postgresai 0.15.0-dev.1 → 0.15.0-dev.10
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +3 -1
- package/bin/postgres-ai.ts +606 -105
- package/bun.lock +4 -4
- package/dist/bin/postgres-ai.js +2355 -577
- package/instances.demo.yml +14 -0
- package/lib/checkup-api.ts +25 -6
- package/lib/checkup.ts +241 -10
- package/lib/config.ts +3 -0
- package/lib/init.ts +196 -4
- package/lib/issues.ts +72 -72
- package/lib/mcp-server.ts +90 -0
- package/lib/metrics-loader.ts +3 -1
- package/lib/reports.ts +373 -0
- package/lib/storage.ts +291 -0
- package/lib/supabase.ts +8 -1
- package/lib/util.ts +7 -1
- package/package.json +4 -4
- package/scripts/embed-checkup-dictionary.ts +9 -0
- package/scripts/embed-metrics.ts +2 -0
- package/test/PERMISSION_CHECK_TEST_SUMMARY.md +139 -0
- package/test/checkup.test.ts +1316 -2
- package/test/compose-cmd.test.ts +120 -0
- package/test/config-consistency.test.ts +321 -5
- package/test/init.integration.test.ts +27 -28
- package/test/init.test.ts +534 -6
- package/test/issues.cli.test.ts +230 -1
- package/test/mcp-server.test.ts +459 -0
- package/test/monitoring.test.ts +78 -0
- package/test/permission-check-sql.test.ts +116 -0
- package/test/reports.cli.test.ts +793 -0
- package/test/reports.test.ts +977 -0
- package/test/schema-validation.test.ts +81 -0
- package/test/storage.test.ts +761 -0
- package/test/test-utils.ts +51 -2
- package/test/upgrade.test.ts +422 -0
|
@@ -0,0 +1,14 @@
|
|
|
1
|
+
# Demo PostgreSQL instance for `pgai mon local-install --demo`
|
|
2
|
+
# This file is copied to instances.yml during demo mode setup.
|
|
3
|
+
|
|
4
|
+
- name: target_database
|
|
5
|
+
conn_str: postgresql://monitor:monitor_pass@target-db:5432/target_database
|
|
6
|
+
preset_metrics: full
|
|
7
|
+
custom_metrics:
|
|
8
|
+
is_enabled: true
|
|
9
|
+
group: default
|
|
10
|
+
custom_tags:
|
|
11
|
+
env: demo
|
|
12
|
+
cluster: local
|
|
13
|
+
node_name: node-01
|
|
14
|
+
sink_type: ~sink_type~ # sed token substituted by generate-pgwatch-sources.sh; values: postgres, prometheus
|
package/lib/checkup-api.ts
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
import * as http from "http";
|
|
1
2
|
import * as https from "https";
|
|
2
3
|
import { URL } from "url";
|
|
3
4
|
import { normalizeBaseUrl } from "./util";
|
|
@@ -27,7 +28,7 @@ function isRetryableError(err: unknown): boolean {
|
|
|
27
28
|
// Retry on server errors (5xx), not on client errors (4xx)
|
|
28
29
|
return err.statusCode >= 500 && err.statusCode < 600;
|
|
29
30
|
}
|
|
30
|
-
|
|
31
|
+
|
|
31
32
|
// Check for Node.js error codes (works on Error and Error-like objects)
|
|
32
33
|
if (typeof err === "object" && err !== null && "code" in err) {
|
|
33
34
|
const code = String((err as { code: unknown }).code);
|
|
@@ -35,7 +36,7 @@ function isRetryableError(err: unknown): boolean {
|
|
|
35
36
|
return true;
|
|
36
37
|
}
|
|
37
38
|
}
|
|
38
|
-
|
|
39
|
+
|
|
39
40
|
if (err instanceof Error) {
|
|
40
41
|
const msg = err.message.toLowerCase();
|
|
41
42
|
// Retry on network-related errors based on message content
|
|
@@ -49,7 +50,7 @@ function isRetryableError(err: unknown): boolean {
|
|
|
49
50
|
msg.includes("network")
|
|
50
51
|
);
|
|
51
52
|
}
|
|
52
|
-
|
|
53
|
+
|
|
53
54
|
return false;
|
|
54
55
|
}
|
|
55
56
|
|
|
@@ -226,7 +227,25 @@ async function postRpc<T>(params: {
|
|
|
226
227
|
resolve(value);
|
|
227
228
|
};
|
|
228
229
|
|
|
229
|
-
|
|
230
|
+
// Transport is picked from the URL protocol so the CLI can talk to a
|
|
231
|
+
// local-dev PostgREST over plain HTTP. Production URLs are always HTTPS;
|
|
232
|
+
// to guard against typos (e.g. a missing 's' in 'https://') silently
|
|
233
|
+
// leaking the API key in cleartext, refuse HTTP to non-loopback hosts
|
|
234
|
+
// unless the operator explicitly opts in via CHECKUP_ALLOW_HTTP=1.
|
|
235
|
+
if (url.protocol === "http:") {
|
|
236
|
+
// WHATWG URL keeps IPv6 literals bracketed in .hostname
|
|
237
|
+
// (e.g. `[::1]`), so strip the brackets before matching the allowlist.
