Compare commits

..
Author SHA1 Message Date
Garry TanandClaude Fable 5 d98e6b507f test(config): use withEnv() for TTL env mutation to satisfy check-test-isolation R1
CI verify failed: test/postgres-engine-config-cache.test.ts mutated
process.env.GBRAIN_CONFIG_CACHE_TTL_MS directly in beforeEach/afterEach.
Wrap each test body in withEnv() (test/helpers/with-env.ts) instead.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-22 10:58:53 -07:00
d4bbf6eeae perf(config): batch + cache engine.getConfig to kill ~85 round-trips per query
Takeover of #1694: a single search fires ~85 serial getConfig() reads
(loadConfigWithEngine x2 plus the mode/cache/intent/rerank/graph-signals
resolvers); on a remote pooler each read is a round-trip, dominating query
latency and risking cli.ts's 10s disconnect force-exit truncating stdout.

The first read now batch-loads the whole config table into a process-
lifetime Map (single-flight under concurrency); setConfig/unsetConfig
write through; a 30s TTL bounds multi-writer staleness and
GBRAIN_CONFIG_CACHE_TTL_MS=0 restores per-key reads.

Rebased onto current master: unlike the original diff, both the batch
load and the TTL=0 per-key fallback stay inside connRetry() so the
#1603/#1891 retry+reconnect posture (pooler-drop self-heal) is preserved.
PGLite stays uncached (in-process, zero round-trips; raw-SQL test
fixtures rely on fresh reads). E2E helpers pin the cache off since those
suites seed config via raw SQL.

Co-authored-by: Omerbahari <Omerbahari@users.noreply.github.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-21 14:26:28 -07:00
9 changed files with 189 additions and 141 deletions
+5 -8
View File
@@ -21,17 +21,14 @@ GBrain is tuned for the Supabase **Transaction pooler** (port 6543): it
auto-disables prepared statements there and routes `engine.transaction()`
(migrations, DDL, sync imports) to a derived **direct** connection
(`db.<ref>.supabase.co:5432`). That direct host is IPv6-only, so on an
IPv4-only host it is unreachable. When that happens gbrain now falls back to
the pooler automatically (one stderr warning, then single-pool mode for the
rest of the process) — but the pooler's ~2-min statement timeout can truncate
very long migrations or bulk imports.
IPv4-only host, reads work but sync **silently skips most pages**. This is the
number one cause of "sync ran but nothing happened."
Fix: make the direct connection reachable over IPv4. Either set
`GBRAIN_DIRECT_DATABASE_URL` to the **Session pooler** string (port 5432 on the
`pooler.supabase.com` host, IPv4), or enable Supabase's IPv4 add-on.
`GBRAIN_DISABLE_DIRECT_POOL=1` skips the direct pool (and the fallback warning)
entirely. Verify by running `gbrain sync` and checking that the page count in
`gbrain stats` matches the syncable file count in the repo.
`pooler.supabase.com` host, IPv4), or enable Supabase's IPv4 add-on. Verify by
running `gbrain sync` and checking that the page count in `gbrain stats` matches
the syncable file count in the repo.
### The Primitives
+5 -8
View File
@@ -2720,17 +2720,14 @@ GBrain is tuned for the Supabase **Transaction pooler** (port 6543): it
auto-disables prepared statements there and routes `engine.transaction()`
(migrations, DDL, sync imports) to a derived **direct** connection
(`db.<ref>.supabase.co:5432`). That direct host is IPv6-only, so on an
IPv4-only host it is unreachable. When that happens gbrain now falls back to
the pooler automatically (one stderr warning, then single-pool mode for the
rest of the process) — but the pooler's ~2-min statement timeout can truncate
very long migrations or bulk imports.
IPv4-only host, reads work but sync **silently skips most pages**. This is the
number one cause of "sync ran but nothing happened."
