INTO v_has_pg_stat_statements;
-- Get stats_reset, base I/O load, and base memory hit ratio for the current database
SELECT SELECT
d.stats_reset, d.stats_reset,
(d.blks_read + d.blks_hit) / EXTRACT(EPOCH FROM (NOW() - d.stats_reset)), (d.blks_read + d.blks_hit) / EXTRACT(EPOCH FROM (NOW() - d.stats_reset)),
CASE CASE WHEN b.buffers_clean + b.buffers_backend = 0 THEN 1.0
WHEN b.buffers_clean + b.buffers_backend = 0 THEN 1.0 ELSE b.buffers_clean::numeric / (b.buffers_clean + b.buffers_backend) END
ELSE b.buffers_clean::numeric / (b.buffers_clean + b.buffers_backend)
END
INTO v_stats_reset, v_base_io_load, v_base_mem_hit_ratio INTO v_stats_reset, v_base_io_load, v_base_mem_hit_ratio
FROM pg_stat_database d FROM pg_stat_database d
CROSS JOIN pg_stat_bgwriter b CROSS JOIN pg_stat_bgwriter b
WHERE d.datname = current_database(); WHERE d.datname = current_database();
-- Calculate average refresh time across all MVs
SELECT AVG(refresh_mv_time_total / GREATEST(refresh_count, 1))
INTO v_avg_refresh_time
FROM public.c77_dbh_mv_stats
WHERE refresh_count > 0;
RETURN ( RETURN (
SELECT json_agg( SELECT json_agg(
json_build_object( json_build_object(
'metric', 'Materialized View Refresh', 'metric', 'Materialized View Refresh',
'mv_name', m.mv_name, 'mv_name', m.mv_name,
'status', CASE 'status', CASE
WHEN m.refresh_mv_time_last > m.refresh_mv_time_max * 0.9 THEN 'Red' WHEN m.refresh_mv_time_last > v_avg_refresh_time * 2 THEN 'Red'
WHEN m.refresh_mv_time_last > (m.refresh_mv_time_total / GREATEST(m.refresh_count, 1)) * 1.2 THEN 'Yellow' WHEN m.refresh_mv_time_last > v_avg_refresh_time * 1.5 THEN 'Yellow'
ELSE 'Green' ELSE 'Green'
END, END,
'severity', CASE 'severity', CASE
WHEN m.refresh_mv_time_last > m.refresh_mv_time_max * 0.9 THEN 5 WHEN m.refresh_mv_time_last > v_avg_refresh_time * 2 THEN 5
WHEN m.refresh_mv_time_last > (m.refresh_mv_time_total / GREATEST(m.refresh_count, 1)) * 1.2 THEN 3 WHEN m.refresh_mv_time_last > v_avg_refresh_time * 1.5 THEN 3
ELSE 1 ELSE 1
END, END,
'insight', CASE 'insight', CASE
WHEN v_io_load > 100 THEN 'Refreshes are slower due to high I/O load.' WHEN v_io_load > 100 THEN 'Refreshes are slow due to high I/O load.'
WHEN v_cpu_time > (CASE WHEN v_has_pg_stat_statements THEN 1000 ELSE 0.5 END) THEN 'Refreshes are slower due to CPU usage.' WHEN v_cpu_time > (CASE WHEN v_has_pg_stat_statements THEN 1000 ELSE 0.5 END) THEN 'Refreshes are slow due to CPU usage.'
WHEN v_mem_hit_ratio < 0.9 THEN 'Refreshes are slower due to memory pressure.' WHEN v_mem_hit_ratio < 0.9 THEN 'Refreshes are slow due to memory pressure.'
ELSE 'Refreshes are performing normally.' ELSE 'Refresh performance is within normal range.'
END, END,
'action', CASE 'action', CASE
WHEN v_io_load > 100 THEN 'Optimize I/O or schedule during off-peak hours.' WHEN v_io_load > 100 THEN 'Optimize I/O or schedule during off-peak hours.'