|
|
238
|
+
const hostname = url.hostname.replace(/^\[|\]$/g, "");
|
|
239
|
+
const isLoopback = ["localhost", "127.0.0.1", "::1"].includes(hostname);
|
|
240
|
+
if (!isLoopback && process.env.CHECKUP_ALLOW_HTTP !== "1") {
|
|
241
|
+
throw new Error(
|
|
242
|
+
`Refusing to send API key over plaintext HTTP to '${url.host}'. ` +
|
|
243
|
+
`Use https://, a loopback hostname, or set CHECKUP_ALLOW_HTTP=1.`
|
|
244
|
+
);
|
|
245
|
+
}
|
|
246
|
+
}
|
|
247
|
+
const transport = url.protocol === "http:" ? http : https;
|
|
248
|
+
const req = transport.request(
|
|
230
249
|
url,
|
|
231
250
|
{
|
|
232
251
|
method: "POST",
|
|
@@ -277,7 +296,7 @@ async function postRpc<T>(params: {
|
|
|
277
296
|
req.destroy(); // Backup: ensure request is terminated
|
|
278
297
|
settledReject(new Error(`RPC ${rpcName} timed out after ${timeoutMs}ms (no response)`));
|
|
279
298
|
}, timeoutMs);
|
|
280
|
-
|
|
299
|
+
|
|
281
300
|
req.on("error", (err: Error) => {
|
|
282
301
|
// Handle abort as timeout (may already be rejected by timeout handler)
|
|
283
302
|
if (err.name === "AbortError" || (err as any).code === "ABORT_ERR") {
|
|
@@ -295,7 +314,7 @@ async function postRpc<T>(params: {
|
|
|
295
314
|
settledReject(err);
|
|
296
315
|
}
|
|
297
316
|
});
|
|
298
|
-
|
|
317
|
+
|
|
299
318
|
req.write(body);
|
|
300
319
|
req.end();
|
|
301
320
|
});
|
package/lib/checkup.ts
CHANGED
|
@@ -243,6 +243,50 @@ export interface RedundantIndex {
|
|
|
243
243
|
redundant_to_parse_error?: string;
|
|
244
244
|
}
|
|
245
245
|
|
|
246
|
+
/**
|
|
247
|
+
* I/O statistics by backend type (I001) - matches I001.schema.json backendIOStats
|
|
248
|
+
*/
|
|
249
|
+
export interface BackendIOStats {
|
|
250
|
+
backend_type: string;
|
|
251
|
+
reads: number;
|
|
252
|
+
/** Read MiB. The historical `_mb` suffix is retained for schema compatibility. */
|
|
253
|
+
read_bytes_mb: number;
|
|
254
|
+
read_time_ms: number;
|
|
255
|
+
writes: number;
|
|
256
|
+
/** Written MiB. The historical `_mb` suffix is retained for schema compatibility. */
|
|
257
|
+
write_bytes_mb: number;
|
|
258
|
+
write_time_ms: number;
|
|
259
|
+
writebacks: number;
|
|
260
|
+
/** Writeback MiB. The historical `_mb` suffix is retained for schema compatibility. */
|
|
261
|
+
writeback_bytes_mb: number;
|
|
262
|
+
writeback_time_ms: number;
|
|
263
|
+
fsyncs: number;
|
|
264
|
+
fsync_time_ms: number;
|
|
265
|
+
/** Relation extension operations reported by pg_stat_io for PostgreSQL 16+. */
|
|
266
|
+
extends?: number;
|
|
267
|
+
/** Extended MiB; PG16 derives extends * op_bytes, PG18+ uses native extend_bytes. */
|
|
268
|
+
extend_bytes_mb?: number;
|
|
269
|
+
hits: number;
|
|
270
|
+
evictions: number;
|
|
271
|
+
reuses: number;
|
|
272
|
+
}
|
|
273
|
+
|
|
274
|
+
/**
|
|
275
|
+
* I/O statistics analysis summary (I001)
|
|
276
|
+
*/
|
|
277
|
+
export interface IOAnalysis {
|
|
278
|
+
total_read_mb: number;
|
|
279
|
+
total_write_mb: number;
|
|
280
|
+
/** read_time_ms + write_time_ms across backends. Excludes writeback and fsync time. */
|
|
281
|
+
total_io_time_ms: number;
|
|
282
|
+
/** Buffer hit ratio: hits / (hits + reads) * 100. */
|
|
283
|
+
read_hit_ratio_pct: number;
|
|
284
|
+
/** Average read latency, or null when there are no reads. */
|
|
285
|
+
avg_read_time_ms: number | null;
|
|
286
|
+
/** Average write latency, or null when there are no writes. */
|
|
287
|
+
avg_write_time_ms: number | null;
|
|
288
|
+
}
|
|
289
|
+
|
|
246
290
|
/**
|
|
247
291
|
* Node result for reports
|
|
248
292
|
*/
|
|
@@ -284,7 +328,7 @@ export function parseVersionNum(versionNum: string): { major: string; minor: str
|
|
|
284
328
|
} catch (err) {
|
|
285
329
|
// parseInt shouldn't throw, but handle edge cases defensively
|
|
286
330
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
287
|
-
console.
|
|
331
|
+
console.error(`[parseVersionNum] Warning: Failed to parse "${versionNum}": ${errorMsg}`);
|
|
288
332
|
return { major: "", minor: "" };
|
|
289
333
|
}
|
|
290
334
|
}
|
|
@@ -729,7 +773,7 @@ export async function getStatsReset(client: Client, pgMajorVersion: number = 16)
|
|
|
729
773
|
} catch (err) {
|
|
730
774
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
731
775
|
postmasterStartupError = `Failed to query postmaster start time: ${errorMsg}`;
|
|
732
|
-
console.
|
|
776
|
+
console.error(`[getStatsReset] Warning: ${postmasterStartupError}`);
|
|
733
777
|
}
|
|
734
778
|
|
|
735
779
|
const statsResult: StatsReset = {
|
|
@@ -811,7 +855,7 @@ export async function getRedundantIndexes(client: Client, pgMajorVersion: number
|
|
|
811
855
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
812
856
|
const indexName = String(transformed.index_name || "unknown");
|
|
813
857
|
parseError = `Failed to parse redundant_to_json: ${errorMsg}`;
|
|
814
|
-
console.
