fix(fcm): A workaround for the concurrency issues in googleapiclient by hiranya911 · Pull Request #558

fix(fcm): A workaround for the concurrency issues in googleapiclient by hiranya911 · Pull Request #558 · firebase/firebase-admin-python

Conversation

This is a possible solution to the problem outlined in #507. I wasn't quite able to repro the exception in the bug report, but the threading issues in googleapiclient are well known, and I can easily reproduce some concurrency issues by calling send_all() from just 3-4 parallel threads. The proposed solution is to instantiate a new transport instance for each invocation of send_all() (as opposed to using a shared transport across calls), and tests indeed confirm that it resolves the issues I managed to reproduce.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Using a new transport makes sense to me. Thanks!


		batch = http.BatchHttpRequest(
		callback=batch_callback, batch_uri=_MessagingService.FCM_BATCH_URL)
		transport = self._build_transport(self._credential)

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is the reason to use _build_trasnport here (instead of directly using _auth.authorized_http) so we can write tests easily?

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes.

Conversation

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

2 participants