Closing the PubSub client does not cleanup grpc connection

Environment details

• OS: MacOS 11.1 (20C69)
• Node.js version: 14.15.3
• npm version: 6.14.9
• @google-cloud/pubsub version: 2.8.0

Steps to reproduce

I noticed the issue because jest complains about:

Jest did not exit one second after the test run has completed. This usually means that there are asynchronous operations that weren’t stopped in your tests. Consider running Jest with --detectOpenHandles to troubleshoot this issue.

But the issue can also be observed without a test runner:

import log from 'why-is-node-running'
import pubsub from '@google-cloud/pubsub';

(async () => {
  const client1 = new pubsub.PubSub({
    projectId: 'stub',
  });
  const [topic1] = await client1.createTopic('test');
  await topic1.delete();
  await client1.close();
  log();
})();

Output:

There are 10 handle(s) keeping the process running

# DNSCHANNEL
(unknown stack trace)

# TTYWRAP
/Users/michael/www/node_modules/supports-color/index.js:129 - stdout: getSupportLevel(process.stdout),

# SIGNALWRAP
/Users/michael/www/node_modules/supports-color/index.js:129 - stdout: getSupportLevel(process.stdout),

# TTYWRAP
/Users/michael/www/node_modules/supports-color/index.js:130 - stderr: getSupportLevel(process.stderr)

# Timeout
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/backoff-timeout.js:63 - this.timerId = setTimeout(() => {
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:189     - this.backoffTimeout.runOnce();
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:397     - this.startBackoff();
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:554     - if (!this.transitionToState([channel_1.ConnectivityState.IDLE], channel_1.ConnectivityState.CONNECTING)) {

# TCPWRAP
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:261 - return net.connect(this.subchannelAddress);
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:283 - const session = http2.connect(addressScheme + targetAuthority, connectionOptions);
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:361 - this.createSession(result);

# HTTP2SESSION
(unknown stack trace)

# Immediate
/Users/michael/www/node_modules/google-gax/build/src/call.js:70                    - setImmediate(this.callback, ...args);
/Users/michael/www/node_modules/google-gax/build/src/normalCalls/retries.js:82     - callback(null, response, next, rawResponse);
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/client.js:173              - callProperties.callback(null, responseMessage);
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/client-interceptors.js:334 - (_b = listener === null || listener === void 0 ? void 0 : listener.onReceiveStatus) === null || _b === void 0 ? void 0 : _b.call(listener, status);

# TickObject
(unknown stack trace)

# TickObject
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:411     - process.nextTick(() => {
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:441     - this.transitionToState([
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:496     - this.checkBothRefcounts();
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel.js:500     - this.unref();
/Users/michael/www/node_modules/@grpc/grpc-js/build/src/subchannel-pool.js:56 - const refedSubchannels = subchannelObjArray.filter((value) => !value.subchannel.unrefIfOneRef());

while this works fine:

import log from 'why-is-node-running'
import pubsub from '@google-cloud/pubsub';

(async () => {
  const client1 = new pubsub.PubSub({
    projectId: 'stub',
  });
  /// nothing is done
  await client1.close();
  log();
})();

[murgatroid99]

I have published a new version of grpc-js that might help with this. Can you update your dependencies and try again?

[feywind]

It sounds like we can close this, then? This patch looks like it went into a recent @grpc/grpc-js release, so anyone having the problem should be able to pull in the newest version of that.

Please feel free to ping if it needs reopening.