|
|
858
|
+
console.error(`[H004] Warning: ${parseError} for index "${indexName}"`);
|
|
815
859
|
}
|
|
816
860
|
|
|
817
861
|
const result: RedundantIndex = {
|
|
@@ -905,7 +949,7 @@ function resolveBuildTs(): string | null {
|
|
|
905
949
|
// package.json not found is expected in some environments (e.g., bundled) - debug only
|
|
906
950
|
if (process.env.DEBUG) {
|
|
907
951
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
908
|
-
console.
|
|
952
|
+
console.error(`[resolveBuildTs] Could not stat package.json, using current time: ${errorMsg}`);
|
|
909
953
|
}
|
|
910
954
|
return new Date().toISOString();
|
|
911
955
|
}
|
|
@@ -1103,7 +1147,7 @@ async function generateD004(client: Client, nodeName: string): Promise<Report> {
|
|
|
1103
1147
|
}
|
|
1104
1148
|
} catch (err) {
|
|
1105
1149
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1106
|
-
console.
|
|
1150
|
+
console.error(`[D004] Error querying pg_stat_statements: ${errorMsg}`);
|
|
1107
1151
|
pgssError = errorMsg;
|
|
1108
1152
|
}
|
|
1109
1153
|
|
|
@@ -1156,7 +1200,7 @@ async function generateD004(client: Client, nodeName: string): Promise<Report> {
|
|
|
1156
1200
|
}
|
|
1157
1201
|
} catch (err) {
|
|
1158
1202
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1159
|
-
console.
|
|
1203
|
+
console.error(`[D004] Error querying pg_stat_kcache: ${errorMsg}`);
|
|
1160
1204
|
kcacheError = errorMsg;
|
|
1161
1205
|
}
|
|
1162
1206
|
|
|
@@ -1324,7 +1368,10 @@ async function generateF004(client: Client, nodeName: string): Promise<Report> {
|
|
|
1324
1368
|
});
|
|
1325
1369
|
} catch (err) {
|
|
1326
1370
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1327
|
-
console.
|
|
1371
|
+
console.error(`[F004] Error estimating table bloat: ${errorMsg}`);
|
|
1372
|
+
if (errorMsg.includes("postgres_ai.")) {
|
|
1373
|
+
console.error(` Hint: Run "postgresai prepare-db <connection>" to create required objects.`);
|
|
1374
|
+
}
|
|
1328
1375
|
}
|
|
1329
1376
|
|
|
1330
1377
|
// Get database info
|
|
@@ -1439,7 +1486,10 @@ async function generateF005(client: Client, nodeName: string): Promise<Report> {
|
|
|
1439
1486
|
});
|
|
1440
1487
|
} catch (err) {
|
|
1441
1488
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1442
|
-
console.
|
|
1489
|
+
console.error(`[F005] Error estimating index bloat: ${errorMsg}`);
|
|
1490
|
+
if (errorMsg.includes("postgres_ai.")) {
|
|
1491
|
+
console.error(` Hint: Run "postgresai prepare-db <connection>" to create required objects.`);
|
|
1492
|
+
}
|
|
1443
1493
|
}
|
|
1444
1494
|
|
|
1445
1495
|
// Get database info
|
|
@@ -1564,7 +1614,7 @@ async function generateG001(client: Client, nodeName: string): Promise<Report> {
|
|
|
1564
1614
|
}
|
|
1565
1615
|
} catch (err) {
|
|
1566
1616
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1567
|
-
console.
|
|
1617
|
+
console.error(`[G001] Error calculating memory usage: ${errorMsg}`);
|
|
1568
1618
|
memoryError = errorMsg;
|
|
1569
1619
|
}
|
|
1570
1620
|
|
|
@@ -1642,7 +1692,7 @@ async function generateG003(client: Client, nodeName: string): Promise<Report> {
|
|
|
1642
1692
|
}
|
|
1643
1693
|
} catch (err) {
|
|
1644
1694
|
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1645
|
-
console.
|
|
1695
|
+
console.error(`[G003] Error querying deadlock stats: ${errorMsg}`);
|
|
1646
1696
|
deadlockError = errorMsg;
|
|
1647
1697
|
}
|
|
1648
1698
|
|
|
@@ -1658,6 +1708,186 @@ async function generateG003(client: Client, nodeName: string): Promise<Report> {
|
|
|
1658
1708
|
return report;
|
|
1659
1709
|
}
|
|
1660
1710
|
|
|
1711
|
+
/**
|
|
1712
|
+
* Get I/O statistics from pg_stat_io (PostgreSQL 16+).
|
|
1713
|
+
* Uses 'pg_stat_io' metric from metrics.yml.