WHEN v_cpu_time > (CASE WHEN v_has_pg_stat_statements THEN 1000 ELSE 0.5 END) THEN 'Increase CPU capacity or optimize queries.' WHEN v_cpu_time > (CASE WHEN v_has_pg_stat_statements THEN 1000 ELSE 0.5 END) THEN 'Increase CPU capacity or optimize queries.'
WHEN v_mem_hit_ratio < 0.9 THEN 'Adjust memory settings (e.g., work_mem, shared_buffers).' WHEN v_mem_hit_ratio < 0.9 THEN 'Adjust memory settings (e.g., work_mem).'
ELSE 'No action needed.' ELSE 'No action needed.'
END, END,
'last_refresh_time', m.refresh_mv_time_last::text, 'last_refresh_time', m.refresh_mv_time_last::text,
'refresh_count', m.refresh_count, 'refresh_count', m.refresh_count,
'io_load', round(v_io_load)::text || ' blocks/sec', 'io_load', round(v_io_load)::text || ' blocks/sec',
'cpu_time', (CASE 'cpu_time', (CASE WHEN v_has_pg_stat_statements THEN round(v_cpu_time)::text || ' ms/call'
WHEN v_has_pg_stat_statements THEN round(v_cpu_time)::text || ' ms/call' ELSE (v_cpu_time * 100)::text || '% waiting' END),
ELSE (v_cpu_time * 100)::text || '% waiting'
END),
'buffer_hit_ratio', round(v_mem_hit_ratio * 100)::text || '%', 'buffer_hit_ratio', round(v_mem_hit_ratio * 100)::text || '%',
'bottleneck', CASE 'bottleneck', CASE
WHEN v_io_load > 100 THEN 'I/O' WHEN v_io_load > 100 THEN 'I/O'
WHEN v_cpu_time > (CASE WHEN v_has_pg_stat_statements THEN 1000 ELSE 0.5 END) THEN 'CPU' WHEN v_cpu_time > (CASE WHEN v_has_pg_stat_statements THEN 1000 ELSE 0.5 END) THEN 'CPU'
WHEN v_mem_hit_ratio < 0.9 THEN 'RAM' WHEN v_mem_hit_ratio < 0.9 THEN 'RAM'
ELSE 'None' ELSE 'None'
END END,
'details', json_build_object(
'avg_refresh_time', v_avg_refresh_time::text,
'explanation', 'Thresholds based on average refresh time (' || v_avg_refresh_time::text || ').'
)
) )
) )
FROM public.c77_dbh_mv_stats m FROM public.c77_dbh_mv_stats m
CROSS JOIN LATERAL ( CROSS JOIN LATERAL (
-- Calculate bottleneck stats per MV
SELECT SELECT
CASE CASE WHEN v_has_pg_stat_statements THEN
WHEN v_has_pg_stat_statements THEN COALESCE(SUM(s.shared_blks_read) / EXTRACT(EPOCH FROM (NOW() - v_stats_reset)), 0)
COALESCE(SUM(s.shared_blks_read) / EXTRACT(EPOCH FROM (NOW() - v_stats_reset)), 0) ELSE v_base_io_load END AS v_io_load,
ELSE CASE WHEN v_has_pg_stat_statements THEN
v_base_io_load COALESCE(SUM(s.total_time) / GREATEST(SUM(s.calls), 1), 0)
END AS v_io_load, ELSE CASE WHEN EXISTS (
CASE SELECT 1 FROM pg_stat_activity
WHEN v_has_pg_stat_statements THEN WHERE query ILIKE '%REFRESH MATERIALIZED VIEW ' || m.mv_name || '%'
COALESCE(SUM(s.total_time) / GREATEST(SUM(s.calls), 1), 0) AND wait_event_type = 'CPU' AND state = 'active'
ELSE ) THEN 1 ELSE 0 END END AS v_cpu_time,
CASE CASE WHEN v_has_pg_stat_statements THEN
WHEN EXISTS ( CASE WHEN SUM(s.shared_blks_hit + s.shared_blks_read) = 0 THEN 1.0
SELECT 1 ELSE SUM(s.shared_blks_hit)::numeric / (SUM(s.shared_blks_hit) + SUM(s.shared_blks_read)) END