Fix: make the direct connection reachable over IPv4. Either set
`GBRAIN_DIRECT_DATABASE_URL` to the **Session pooler** string (port 5432 on the
`pooler.supabase.com` host, IPv4), or enable Supabase's IPv4 add-on.
`GBRAIN_DISABLE_DIRECT_POOL=1` skips the direct pool (and the fallback warning)
entirely. Verify by running `gbrain sync` and checking that the page count in
`gbrain stats` matches the syncable file count in the repo.
`pooler.supabase.com` host, IPv4), or enable Supabase's IPv4 add-on. Verify by
running `gbrain sync` and checking that the page count in `gbrain stats` matches
the syncable file count in the repo.
### The Primitives
-6
View File
@@ -1078,9 +1078,6 @@ async function initPostgres(opts: {
console.warn(' Direct connections are IPv6 only and fail in many environments.');
console.warn(' Use the Transaction pooler connection string instead (port 6543):');
console.warn(' Supabase Dashboard > Connect (top bar) > Connection String > Transaction pooler');
console.warn(' (With a pooler URL, gbrain derives a direct connection for DDL and falls back');
console.warn(' to the pooler automatically if that host is unreachable. Power users:');
console.warn(' GBRAIN_DIRECT_DATABASE_URL overrides the derived URL; GBRAIN_DISABLE_DIRECT_POOL=1 disables it.)');
console.warn('');
}
@@ -1094,9 +1091,6 @@ async function initPostgres(opts: {
if (databaseUrl.includes('supabase.co') && (msg.includes('ECONNREFUSED') || msg.includes('ETIMEDOUT'))) {
console.error('Connection failed. Supabase direct connections (db.*.supabase.co:5432) are IPv6 only.');
console.error('Use the Transaction pooler connection string instead (port 6543).');
console.error('(gbrain derives its own direct connection from pooler URLs for DDL; if that host is');
console.error('unreachable it falls back to the pooler. GBRAIN_DIRECT_DATABASE_URL overrides the');
console.error('derived URL; GBRAIN_DISABLE_DIRECT_POOL=1 disables the direct pool entirely.)');
}
throw e;
}
+2 -48
View File
@@ -167,25 +167,6 @@ export function deriveDirectUrl(url: string): string | null {
}
}
/**
* Error codes that mean "the direct host is unreachable from this network"
* (#1641). The auto-derived db.<ref>.supabase.co host is IPv6-only without
* the paid IPv4 add-on, so ENOTFOUND/ECONNREFUSED here is expected on
* IPv4-only networks — we fall back to the pooler instead of failing init.
*/
const NETWORK_UNREACHABLE_CODES = [
'ENOTFOUND', 'ECONNREFUSED', 'ENETUNREACH', 'EHOSTUNREACH',
'ETIMEDOUT', 'CONNECT_TIMEOUT',
];
/** True when err looks like a network-unreachable failure (not auth/SQL). */
export function isNetworkUnreachableError(err: unknown): boolean {
const code = (err as { code?: unknown } | null)?.code;
if (typeof code === 'string' && NETWORK_UNREACHABLE_CODES.includes(code)) return true;
const msg = err instanceof Error ? err.message : String(err);
return NETWORK_UNREACHABLE_CODES.some(c => msg.includes(c));
}
/**
* Read kill-switch state from env. Subordinate to parent manager's state
* when present (A2 inheritance).
@@ -338,30 +319,7 @@ export class ConnectionManager {
throw err;
});
}
let pool: Sql | null;
try {
pool = await this._directInit;
} catch (err) {
// #1641: the derived direct host (db.<ref>.supabase.co) is IPv6-only
// without Supabase's IPv4 add-on. On IPv4-only networks the direct
// pool can never connect — permanently fall back to the read pool
// (self-activating kill-switch) instead of failing init/migrations.