|
|
1714
|
+
*
|
|
1715
|
+
* @param client - Connected PostgreSQL client
|
|
1716
|
+
* @param pgMajorVersion - PostgreSQL major version; defaults to 0 so omitted versions return unavailable
|
|
1717
|
+
* @param metricSqlOverride - Optional SQL override; empty or placeholder SQL returns [] without querying
|
|
1718
|
+
* @returns Array of I/O stats by backend type, or empty array if unavailable
|
|
1719
|
+
*/
|
|
1720
|
+
export async function getIOStatistics(
|
|
1721
|
+
client: Client,
|
|
1722
|
+
pgMajorVersion: number = 0,
|
|
1723
|
+
metricSqlOverride?: string
|
|
1724
|
+
): Promise<BackendIOStats[]> {
|
|
1725
|
+
// pg_stat_io requires PostgreSQL 16+
|
|
1726
|
+
if (pgMajorVersion < 16) {
|
|
1727
|
+
return [];
|
|
1728
|
+
}
|
|
1729
|
+
|
|
1730
|
+
try {
|
|
1731
|
+
const sql = metricSqlOverride ?? getMetricSql(METRIC_NAMES.I001, pgMajorVersion);
|
|
1732
|
+
// Skip if metric returns empty/placeholder SQL
|
|
1733
|
+
if (!sql || sql.trim().startsWith(";")) {
|
|
1734
|
+
return [];
|
|
1735
|
+
}
|
|
1736
|
+
|
|
1737
|
+
const result = await client.query(sql);
|
|
1738
|
+
return result.rows.map((row) => {
|
|
1739
|
+
const transformed = transformMetricRow(row);
|
|
1740
|
+
return {
|
|
1741
|
+
backend_type: String(transformed.backend_type || "unknown"),
|
|
1742
|
+
reads: parseInt(String(transformed.reads || 0), 10),
|
|
1743
|
+
read_bytes_mb: parseInt(String(transformed.read_bytes_mb || 0), 10),
|
|
1744
|
+
read_time_ms: parseInt(String(transformed.read_time_ms || 0), 10),
|
|
1745
|
+
writes: parseInt(String(transformed.writes || 0), 10),
|
|
1746
|
+
write_bytes_mb: parseInt(String(transformed.write_bytes_mb || 0), 10),
|
|
1747
|
+
write_time_ms: parseInt(String(transformed.write_time_ms || 0), 10),
|
|
1748
|
+
writebacks: parseInt(String(transformed.writebacks || 0), 10),
|
|
1749
|
+
writeback_bytes_mb: parseInt(String(transformed.writeback_bytes_mb || 0), 10),
|
|
1750
|
+
writeback_time_ms: parseInt(String(transformed.writeback_time_ms || 0), 10),
|
|
1751
|
+
fsyncs: parseInt(String(transformed.fsyncs || 0), 10),
|
|
1752
|
+
fsync_time_ms: parseInt(String(transformed.fsync_time_ms || 0), 10),
|
|
1753
|
+
extends: parseInt(String(transformed.extends || 0), 10),
|
|
1754
|
+
extend_bytes_mb: parseInt(String(transformed.extend_bytes_mb || 0), 10),
|
|
1755
|
+
hits: parseInt(String(transformed.hits || 0), 10),
|
|
1756
|
+
evictions: parseInt(String(transformed.evictions || 0), 10),
|
|
1757
|
+
reuses: parseInt(String(transformed.reuses || 0), 10),
|
|
1758
|
+
};
|
|
1759
|
+
});
|
|
1760
|
+
} catch (err) {
|
|
1761
|
+
const errorMsg = err instanceof Error ? err.message : String(err);
|
|
1762
|
+
console.log(`[I001] Error fetching I/O statistics: ${errorMsg}`);
|
|
1763
|
+
return [];
|
|
1764
|
+
}
|
|
1765
|
+
}
|
|
1766
|
+
|
|
1767
|
+
/**
|
|
1768
|
+
* Generate I001 report - I/O statistics (pg_stat_io)
|
|
1769
|
+
*
|
|
1770
|
+
* This report collects I/O statistics from pg_stat_io (PostgreSQL 16+),
|
|
1771
|
+
* providing insights into read/write operations by backend type.
|
|
1772
|
+
*
|
|
1773
|
+
* @param client - Connected PostgreSQL client
|
|
1774
|
+
* @param nodeName - Node name for the report payload
|
|
1775
|
+
* @returns I001 report payload
|
|
1776
|
+
*/
|
|
1777
|
+
async function generateI001(client: Client, nodeName: string): Promise<Report> {
|
|
1778
|
+
const report = createBaseReport("I001", "I/O statistics (pg_stat_io)", nodeName);
|
|
1779
|
+
const postgresVersion = await getPostgresVersion(client);
|
|
1780
|
+
const parsedPgMajorVersion = parseInt(postgresVersion.server_major_ver, 10);
|
|
1781
|
+
const pgMajorVersion = Number.isFinite(parsedPgMajorVersion) ? parsedPgMajorVersion : 0;
|
|
1782
|
+
|
|
1783
|
+
// pg_stat_io requires PostgreSQL 16+