FROM pg_stat_activity ELSE v_base_mem_hit_ratio END AS v_mem_hit_ratio
WHERE query ILIKE '%REFRESH MATERIALIZED VIEW ' || m.mv_name || '%' FROM (SELECT 1 AS dummy) AS dummy
AND wait_event_type = 'CPU'
AND state = 'active'
) THEN 1
ELSE 0
END
END AS v_cpu_time,
CASE
WHEN v_has_pg_stat_statements THEN
CASE
WHEN SUM(s.shared_blks_hit + s.shared_blks_read) = 0 THEN 1.0
ELSE SUM(s.shared_blks_hit)::numeric / (SUM(s.shared_blks_hit) + SUM(s.shared_blks_read))
END
ELSE
v_base_mem_hit_ratio
END AS v_mem_hit_ratio
FROM (
-- Subquery to avoid duplicate logic
SELECT 1 AS dummy
) AS dummy
LEFT JOIN pg_stat_statements s ON v_has_pg_stat_statements LEFT JOIN pg_stat_statements s ON v_has_pg_stat_statements
AND s.query ILIKE '%REFRESH MATERIALIZED VIEW ' || m.mv_name || '%' AND s.query ILIKE '%REFRESH MATERIALIZED VIEW ' || m.mv_name || '%'
AND s.dbid = (SELECT datid FROM pg_stat_database WHERE datname = current_database()) AND s.dbid = (SELECT datid FROM pg_stat_database WHERE datname = current_database())
@ -114,7 +99,6 @@ END;
$$ LANGUAGE plpgsql; $$ LANGUAGE plpgsql;
CREATE OR REPLACE FUNCTION public.c77_dbh_get_mv_details(p_mv_name text) CREATE OR REPLACE FUNCTION public.c77_dbh_get_mv_details(p_mv_name text)
RETURNS JSON AS $$ RETURNS JSON AS $$
DECLARE DECLARE

89
public.c77_dbh_get_vacuum_health.sql
View File

@ -1,49 +1,96 @@
CREATE OR REPLACE FUNCTION public.c77_dbh_get_vacuum_health() CREATE OR REPLACE FUNCTION public.c77_dbh_get_vacuum_health()
RETURNS JSON AS $$ RETURNS JSON AS $$
DECLARE DECLARE
v_median_bloat numeric;
v_bloat_stddev numeric;
v_has_pgstattuple boolean; v_has_pgstattuple boolean;
BEGIN BEGIN
-- Check if pgstattuple is installed
SELECT EXISTS ( SELECT EXISTS (
SELECT 1 FROM pg_extension WHERE extname = 'pgstattuple' SELECT 1 FROM pg_extension WHERE extname = 'pgstattuple'
) INTO v_has_pgstattuple; ) INTO v_has_pgstattuple;
-- Calculate median and stddev including TOAST-adjusted bloat
WITH toast_bloat AS (
SELECT
(n.nspname || '.' || c.relname) AS table_name,
CASE WHEN v_has_pgstattuple THEN
COALESCE((pgstattuple(c.oid)).free_percent + (pgstattuple(c.oid)).dead_tuple_percent, 0)
ELSE
CASE WHEN s.n_live_tup + s.n_dead_tup = 0 THEN 0
ELSE (s.n_dead_tup::numeric / (s.n_live_tup + s.n_dead_tup)) * 100 END
END AS bloat_estimate
FROM pg_stat_all_tables s
JOIN pg_class c ON c.oid = s.relid
JOIN pg_namespace n ON n.oid = c.relnamespace
WHERE n.nspname NOT IN ('pg_catalog', 'information_schema')
)
SELECT
percentile_cont(0.5) WITHIN GROUP (ORDER BY bloat_estimate),
stddev(bloat_estimate)
INTO v_median_bloat, v_bloat_stddev
FROM toast_bloat;
RETURN ( RETURN (
SELECT json_agg( SELECT json_agg(
json_build_object( json_build_object(
'metric', 'Table Vacuum Health', 'metric', 'Table Vacuum Health',
'table_name', table_name, 'table_name', v.table_name,
'status', CASE 'status', CASE
