"TimeoutError: [Errno 110] Connection timed out" from Google Colab Notebook

[natemurthy]

Bug Description

I am trying to connect to a trial AlloyDB instance from a Google Colab Notebook. Both the my Notebook / GDrive are managed under the same GCP organization as the DB instance. Unfortunately I observe a timeout (already have my ADC with correct database user credentials, password hidden). The code I'm evaluating is available here: https://colab.research.google.com/drive/1_Xd4rfncFBfxbgQwYu_ZL4Tu76VApcn2#scrollTo=MA7sOm6qF2b0

from google.cloud.alloydb.connector import Connector

connector = Connector()

# https://github.com/GoogleCloudPlatform/alloydb-python-connector/tree/main
conn = connector.connect(
    "projects/dynamic-branch-380821/locations/us-central1/clusters/tmp-trial-alloydb-cluster/instances/primary-instance",
    "pg8000",
    user="analyst",
    password=db_password,
    db="verse-db"
)

The timeout occurs in the connecter.connect() method.

Example code (or command)

No response

Stacktrace

TimeoutError                              Traceback (most recent call last)
<ipython-input-12-ecf23c99c5ae> in <cell line: 6>()
      4 
      5 # https://github.com/GoogleCloudPlatform/alloydb-python-connector/tree/main
----> 6 conn = connector.connect(
      7     "projects/dynamic-branch-380821/locations/us-central1/clusters/tmp-trial-alloydb-cluster/instances/primary-instance",
      8     "pg8000",

7 frames
/usr/lib/python3.10/socket.py in create_connection(address, timeout, source_address)
    831             if source_address:
    832                 sock.bind(source_address)
--> 833             sock.connect(sa)
    834             # Break explicitly a reference cycle
    835             err = None

TimeoutError: [Errno 110] Connection timed out

Steps to reproduce?

To reproduce, just copy the notebook and execute within your GCP+Gdrive/Colab environment

Environment

This is already documented in the notebook

Additional Details

No response

[enocom]

By default the Python Connector tries to connect with private IP. Unless the VM that runs your notebook code is within the same VPC as your PSA attachment, you'll need to update your instance to have a public IP and update your connector code to be:

# https://github.com/GoogleCloudPlatform/alloydb-python-connector/tree/main
conn = connector.connect(
    "projects/dynamic-branch-380821/locations/us-central1/clusters/tmp-trial-alloydb-cluster/instances/primary-instance",
    "pg8000",
    user="analyst",
    password=db_password,
    db="verse-db",
    ip_type="PUBLIC",
)

The reason you're seeing a timeout error is there's no network path between the machine running your notebook and your private IP address.

[enocom]

You can assign a public IP with:

!gcloud beta alloydb instances update {instance_name} --region={region} --cluster={cluster_name} --assign-inbound-public-ip=ASSIGN_IPV4

[natemurthy]

Cool, that worked! Maybe that detail above can be added to the top-level README? I noticed for the cloud-sql-python-connector["pg8000"] library the Connector.connect() by default used the public IP. So that might be a small thing that should be consistent between the two class methods.

[enocom]

Glad to hear it and good idea. I've sent a PR to update the README to make this clear.