|
|
1784
|
+
if (pgMajorVersion < 16) {
|
|
1785
|
+
report.results[nodeName] = {
|
|
1786
|
+
data: {
|
|
1787
|
+
available: false,
|
|
1788
|
+
min_version_required: "16",
|
|
1789
|
+
by_backend_type: [],
|
|
1790
|
+
analysis: {
|
|
1791
|
+
total_read_mb: 0,
|
|
1792
|
+
total_write_mb: 0,
|
|
1793
|
+
total_io_time_ms: 0,
|
|
1794
|
+
read_hit_ratio_pct: 0,
|
|
1795
|
+
avg_read_time_ms: null,
|
|
1796
|
+
avg_write_time_ms: null,
|
|
1797
|
+
},
|
|
1798
|
+
stats_reset_s: null,
|
|
1799
|
+
},
|
|
1800
|
+
postgres_version: postgresVersion,
|
|
1801
|
+
};
|
|
1802
|
+
return report;
|
|
1803
|
+
}
|
|
1804
|
+
|
|
1805
|
+
const ioStats = await getIOStatistics(client, pgMajorVersion);
|
|
1806
|
+
|
|
1807
|
+
// Sort by backend_type, putting 'total' first if present
|
|
1808
|
+
ioStats.sort((a, b) => {
|
|
1809
|
+
if (a.backend_type === "total") return -1;
|
|
1810
|
+
if (b.backend_type === "total") return 1;
|
|
1811
|
+
return a.backend_type.localeCompare(b.backend_type);
|
|
1812
|
+
});
|
|
1813
|
+
|
|
1814
|
+
// Find 'total' row for analysis, or sum all rows if not present
|
|
1815
|
+
let totalStats = ioStats.find((s) => s.backend_type === "total");
|
|
1816
|
+
if (!totalStats && ioStats.length > 0) {
|
|
1817
|
+
totalStats = {
|
|
1818
|
+
backend_type: "total",
|
|
1819
|
+
reads: ioStats.reduce((sum, s) => sum + s.reads, 0),
|
|
1820
|
+
read_bytes_mb: ioStats.reduce((sum, s) => sum + s.read_bytes_mb, 0),
|
|
1821
|
+
read_time_ms: ioStats.reduce((sum, s) => sum + s.read_time_ms, 0),
|
|
1822
|
+
writes: ioStats.reduce((sum, s) => sum + s.writes, 0),
|
|
1823
|
+
write_bytes_mb: ioStats.reduce((sum, s) => sum + s.write_bytes_mb, 0),
|
|
1824
|
+
write_time_ms: ioStats.reduce((sum, s) => sum + s.write_time_ms, 0),
|
|
1825
|
+
writebacks: ioStats.reduce((sum, s) => sum + s.writebacks, 0),
|
|
1826
|
+
writeback_bytes_mb: ioStats.reduce((sum, s) => sum + s.writeback_bytes_mb, 0),
|
|
1827
|
+
writeback_time_ms: ioStats.reduce((sum, s) => sum + s.writeback_time_ms, 0),
|
|
1828
|
+
fsyncs: ioStats.reduce((sum, s) => sum + s.fsyncs, 0),
|
|
1829
|
+
fsync_time_ms: ioStats.reduce((sum, s) => sum + s.fsync_time_ms, 0),
|
|
1830
|
+
extends: ioStats.reduce((sum, s) => sum + (s.extends || 0), 0),
|
|
1831
|
+
extend_bytes_mb: ioStats.reduce((sum, s) => sum + (s.extend_bytes_mb || 0), 0),
|
|
1832
|
+
hits: ioStats.reduce((sum, s) => sum + s.hits, 0),
|
|
1833
|
+
evictions: ioStats.reduce((sum, s) => sum + s.evictions, 0),
|
|
1834
|
+
reuses: ioStats.reduce((sum, s) => sum + s.reuses, 0),
|
|
1835
|
+
};
|
|
1836
|
+
}
|
|
1837
|
+
|
|
1838
|
+
// Calculate analysis
|
|
1839
|
+
const totalReadMb = totalStats?.read_bytes_mb || 0;
|
|
1840
|
+
const totalWriteMb = totalStats?.write_bytes_mb || 0;
|
|
1841
|
+
const totalReadTime = totalStats?.read_time_ms || 0;
|
|
1842
|
+
const totalWriteTime = totalStats?.write_time_ms || 0;
|
|
1843
|
+
const totalIoTimeMs = totalReadTime + totalWriteTime;
|
|
1844
|
+
const totalReads = totalStats?.reads || 0;
|
|
1845
|
+
const totalWrites = totalStats?.writes || 0;
|
|
1846
|
+
const totalHits = totalStats?.hits || 0;
|
|
1847
|
+
|
|
1848
|
+
// Hit ratio: hits / (hits + reads) * 100
|
|
1849
|
+
const totalRequests = totalHits + totalReads;
|
|
1850
|
+
const readHitRatioPct = totalRequests > 0 ? Math.round((totalHits / totalRequests) * 10000) / 100 : 0;
|
|
1851
|
+
|
|
1852
|
+
// Average times
|
|
1853
|
+
const avgReadTimeMs = totalReads > 0 ? Math.round((totalReadTime / totalReads) * 1000) / 1000 : null;
|
|
1854
|
+
const avgWriteTimeMs = totalWrites > 0 ? Math.round((totalWriteTime / totalWrites) * 1000) / 1000 : null;
|
|
1855
|
+
|
|
1856
|
+
// Direct-connect checkup queries stats_reset separately instead of reading it from pgwatch metrics.