WHEN bloat_estimate > 50 THEN 'Red' WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + 2 * v_bloat_stddev THEN 'Red'
WHEN bloat_estimate > 20 THEN 'Yellow' WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + v_bloat_stddev THEN 'Yellow'
ELSE 'Green' ELSE 'Green'
END, END,
'severity', CASE 'severity', CASE
WHEN bloat_estimate > 50 THEN 5 WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + 2 * v_bloat_stddev THEN 5
WHEN bloat_estimate > 20 THEN 3 WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + v_bloat_stddev THEN 3
ELSE 1 ELSE 1
END, END,
'insight', CASE 'insight', CASE
WHEN bloat_estimate > 50 THEN 'Table has critical bloat; vacuum and analyze urgently.' WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + 2 * v_bloat_stddev THEN
WHEN bloat_estimate > 20 THEN 'Table has moderate bloat; consider vacuuming.' 'Table bloat is critically high' ||
ELSE 'Table is in good health.' (CASE WHEN t.toast_bloat > v.bloat_estimate THEN '; TOAST overhead is significant.' ELSE '.' END)
WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + v_bloat_stddev THEN
'Table bloat is above average' ||
(CASE WHEN t.toast_bloat > v.bloat_estimate THEN '; TOAST overhead notable.' ELSE '.' END)
ELSE 'Table bloat is within normal range.'
END, END,
'action', CASE 'action', CASE
WHEN bloat_estimate > 50 THEN 'Run VACUUM FULL or CLUSTER; check autovacuum settings.' WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + 2 * v_bloat_stddev THEN
WHEN bloat_estimate > 20 THEN 'Run VACUUM and ANALYZE; monitor autovacuum.' 'Run VACUUM FULL or CLUSTER' ||
ELSE 'No action needed.' (CASE WHEN t.toast_bloat > v.bloat_estimate THEN '; review large fields for TOAST impact.' ELSE '.' END)
WHEN (v.bloat_estimate + COALESCE(t.toast_bloat, 0)) > v_median_bloat + v_bloat_stddev THEN
'Run VACUUM and ANALYZE' ||
(CASE WHEN t.toast_bloat > v.bloat_estimate THEN '; check TOAST usage.' ELSE '.' END)
ELSE 'No immediate action required.'
END, END,
'last_vacuum', last_vacuum::text, 'last_vacuum', v.last_vacuum::text,
'last_autovacuum', last_autovacuum::text, 'last_autovacuum', v.last_autovacuum::text,
'vacuum_count', vacuum_count, 'vacuum_count', v.vacuum_count,
'dead_tuples', dead_tuples, 'dead_tuples', v.dead_tuples,
'live_tuples', live_tuples, 'live_tuples', v.live_tuples,
'table_size', table_size::text || ' bytes', 'table_size', pg_size_pretty(v.table_size),
'bloat_estimate', bloat_estimate::text || '%' || 'bloat_estimate', v.bloat_estimate::text || '%' ||
(CASE WHEN v_has_pgstattuple THEN ' (precise)' ELSE ' (estimated)' END) (CASE WHEN v_has_pgstattuple THEN ' (precise)' ELSE ' (estimated)' END),
'details', json_build_object(
'median_bloat', round(v_median_bloat)::text || '%',
'bloat_stddev', round(v_bloat_stddev)::text || '%',
'toast_bloat', round(COALESCE(t.toast_bloat, 0))::text || '%',
'explanation', 'Includes TOAST bloat in total health assessment.'