// Non-network errors (auth, SQL) still throw: they mean misconfig,
// not unreachability.
if (isNetworkUnreachableError(err)) {
const alreadyWarned = this._killSwitch;
this._killSwitch = true;
const msg = err instanceof Error ? err.message : String(err);
if (!alreadyWarned) console.error(
`gbrain: direct connection to ${this._directUrl ? this.hostOnly(this._directUrl) : 'unknown host'} unreachable (${msg}); ` +
'falling back to the pooler for DDL/bulk (long migrations may hit the pooler statement timeout). ' +
'Set GBRAIN_DIRECT_DATABASE_URL to a reachable direct URL (e.g. the Session pooler, port 5432) or enable the Supabase IPv4 add-on; ' +
'GBRAIN_DISABLE_DIRECT_POOL=1 silences this.',
);
return this.getReadPool();
}
throw err;
}
const pool = await this._directInit;
if (!pool) {
// Defensive — initDirectPool should have thrown.
throw new Error('connection-manager: direct pool init returned null');
@@ -392,9 +350,8 @@ export class ConnectionManager {
},
};
const t0 = Date.now();
let pool: Sql | null = null;
try {
pool = postgres(this._directUrl, opts);
const pool = postgres(this._directUrl, opts);
// Probe to validate connectivity early.
await pool`SELECT 1`;
logConnectionEvent({
@@ -405,9 +362,6 @@ export class ConnectionManager {
});
return pool;
} catch (err) {
// Don't leak the failed pool's sockets/timers (#1641 fallback keeps
// the process running afterward).
if (pool) await endPoolBounded(pool);
logConnectionEvent({
pool: 'ddl',
op: 'error',
+55 -7
View File
@@ -5565,30 +5565,78 @@ export class PostgresEngine implements BrainEngine {
});
}
/**
* perf (#1694 by @Omerbahari): process-lifetime config cache. A single
* search fires ~85 getConfig() reads (loadConfigWithEngine x2, plus
* mode/cache/intent/rerank/graph-signals resolvers). On a remote pooler
* each read is a round-trip; serial they dominate query latency and can
* push the op handler past cli.ts's 10s disconnect force-exit, truncating
* stdout. First read batch-loads the whole `config` table into this Map
* (inside the same connRetry posture as the per-key read #1603/#1891);
* setConfig/unsetConfig write through. TTL bounds staleness for
* multi-writer processes; GBRAIN_CONFIG_CACHE_TTL_MS=0 disables.
* Only present keys are stored Map.has() distinguishes known-absent.
*/
private _configCache: Map<string, string> | null = null;
private _configCacheLoadedAt = 0;
private _configCacheLoad: Promise<void> | null = null;
private get _configCacheTtlMs(): number {
const raw = process.env.GBRAIN_CONFIG_CACHE_TTL_MS;
if (raw !== undefined) {
const n = parseInt(raw, 10);
if (Number.isFinite(n) && n >= 0) return n;
}
return 30_000;
}
async getConfig(key: string): Promise<string | null> {
// #1603: a transient pooler drop on this read used to throw / fall through
// to defaults silently — which on remote Postgres surfaces as the wrong
// search mode/knobs and empty-stdout queries.
return this.connRetry(async () => {
const rows = await this.sql`SELECT value FROM config WHERE key = ${key}`;
return rows.length > 0 ? (rows[0].value as string) : null;
});
// search mode/knobs and empty-stdout queries. Both the batch load and the
// cache-off per-key read keep the connRetry reconnect posture.
const ttl = this._configCacheTtlMs;
if (ttl === 0) {
return this.connRetry(async () => {
const rows = await this.sql`SELECT value FROM config WHERE key = ${key}`;
return rows.length > 0 ? (rows[0].value as string) : null;
});
}
if (this._configCache === null || Date.now() - this._configCacheLoadedAt >= ttl) {
// Single-flight: concurrent cold reads share one batch load.