|
|
1857
|
+
let statsResetS: number | null = null;
|
|
1858
|
+
try {
|
|
1859
|
+
const resetResult = await client.query(`
|
|
1860
|
+
select max(extract(epoch from now() - stats_reset)::int) as stats_reset_s
|
|
1861
|
+
from pg_stat_io
|
|
1862
|
+
`);
|
|
1863
|
+
if (resetResult.rows.length > 0 && resetResult.rows[0].stats_reset_s !== null) {
|
|
1864
|
+
const parsedStatsResetS = parseInt(resetResult.rows[0].stats_reset_s, 10);
|
|
1865
|
+
statsResetS = Number.isFinite(parsedStatsResetS) ? parsedStatsResetS : null;
|
|
1866
|
+
}
|
|
1867
|
+
} catch (err) {
|
|
1868
|
+
// Ignore errors getting stats_reset - not critical
|
|
1869
|
+
}
|
|
1870
|
+
|
|
1871
|
+
report.results[nodeName] = {
|
|
1872
|
+
data: {
|
|
1873
|
+
available: ioStats.length > 0,
|
|
1874
|
+
by_backend_type: ioStats,
|
|
1875
|
+
analysis: {
|
|
1876
|
+
total_read_mb: totalReadMb,
|
|
1877
|
+
total_write_mb: totalWriteMb,
|
|
1878
|
+
total_io_time_ms: totalIoTimeMs,
|
|
1879
|
+
read_hit_ratio_pct: readHitRatioPct,
|
|
1880
|
+
avg_read_time_ms: avgReadTimeMs,
|
|
1881
|
+
avg_write_time_ms: avgWriteTimeMs,
|
|
1882
|
+
},
|
|
1883
|
+
stats_reset_s: statsResetS,
|
|
1884
|
+
},
|
|
1885
|
+
postgres_version: postgresVersion,
|
|
1886
|
+
};
|
|
1887
|
+
|
|
1888
|
+
return report;
|
|
1889
|
+
}
|
|
1890
|
+
|
|
1661
1891
|
/**
|
|
1662
1892
|
* Available report generators
|
|
1663
1893
|
*/
|
|
@@ -1677,6 +1907,7 @@ export const REPORT_GENERATORS: Record<string, (client: Client, nodeName: string
|
|
|
1677
1907
|
H001: generateH001,
|
|
1678
1908
|
H002: generateH002,
|
|
1679
1909
|
H004: generateH004,
|
|
1910
|
+
I001: generateI001,
|
|
1680
1911
|
};
|
|
1681
1912
|
|
|
1682
1913
|
/**
|
package/lib/config.ts
CHANGED
|
@@ -8,6 +8,7 @@ import * as os from "os";
|
|
|
8
8
|
export interface Config {
|
|
9
9
|
apiKey: string | null;
|
|
10
10
|
baseUrl: string | null;
|
|
11
|
+
storageBaseUrl: string | null;
|
|
11
12
|
orgId: number | null;
|
|
12
13
|
defaultProject: string | null;
|
|
13
14
|
/** Docker Compose project name for monitoring stack */
|
|
@@ -48,6 +49,7 @@ export function readConfig(): Config {
|
|
|
48
49
|
const config: Config = {
|
|
49
50
|
apiKey: null,
|
|
50
51
|
baseUrl: null,
|
|
52
|
+
storageBaseUrl: null,
|
|
51
53
|
orgId: null,
|
|
52
54
|
defaultProject: null,
|
|
53
55
|
projectName: null,
|
|
@@ -61,6 +63,7 @@ export function readConfig(): Config {
|
|
|
61
63
|
const parsed = JSON.parse(content);
|
|
62
64
|
config.apiKey = parsed.apiKey ?? null;
|
|
63
65
|
config.baseUrl = parsed.baseUrl ?? null;
|
|
66
|
+
config.storageBaseUrl = parsed.storageBaseUrl ?? null;
|
|
64
67
|
config.orgId = parsed.orgId ?? null;
|
|
65
68
|
config.defaultProject = parsed.defaultProject ?? null;
|
|
66
69
|
config.projectName = parsed.projectName ?? null;
|
package/lib/init.ts
CHANGED
|
@@ -127,8 +127,10 @@ export async function connectWithSslFallback(
|
|
|
127
127
|
verbose?: boolean
|
|
128
128
|
): Promise<{ client: PgClient; usedSsl: boolean }> {
|
|
129
129
|
const tryConnect = async (config: PgClientConfig): Promise<PgClient> => {
|
|
130
|
-
const client = new ClientClass(config);
|
|
130
|
+
const client = new ClientClass({ ...config, connectionTimeoutMillis: 10_000 } as any);
|
|
131
131
|
await client.connect();
|
|
132
|
+
// Set a default statement timeout to prevent runaway queries
|
|
133
|
+
await client.query("SET statement_timeout = '30s'");
|
|
132
134
|
return client;
|
|
133
135
|
};
|
|
134
136
|
|
|
@@ -149,7 +151,7 @@ export async function connectWithSslFallback(
|
|
|
149
151
|
}
|
|
150
152
|
|
|
151
153
|
if (verbose) {
|
|
152
|
-
console.
|
|
154
|
+
console.error("SSL connection failed, retrying without SSL...");
|
|
153
155
|
}
|
|
154
156
|
|
|
155
157
|
// Retry without SSL
|
|
@@ -454,8 +456,18 @@ export function resolveAdminConnection(opts: {
|
|
|
454
456
|
return { clientConfig: cfg, display: describePgConfig(cfg), sslFallbackEnabled: true };
|
|
455
457
|
}
|
|
456
458
|
|
|
459
|
+
/**
|
|
460
|
+
* Generate a cryptographically secure random password for the monitoring role.
|
|
461
|
+
*
|
|
462
|
+
* Encoding note — bytes vs output length:
|
|
463
|
+
* - hex: N bytes → 2N characters (24 bytes → 48 hex chars)
|
|
464
|
+
* - base64: N bytes → ⌈4N/3⌉ chars (24 bytes → 32 base64url chars, no padding)
|
|
465
|
+
*
|
|
466
|
+
* We use base64url (RFC 4648 §5) because it is shorter than hex and safe in URLs,
|
|
467
|
+
* connection strings, and shell variables without quoting.
|
|
468
|
+
*/
|
|
457
469
|
function generateMonitoringPassword(): string {
|
|
458
|
-
//
|
|
470
|
+
// 24 random bytes → 32 base64url characters (no padding).
|
|
459
471
|
// Note: randomBytes() throws on failure; we add a tiny sanity check for unexpected output.
|
|
460
472
|
const password = randomBytes(24).toString("base64url");
|
|
461
473
|
if (password.length < 30) {
|
|
@@ -659,6 +671,36 @@ export type VerifyInitResult = {
|
|
|
659
671
|
missingOptional: string[];
|
|
660
672
|
};
|
|
661
673
|
|
|
674
|
+
/** A single permission check result from the preflight query. */
|
|
675
|
+
export type PermissionCheckRow = {
|
|
676
|
+
permission_name: string;
|
|
677
|
+
status: "required" | "optional";
|
|
678
|
+
/**
|
|
679
|
+
* Whether the permission is granted.