)
) )
) )
FROM public.c77_dbh_vacuum_stats FROM public.c77_dbh_vacuum_stats v
LEFT JOIN (
SELECT
(n.nspname || '.' || c.relname) AS table_name,
CASE WHEN v_has_pgstattuple THEN
COALESCE((pgstattuple(tst.oid)).free_percent + (pgstattuple(tst.oid)).dead_tuple_percent, 0)
ELSE 0 END AS toast_bloat
FROM pg_class c
JOIN pg_namespace n ON n.oid = c.relnamespace
LEFT JOIN pg_class tst ON tst.oid = c.reltoastrelid
WHERE n.nspname NOT IN ('pg_catalog', 'information_schema') AND c.reltoastrelid != 0
) t ON v.table_name = t.table_name
); );
END; END;
$$ LANGUAGE plpgsql; $$ LANGUAGE plpgsql;

174
public.c77_dbh_index_stats.sql Normal file
View File

@ -0,0 +1,174 @@
CREATE TABLE IF NOT EXISTS public.c77_dbh_index_stats (
index_name text COLLATE pg_catalog."default" PRIMARY KEY, -- Fully qualified index name
table_name text, -- Associated table
index_size bigint DEFAULT 0, -- Size in bytes
scan_count bigint DEFAULT 0, -- Number of scans
bloat_estimate numeric DEFAULT 0, -- Estimated bloat percentage
last_updated timestamp without time zone DEFAULT now() -- Last stats update
);
CREATE OR REPLACE FUNCTION public.c77_dbh_update_index_stats()
RETURNS void AS $$
DECLARE
v_has_pgstattuple boolean;
BEGIN
SELECT EXISTS (
SELECT 1 FROM pg_extension WHERE extname = 'pgstattuple'
) INTO v_has_pgstattuple;
INSERT INTO public.c77_dbh_index_stats (
index_name, table_name, index_size, scan_count, bloat_estimate, last_updated
)
SELECT
(n.nspname || '.' || i.relname) AS index_name,
(n.nspname || '.' || t.relname) AS table_name,
pg_relation_size(i.oid) AS index_size,
COALESCE(s.idx_scan, 0) AS scan_count,
CASE WHEN v_has_pgstattuple THEN
ROUND(CAST(COALESCE((pgstattuple(i.oid)).free_percent + (pgstattuple(i.oid)).dead_tuple_percent, 0) AS numeric), 2)
ELSE
0 -- Placeholder; could estimate based on size vs. usage
END AS bloat_estimate,
now() AS last_updated
FROM pg_index ix
JOIN pg_class i ON i.oid = ix.indexrelid
JOIN pg_class t ON t.oid = ix.indrelid
JOIN pg_namespace n ON n.oid = i.relnamespace
LEFT JOIN pg_stat_all_indexes s ON s.indexrelid = i.oid
WHERE n.nspname NOT IN ('pg_catalog', 'information_schema')
ON CONFLICT (index_name) DO UPDATE
SET
table_name = EXCLUDED.table_name,
index_size = EXCLUDED.index_size,
scan_count = EXCLUDED.scan_count,
bloat_estimate = EXCLUDED.bloat_estimate,
last_updated = EXCLUDED.last_updated;
END;
$$ LANGUAGE plpgsql;
CREATE OR REPLACE FUNCTION public.c77_dbh_get_index_health()
RETURNS JSON AS $$
DECLARE
v_median_bloat numeric;
v_median_usage numeric;
BEGIN
SELECT
percentile_cont(0.5) WITHIN GROUP (ORDER BY bloat_estimate),
percentile_cont(0.5) WITHIN GROUP (ORDER BY scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0))
INTO v_median_bloat, v_median_usage
FROM public.c77_dbh_index_stats
WHERE index_name NOT LIKE 'pg_toast.%';
RETURN (
SELECT json_agg(
json_build_object(
'metric', 'Index Health',
'index_name', index_name,
'table_name', table_name,
'status', CASE
WHEN bloat_estimate > v_median_bloat * 2 OR scan_count = 0 THEN 'Red'
WHEN bloat_estimate > v_median_bloat * 1.2 OR (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 'Yellow'
ELSE 'Green'
END,
'severity', CASE
WHEN bloat_estimate > v_median_bloat * 2 OR scan_count = 0 THEN 5
WHEN bloat_estimate > v_median_bloat * 1.2 OR (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 3
ELSE 1
END,
'insight', CASE
WHEN scan_count = 0 THEN 'Index is unused; consider dropping.'
WHEN bloat_estimate > v_median_bloat * 2 THEN 'Index bloat is critically high.'
WHEN (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 'Index usage is low relative to size.'
ELSE 'Index is healthy and well-utilized.'
