test(api): let a shutdown's closed connections leave pg_stat_activity
CI / verify (pull_request) Successful in 1m16s
CI / infrastructure-smoke (pull_request) Successful in 40s
CI / package-release (pull_request) Successful in 57s
CI / image-build (pull_request) Successful in 1m1s
CI / migration-smoke (pull_request) Successful in 2m30s
CI / deploy-smoke (same-origin) (pull_request) Successful in 50s
CI / package-publication (pull_request) Successful in 55s
CI / deploy-smoke (split) (pull_request) Successful in 2m41s

Four integration tests shut the API down and immediately count the
database's rows in pg_stat_activity, expecting zero. PostgreSQL ends a
backend asynchronously: pool.end() resolves once the client side has
closed, and the server can still list that backend for a moment. On a
loaded runner an immediate count saw one, failing migration-smoke on a
docs-only PR (Ekklesia CI run #238, module-database.integration.ts:764,
"expected 1 to be +0").

Reproduced against postgres:18.4-alpine3.24 limited to 0.05 CPU: after
pool.end() resolved, the backend was still listed in 7 of 300 trials,
and gone about a millisecond later.

The checks now wait up to 3 s for the count to reach zero. A connection
that is genuinely left open is still caught: an idle pooled client
stays listed for pg-pool's 10 s idle timeout, far past the deadline.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
This commit is contained in:
ekklesia
2026-10-13 23:13:21 -05:00
co-authored by Claude Opus 5.5
parent 63fda0f8d6
commit f2fc069e66
2 changed files with 50 additions and 4 deletions
+24 -1
View File
@@ -195,6 +195,29 @@ async function activeConnectionCount(databaseName: string): Promise<number> {
}
}
/** How long a shutdown's closed connections may take to leave `pg_stat_activity`. Far below pg's
* 10 s idle timeout, so a connection that is genuinely still open is still caught. */
const SHUTDOWN_SETTLE_MS = 3_000;
/**
* The connections left on `databaseName` once a shutdown has had a moment to land.
*
* PostgreSQL ends a backend asynchronously: `pool.end()` resolves once the client side has closed,
* and the server can still list that backend for a moment afterwards. On a loaded CI runner that
* moment was long enough for an immediate count to see one. So this waits, briefly, for the count to
* reach zero; a leaked pooled client would stay idle and listed far past the deadline.
*/
async function connectionsLeftAfterShutdown(databaseName: string): Promise<number> {
const deadline = Date.now() + SHUTDOWN_SETTLE_MS;
for (;;) {
const count = await activeConnectionCount(databaseName);
if (count === 0 || Date.now() >= deadline) {
return count;
}
await new Promise((resolve) => setTimeout(resolve, 50));
}
}
async function createDisposableDatabase(label: string): Promise<{ name: string; url: string }> {
const name = makeDatabaseName(label);
await createDatabase(name);
@@ -760,7 +783,7 @@ describe.sequential("live API — module database access through ModuleRuntimeSe
await running.shutdown();
const remainingConnections = await activeConnectionCount(database.name);
const remainingConnections = await connectionsLeftAfterShutdown(database.name);
expect(remainingConnections).toBe(0);
const eventsAfterShutdown = readRecorder();