fix: handle cloud-sql-connector unhandled exceptions on close

[fredzqm]

Description

This commit resolves an issue where deploying a Data Connect service backed by Cloud SQL Postgres fails with an unexpected unhandled exception (CloudSQLConnectorError: The connector was closed).

In @google-cloud/cloud-sql-connector version >= 1.7.0, the connector.close() method destroys underlying sockets by asynchronously emitting an error event. Previously, firebase-tools closed the pg.Pool first, which caused the database client to detach all error listeners from its sockets. Consequently, the later connector.close() call caused Node.js to crash due to an unhandled exception on those sockets.

To prevent this:

1. Re-ordered cleanup sequences so that connector.close() is executed prior to pool.end().
2. Attached an explicit 'error' handler to pg.Pool to ensure any connection teardown errors on idle clients are safely logged and absorbed.
3. Wrapped cleanup functions inside try/catch blocks for added defensiveness.

Fixes #10549

Scenarios Tested

• Successfully ran all dataconnect and cloudsql unit tests.
• Verified teardown sequence completes gracefully without crash using a local reproduction script.
• Verified fix handles both connect.ts deployments and dataconnect-sql-shell.ts shell exits.

Sample Commands

firebase deploy --only dataconnect --project <project-id>

Description

Scenarios Tested

Sample Commands

[gemini-code-assist]

Code Review

This pull request adds error handling to the PostgreSQL pool and wraps cleanup operations in try/catch blocks to prevent unhandled exceptions during Data Connect deployments. However, the reviewer identified critical issues where resource leaks can occur if pool.connect() fails, as the cleanup functions would not be executed. Additionally, declaring conn after cleanUpFn in connect.ts can trigger linter warnings due to referencing a variable before its declaration.

[gemini-code-assist]

If pool.connect() fails and throws an error, the command action will reject immediately without executing the cleanup block at the end of the function. This results in a resource leak where the connector and pool are left open.

Wrapping pool.connect() in a try/catch block to clean up resources on failure prevents this leak.

Suggested change

	const conn: pg.PoolClient = await pool.connect();
	let conn: pg.PoolClient;
	try {
	conn = await pool.connect();
	} catch (err) {
	try {
	connector.close();
	} catch (closeErr) {
	logger.debug("Error closing Cloud SQL connector:", closeErr);
	}
	try {
	await pool.end();
	} catch (endErr) {
	logger.debug("Error ending pg pool:", endErr);
	}
	throw err;
	}

[joehan]

Even better than pinning - thanks!