|
|
680
|
+
* - `true` — permission is granted
|
|
681
|
+
* - `false` — permission is explicitly denied
|
|
682
|
+
* - `null` — check was skipped (e.g., object does not exist, so the privilege
|
|
683
|
+
* check is inapplicable — such as SELECT on a view that hasn't been created)
|
|
684
|
+
*/
|
|
685
|
+
granted: boolean | null;
|
|
686
|
+
fix_command: string | null;
|
|
687
|
+
};
|
|
688
|
+
|
|
689
|
+
/**
|
|
690
|
+
* Result of the preflight permission check for the current DB user.
|
|
691
|
+
*
|
|
692
|
+
* - `ok` is `true` when `missingRequired` is empty.
|
|
693
|
+
* - `rows` contains every check (for inspection / logging).
|
|
694
|
+
* - `missingRequired` / `missingOptional` are filtered subsets of `rows`
|
|
695
|
+
* where the permission is not granted (`granted !== true`).
|
|
696
|
+
*/
|
|
697
|
+
export type PreflightPermissionResult = {
|
|
698
|
+
ok: boolean;
|
|
699
|
+
rows: PermissionCheckRow[];
|
|
700
|
+
missingRequired: PermissionCheckRow[];
|
|
701
|
+
missingOptional: PermissionCheckRow[];
|
|
702
|
+
};
|
|
703
|
+
|
|
662
704
|
export type UninitPlan = {
|
|
663
705
|
monitoringUser: string;
|
|
664
706
|
database: string;
|
|
@@ -813,7 +855,12 @@ export async function verifyInitSetup(params: {
|
|
|
813
855
|
missingRequired.push("USAGE on schema postgres_ai");
|
|
814
856
|
}
|
|
815
857
|
|
|
816
|
-
const viewExistsRes = await params.client.query(
|
|
858
|
+
const viewExistsRes = await params.client.query(`
|
|
859
|
+
select case
|
|
860
|
+
when not has_schema_privilege(current_user, 'postgres_ai', 'USAGE') then null
|
|
861
|
+
else to_regclass('postgres_ai.pg_statistic') is not null
|
|
862
|
+
end as ok
|
|
863
|
+
`);
|
|
817
864
|
if (!viewExistsRes.rows?.[0]?.ok) {
|
|
818
865
|
missingRequired.push("view postgres_ai.pg_statistic exists");
|
|
819
866
|
} else {
|
|
@@ -936,4 +983,149 @@ export async function verifyInitSetup(params: {
|
|
|
936
983
|
}
|
|
937
984
|
}
|
|
938
985
|
|
|
986
|
+
/**
|
|
987
|
+
* Check that the currently connected DB user has sufficient permissions for
|
|
988
|
+
* monitoring operations. Returns structured results with fix commands.
|
|
989
|
+
*
|
|
990
|
+
* Required permissions cause startup to fail; optional ones produce warnings.
|
|
991
|
+
*
|
|
992
|
+
* @param client An already-connected PostgreSQL client.
|
|
993
|
+
* @returns A {@link PreflightPermissionResult} with per-check rows and
|
|
994
|
+
* filtered `missingRequired` / `missingOptional` arrays.
|
|
995
|
+
* @throws Propagates database errors (network, permission denied on catalog
|
|
996
|
+
* tables, timeout) to the caller.
|
|
997
|
+
*/
|
|
998
|
+
export async function checkCurrentUserPermissions(
|
|
999
|
+
client: PgClient
|
|
1000
|
+
): Promise<PreflightPermissionResult> {
|
|
1001
|
+
const sql = `
|
|
1002
|
+
with permission_checks as (
|
|
1003
|
+
select
|
|
1004
|
+
format('connect on database %I', current_database()) as permission_name,
|
|
1005
|
+
'required' as status,
|
|
1006
|
+
has_database_privilege(current_user, current_database(), 'connect') as granted
|
|
1007
|
+
|
|
1008
|
+
union all
|
|
1009
|
+
|
|
1010
|
+
select
|
|
1011
|
+
'pg_monitor role membership' as permission_name,
|
|
1012
|
+
'required' as status,
|
|
1013
|
+
-- CASE guarantees evaluation order: pg_has_role() is only called if the
|
|
1014
|
+
-- pg_monitor role exists, avoiding ERROR on PostgreSQL < 10 or when dropped.