this._configCacheLoad ??= this.connRetry(async () => {
const rows = await this.sql`SELECT key, value FROM config` as unknown as
Array<{ key: string; value: string | null }>;
const map = new Map<string, string>();
for (const r of rows) if (r.value != null) map.set(r.key, r.value);
this._configCache = map;
this._configCacheLoadedAt = Date.now();
}).finally(() => {
this._configCacheLoad = null;
});
await this._configCacheLoad;
}
return this._configCache!.has(key) ? this._configCache!.get(key)! : null;
}
async setConfig(key: string, value: string): Promise<void> {
return this.connRetry(async () => {
await this.connRetry(async () => {
await this.sql`
INSERT INTO config (key, value) VALUES (${key}, ${value})
ON CONFLICT (key) DO UPDATE SET value = EXCLUDED.value
`;
});
// Write-through so a long-lived process never serves stale config.
this._configCache?.set(key, value);
}
async unsetConfig(key: string): Promise<number> {
return this.connRetry(async () => {
const count = await this.connRetry(async () => {
const result = await this.sql`DELETE FROM config WHERE key = ${key}` as unknown as { count: number };
return result.count ?? 0;
});
// Write-through: known-absent, so the cache doesn't serve a stale value.
this._configCache?.delete(key);
return count;
}
async listConfigKeys(prefix: string): Promise<string[]> {
-63
View File
@@ -3,7 +3,6 @@ import {
isSupabasePoolerUrl,
deriveDirectUrl,
readKillSwitchEnv,
isNetworkUnreachableError,
resolveDirectPoolSize,
ConnectionManager,
DEFAULT_DIRECT_POOL_SIZE,
@@ -239,65 +238,3 @@ describe('ConnectionManager — parent inheritance (A2)', () => {
}
});
});
describe('isNetworkUnreachableError (#1641)', () => {
test('classifies network codes as unreachable', () => {
for (const code of ['ENOTFOUND', 'ECONNREFUSED', 'ENETUNREACH', 'EHOSTUNREACH', 'ETIMEDOUT', 'CONNECT_TIMEOUT']) {
const err = Object.assign(new Error('connect failed'), { code });
expect(isNetworkUnreachableError(err)).toBe(true);
}
});
test('classifies by message when code absent', () => {
expect(isNetworkUnreachableError(new Error('getaddrinfo ENOTFOUND db.abc.supabase.co'))).toBe(true);
});
test('auth/SQL errors are NOT unreachable', () => {
expect(isNetworkUnreachableError(new Error('password authentication failed for user "postgres"'))).toBe(false);
expect(isNetworkUnreachableError(new Error('syntax error at or near "SELEC"'))).toBe(false);
expect(isNetworkUnreachableError(null)).toBe(false);
});
});
describe('ConnectionManager — direct-pool fallback on unreachable host (#1641)', () => {
let originalKillSwitch: string | undefined;
let originalError: typeof console.error;
let errLines: string[];
beforeEach(() => {
originalKillSwitch = process.env.GBRAIN_DISABLE_DIRECT_POOL;
delete process.env.GBRAIN_DISABLE_DIRECT_POOL;
originalError = console.error;
errLines = [];
console.error = (...args: unknown[]) => { errLines.push(args.join(' ')); };
});
afterEach(() => {
console.error = originalError;
if (originalKillSwitch === undefined) delete process.env.GBRAIN_DISABLE_DIRECT_POOL;
else process.env.GBRAIN_DISABLE_DIRECT_POOL = originalKillSwitch;
});
test('ddl() falls back to the read pool when the direct host is unreachable', async () => {
const cm = new ConnectionManager({
url: 'postgresql://postgres.abc:p@aws.pooler.supabase.com:6543/db',
// 127.0.0.1:9 (discard) → instant ECONNREFUSED, the IPv4-only-network shape.
