Firestore: watch.unsubscribe() causes Google Cloud Function to timeout.

[Kharms]

Environment details

Environment: GCF python37 runtime
Version: google-cloud-firestore==1.4.0

Steps to reproduce

1. Initialize a fs document:
2. Deploy a HTTP triggered GCF that creates a listener, then unsubscribes.
3. View logs.

Code example

from google.cloud import firestore


# Initialization code, run prior to deploy.
# fs_client = firestore.Client()
# doc_ref = fs_client.collection("test").document("subtest")
# doc_ref.set({"hello": "world"})


def abort_now_listener(doc_snapshot, changes, read_time):
    for doc in doc_snapshot:
        doc_vals = doc.to_dict()
        print(doc_vals["hello"])


# gcloud functions deploy watch_fs --runtime python37 --trigger-http
def watch_fs(request):
    print("creating client")
    fs_client = firestore.Client()
    doc_ref = fs_client.collection("test").document("subtest")
    print(doc_ref)
    watcher = doc_ref.on_snapshot(abort_now_listener)
    print("after watcher")
    watcher.unsubscribe()
    print("the end!")
    return "hello world"

GCF Logs attached.

[Kharms]

Rolling back to 1.3.0 fixes the timeout issue, reintroduces: #7826

Might also be related to #8616

[tseaver]

@Kharms Thanks for the report, and for investigating rolling back. As a workaround, I would recommend just avoiding the unsubscribe call

[jlara310]

Issue not dependent of Cloud Function environment: https://stackoverflow.com/questions/58224392/detaching-firestore-listener-in-python-causes-program-to-freeze

[tseaver]

Via #9337.