END,
'action', CASE
WHEN scan_count = 0 THEN 'Drop the index if not needed.'
WHEN bloat_estimate > v_median_bloat * 2 THEN 'Run REINDEX; check autovacuum.'
WHEN (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 'Review query patterns or consider reindexing.'
ELSE 'No action needed.'
END,
'index_size', pg_size_pretty(index_size),
'scan_count', scan_count,
'bloat_estimate', bloat_estimate::text || '%',
'details', json_build_object(
'usage_rate', round((scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0))::numeric, 2)::text || ' scans/MB',
'median_bloat', round(v_median_bloat)::text || '%',
'median_usage', round(v_median_usage, 2)::text || ' scans/MB'
)
)
)
FROM public.c77_dbh_index_stats
WHERE index_name NOT LIKE 'pg_toast.%'
);
END;
$$ LANGUAGE plpgsql;
CREATE OR REPLACE FUNCTION public.c77_dbh_get_toast_health()
RETURNS JSON AS $$
DECLARE
v_median_bloat numeric;
v_median_usage numeric;
v_has_pgstattuple boolean;
BEGIN
SELECT EXISTS (
SELECT 1 FROM pg_extension WHERE extname = 'pgstattuple'
) INTO v_has_pgstattuple;
SELECT
percentile_cont(0.5) WITHIN GROUP (ORDER BY bloat_estimate),
percentile_cont(0.5) WITHIN GROUP (ORDER BY scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0))
INTO v_median_bloat, v_median_usage
FROM public.c77_dbh_index_stats
WHERE index_name LIKE 'pg_toast.%';
RETURN (
SELECT json_agg(
json_build_object(
'metric', 'TOAST Index Health',
'index_name', index_name,
'parent_table', table_name,
'status', CASE
WHEN bloat_estimate > v_median_bloat * 2 OR scan_count = 0 THEN 'Red'
WHEN bloat_estimate > v_median_bloat * 1.2 OR (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 'Yellow'
ELSE 'Green'
END,
'severity', CASE
WHEN bloat_estimate > v_median_bloat * 2 OR scan_count = 0 THEN 5
WHEN bloat_estimate > v_median_bloat * 1.2 OR (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 3
ELSE 1
END,
'insight', CASE
WHEN scan_count = 0 THEN 'TOAST index is unused; parent table may not need toasted columns indexed.'
WHEN bloat_estimate > v_median_bloat * 2 THEN 'TOAST index bloat is critically high.'
WHEN (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 'TOAST index usage is low relative to size.'
ELSE 'TOAST index is healthy and well-utilized.'
END,
'action', CASE
WHEN scan_count = 0 THEN 'Review parent table queries; consider dropping TOAST index.'
WHEN bloat_estimate > v_median_bloat * 2 THEN 'Run REINDEX on TOAST table; optimize large fields in parent.'
WHEN (scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0)) < v_median_usage * 0.5 THEN 'Check parent table usage patterns.'
ELSE 'No action needed.'
END,
'index_size', pg_size_pretty(index_size),
'scan_count', scan_count,
'bloat_estimate', bloat_estimate::text || '%' ||
(CASE WHEN v_has_pgstattuple THEN ' (precise)' ELSE ' (estimated)' END),
'details', json_build_object(
'usage_rate', round((scan_count::numeric / NULLIF(index_size::numeric / 1024^2, 0))::numeric, 2)::text || ' scans/MB',
'median_bloat', round(v_median_bloat)::text || '%',
'median_usage', round(v_median_usage, 2)::text || ' scans/MB',
'parent_columns', (
SELECT string_agg(attname, ', ')
FROM pg_attribute a
JOIN pg_class c ON c.oid = (SELECT oid FROM pg_class WHERE relname = split_part(table_name, '.', 2) AND relnamespace = (SELECT oid FROM pg_namespace WHERE nspname = split_part(table_name, '.', 1)))
WHERE a.attrelid = c.oid AND a.attlen = -1 -- Variable-length columns likely toasted
)
)
)
)
FROM public.c77_dbh_index_stats
WHERE index_name LIKE 'pg_toast.%'
);
END;
$$ LANGUAGE plpgsql;