directUrl: 'postgresql://postgres:p@127.0.0.1:9/db',
});
const fakeReadPool = {} as ReturnType<typeof ConnectionManager.prototype.read>;
cm.setReadPool(fakeReadPool);
expect(cm.isDualPoolActive()).toBe(true);
const pool = await cm.ddl(); // without the fix this throws ECONNREFUSED
expect(pool).toBe(fakeReadPool);
// Self-activating kill-switch: subsequent calls skip the direct pool.
expect(cm.isKillSwitchActive()).toBe(true);
expect(cm.isDualPoolActive()).toBe(false);
expect(cm.describeMode().mode).toBe('single (kill-switch)');
// One stderr line mentioning the power-user override.
const warning = errLines.filter(l => l.includes('GBRAIN_DIRECT_DATABASE_URL'));
expect(warning.length).toBe(1);
const again = await cm.ddl();
expect(again).toBe(fakeReadPool);
expect(errLines.filter(l => l.includes('GBRAIN_DIRECT_DATABASE_URL')).length).toBe(1);
}, 20000);
});
+7
View File
@@ -29,6 +29,13 @@ if (existsSync(envPath)) {
}
}
// E2E suites seed/rewrite the config table via raw SQL and expect engine
// reads to see it immediately; disable the process-lifetime config cache
// (#1694) so read semantics match pre-cache behavior. Spawned CLI
// subprocesses inherit this. Cache semantics are pinned by
// test/postgres-engine-config-cache.test.ts.
process.env.GBRAIN_CONFIG_CACHE_TTL_MS ??= '0';
const DATABASE_URL = process.env.DATABASE_URL;
const FIXTURES_DIR = resolve(import.meta.dir, 'fixtures');
+112
View File
@@ -0,0 +1,112 @@
/**
* Process-lifetime config cache (#1694 takeover, by @Omerbahari).
*
* A single search fires ~85 getConfig() reads; on a remote pooler each is a
* round-trip. The first read now batch-loads the whole `config` table into a
* Map; setConfig/unsetConfig write through; TTL bounds multi-writer
* staleness; GBRAIN_CONFIG_CACHE_TTL_MS=0 restores per-key reads.
*
* Pure: stubs `_sql` with a call-counting fake; no real DB.
*/
import { describe, it, expect } from 'bun:test';
import { PostgresEngine } from '../src/core/postgres-engine.ts';
import { withEnv } from './helpers/with-env.ts';
const FAST_RETRY = { maxRetries: 3, delayMs: 1, delayMaxMs: 1, jitter: 'none' as const };
/** Engine whose `sql` records every query's template strings and returns `rows`. */
function makeEngine(rows: unknown[]) {
const e = new PostgresEngine();
const calls: string[] = [];
(e as unknown as { _connectionStyle: string })._connectionStyle = 'instance';
(e as unknown as { _bulkRetryOptsCache: unknown })._bulkRetryOptsCache = FAST_RETRY;
(e as unknown as { _sql: unknown })._sql = (strings: TemplateStringsArray) => {
calls.push(strings.join('?'));
return Promise.resolve(rows);
};
return { engine: e, calls };
}
/** Run `fn` with GBRAIN_CONFIG_CACHE_TTL_MS set (or cleared when undefined). */
const withTtl = (ttl: string | undefined, fn: () => Promise<void>) =>
withEnv({ GBRAIN_CONFIG_CACHE_TTL_MS: ttl }, fn);
describe('PostgresEngine config cache (#1694)', () => {
it('batch-loads once and serves repeat reads from the cache', () => withTtl(undefined, async () => {
const { engine, calls } = makeEngine([
{ key: 'search.mode', value: 'balanced' },
{ key: 'embedding_multimodal', value: 'true' },
]);
expect(await engine.getConfig('search.mode')).toBe('balanced');
expect(await engine.getConfig('embedding_multimodal')).toBe('true');
expect(await engine.getConfig('search.mode')).toBe('balanced');
// One SELECT total — this is the whole point of the fix.