|
|
1015
|
+
case
|
|
1016
|
+
when not exists (select from pg_roles where rolname = 'pg_monitor')
|
|
1017
|
+
then false
|
|
1018
|
+
else pg_has_role(current_user, 'pg_monitor', 'member')
|
|
1019
|
+
end as granted
|
|
1020
|
+
|
|
1021
|
+
union all
|
|
1022
|
+
|
|
1023
|
+
select
|
|
1024
|
+
'select on pg_catalog.pg_index' as permission_name,
|
|
1025
|
+
'required' as status,
|
|
1026
|
+
has_table_privilege(current_user, 'pg_catalog.pg_index', 'select') as granted
|
|
1027
|
+
|
|
1028
|
+
union all
|
|
1029
|
+
|
|
1030
|
+
select
|
|
1031
|
+
'postgres_ai.pg_statistic view exists' as permission_name,
|
|
1032
|
+
'optional' as status,
|
|
1033
|
+
case
|
|
1034
|
+
when not has_schema_privilege(current_user, 'postgres_ai', 'USAGE') then null
|
|
1035
|
+
else to_regclass('postgres_ai.pg_statistic') is not null
|
|
1036
|
+
end as granted
|
|
1037
|
+
|
|
1038
|
+
union all
|
|
1039
|
+
|
|
1040
|
+
select
|
|
1041
|
+
'select on postgres_ai.pg_statistic' as permission_name,
|
|
1042
|
+
'optional' as status,
|
|
1043
|
+
case
|
|
1044
|
+
when not has_schema_privilege(current_user, 'postgres_ai', 'USAGE') then null
|
|
1045
|
+
when to_regclass('postgres_ai.pg_statistic') is null then null
|
|
1046
|
+
else has_table_privilege(current_user, 'postgres_ai.pg_statistic', 'select')
|
|
1047
|
+
end as granted
|
|
1048
|
+
)
|
|
1049
|
+
select
|
|
1050
|
+
permission_name,
|
|
1051
|
+
status,
|
|
1052
|
+
granted,
|
|
1053
|
+
case
|
|
1054
|
+
when status = 'required' and not coalesce(granted, false) then
|
|
1055
|
+
case
|
|
1056
|
+
when permission_name like 'connect%' then
|
|
1057
|
+
format('grant connect on database %I to %I;', current_database(), current_user)
|
|
1058
|
+
when permission_name = 'pg_monitor role membership' then
|
|
1059
|
+
format('grant pg_monitor to %I;', current_user)
|
|
1060
|
+
when permission_name like 'select on pg_catalog.pg_index' then
|
|
1061
|
+
format('grant select on pg_catalog.pg_index to %I;', current_user)
|
|
1062
|
+
end
|
|
1063
|
+
when permission_name = 'postgres_ai.pg_statistic view exists' and granted = false then
|
|
1064
|
+
'-- create postgres_ai.pg_statistic view (see setup script)'
|
|
1065
|
+
when permission_name = 'select on postgres_ai.pg_statistic' and granted = false then
|
|
1066
|
+
format('grant select on postgres_ai.pg_statistic to %I;', current_user)
|
|
1067
|
+
else null
|
|
1068
|
+
end as fix_command
|
|
1069
|
+
from permission_checks
|
|
1070
|
+
order by
|
|
1071
|
+
case status when 'required' then 1 else 2 end,
|
|
1072
|
+
permission_name;
|
|
1073
|
+
`;
|
|
1074
|
+
|
|
1075
|
+
const res = await client.query(sql);
|
|
1076
|
+
const rows: PermissionCheckRow[] = res.rows;
|
|
1077
|
+
|
|
1078
|
+
// Required: treat null (skipped) as not-granted — fail safe.
|
|
1079
|
+
// Optional: only explicit false counts as missing; null means the check was
|
|
1080
|
+
// skipped (e.g., view doesn't exist) and is not actionable.
|
|
1081
|
+
const missingRequired = rows.filter((r) => r.status === "required" && r.granted !== true);
|
|
1082
|
+
const missingOptional = rows.filter((r) => r.status === "optional" && r.granted === false);
|
|
1083
|
+
|
|
1084
|
+
return {
|
|
1085
|
+
ok: missingRequired.length === 0,
|
|
1086
|
+
rows,
|
|
1087
|
+
missingRequired,
|
|
1088
|
+
missingOptional,
|
|
1089
|
+
};
|
|
1090
|
+
}
|
|
1091
|
+
|
|
1092
|
+
/**
|
|
1093
|
+
* Format permission check results into user-facing error/warning lines.
|
|
1094
|
+
*
|
|
1095
|
+
* @returns An object with `warnings` (for optional misses), `errors` (for
|
|
1096
|
+
* required misses including fix SQL), and `failed` (whether required
|
|
1097
|
+
* permissions are missing).
|
|
1098
|
+
*/
|
|
1099
|
+
export function formatPermissionCheckMessages(result: PreflightPermissionResult): {
|
|
1100
|
+
failed: boolean;
|
|
1101
|
+
warnings: string[];
|
|
1102
|
+
errors: string[];
|
|
1103
|
+
} {
|
|
1104
|
+
const warnings: string[] = [];
|
|
1105
|
+
const errors: string[] = [];
|
|
1106
|
+
|
|
1107
|
+
for (const row of result.missingOptional) {
|
|
1108
|
+
const fix = row.fix_command ? ` Fix: ${row.fix_command}` : "";
|
|
1109
|
+
warnings.push(`Warning: optional permission missing — ${row.permission_name}.${fix}`);
|
|
1110
|
+
}
|
|
939
1111
|
|
|
1112
|
+
if (!result.ok) {
|
|
1113
|
+
errors.push("Error: the database user is missing required permissions.\n");
|
|
1114
|
+
errors.push("Missing permissions:");
|
|
1115
|
+
for (const row of result.missingRequired) {
|
|
1116
|
+
errors.push(` - ${row.permission_name}`);
|
|
1117
|
+
}
|
|
1118
|
+
const fixes = result.missingRequired
|
|
1119
|
+
.map((r) => r.fix_command)
|
|
1120
|
+
.filter(Boolean);
|
|
1121
|
+
if (fixes.length > 0) {
|
|
1122
|
+
errors.push("\nTo fix, run the following as a superuser:\n");
|
|
1123
|
+
for (const fix of fixes) {
|
|
1124
|
+
errors.push(` ${fix}`);
|
|
1125
|
+
}
|
|
1126
|
+
}
|
|
1127
|
+
errors.push("\nAlternatively, run 'postgresai prepare-db' to set up permissions automatically.");
|
|
1128
|
+
}
|
|
1129
|
+
|
|
1130
|
+
return { failed: !result.ok, warnings, errors };
|
|
1131
|
+
}
|