expect(calls.length).toBe(1);
expect(calls[0]).toContain('SELECT key, value FROM config');
}));
it('returns null for a known-absent key without an extra round-trip', () => withTtl(undefined, async () => {
const { engine, calls } = makeEngine([{ key: 'a', value: '1' }]);
expect(await engine.getConfig('missing.key')).toBeNull();
expect(await engine.getConfig('missing.key')).toBeNull();
expect(calls.length).toBe(1);
}));
it('setConfig writes through so subsequent reads see the new value', () => withTtl(undefined, async () => {
const { engine, calls } = makeEngine([{ key: 'k', value: 'old' }]);
expect(await engine.getConfig('k')).toBe('old');
await engine.setConfig('k', 'new');
expect(await engine.getConfig('k')).toBe('new');
expect(calls.length).toBe(2); // batch load + upsert; no re-read
}));
it('unsetConfig writes through so subsequent reads see absence', () => withTtl(undefined, async () => {
const { engine } = makeEngine([{ key: 'k', value: 'v' }]);
expect(await engine.getConfig('k')).toBe('v');
await engine.unsetConfig('k');
expect(await engine.getConfig('k')).toBeNull();
}));
it('concurrent cold reads share a single batch load (single-flight)', () => withTtl(undefined, async () => {
const { engine, calls } = makeEngine([{ key: 'k', value: 'v' }]);
const [a, b, c] = await Promise.all([
engine.getConfig('k'),
engine.getConfig('k'),
engine.getConfig('other'),
]);
expect([a, b, c]).toEqual(['v', 'v', null]);
expect(calls.length).toBe(1);
}));
it('GBRAIN_CONFIG_CACHE_TTL_MS=0 disables the cache (per-key reads)', () => withTtl('0', async () => {
const { engine, calls } = makeEngine([{ value: 'v' }]);
expect(await engine.getConfig('k')).toBe('v');
expect(await engine.getConfig('k')).toBe('v');
expect(calls.length).toBe(2);
expect(calls[0]).toContain('SELECT value FROM config WHERE key =');
}));
it('an expired TTL reloads from the database', () => withTtl('1', async () => {
const { engine, calls } = makeEngine([{ key: 'k', value: 'v' }]);
expect(await engine.getConfig('k')).toBe('v');
await new Promise((r) => setTimeout(r, 5));
expect(await engine.getConfig('k')).toBe('v');
expect(calls.length).toBe(2); // two batch loads
}));
it('the batch load keeps the connRetry reconnect posture (#1603/#1891)', () => withTtl(undefined, async () => {
const e = new PostgresEngine();
(e as unknown as { _connectionStyle: string })._connectionStyle = 'instance';
(e as unknown as { _sql: unknown })._sql = null; // torn-down pool → retryable
(e as unknown as { _bulkRetryOptsCache: unknown })._bulkRetryOptsCache = FAST_RETRY;
let reconnects = 0;
(e as unknown as { reconnect: () => Promise<void> }).reconnect = async () => {
reconnects++;
(e as unknown as { _sql: unknown })._sql = () =>
Promise.resolve([{ key: 'k', value: 'v' }]);
};
expect(await e.getConfig('k')).toBe('v');
expect(reconnects).toBe(1);
}));
});
@@ -43,7 +43,9 @@ function makeTornDownEngine(poolResult: unknown): { engine: PostgresEngine; reco
describe('PostgresEngine non-batch config accessors self-heal (PR #1891 takeover)', () => {
it('getConfig reconnects + retries a null instance pool, then returns the value', async () => {
const { engine, reconnects } = makeTornDownEngine([{ value: 'live-value' }]);
// Rows carry `key` too: getConfig's default cached path batch-loads
// `SELECT key, value FROM config` (#1694) through the same connRetry.
const { engine, reconnects } = makeTornDownEngine([{ key: 'some.key', value: 'live-value' }]);
expect(await engine.getConfig('some.key')).toBe('live-value');
expect(reconnects()).toBe(1); // exactly one reconnect closed the gap
});