How to correctly stop Google Cloud Datalab

5
votes

Playing with data is a joy in Junyper/Datalab, but I do not want it to become costly.

Google recommends:

"You can minimize compute charges by stopping/restarting Cloud Datalab instances."

However, if I stop the AppEngine instance or the Compute Engine VM instance, they simply restart ...

So how to correctly stop/pause Google Cloud Datalab, so that I'm only charged for my use, not for the idle time? Is there some kind of trigger that restarts the instances?

1
google-app-enginegoogle-compute-enginegoogle-cloud-datalab
Google Cloud Datalab is currently in beta, so the problem discribed above is likely to disappear, or the solution might change in the future. – Karel Van Roey
I have this problem too...If you delete the lab, then, I think you'll lose all your code. I think Google's documentation on the shutdown is incorrect. github.com/GoogleCloudPlatform/datalab/issues/742 – Mike Chirico
@MikeChirico Thx for sharing the issue link. Deleting the datalab VM instance does not necessarily result in data loss [!]: The notebooks are in a git-repo (can be synced with Github/etc), and your raw data should only temporarily be stored inside the VM. Better options are Google Cloud Storage, BigQuery, ... But having to delete the VM instances is cumbersome and interrupts the flow of normal usage. [!] Try this technique first with non-vital data! – Karel Van Roey

1 Answers

8
votes

Here's what I'm doing. I like keeping everything in the cloud; but, I want to resume work with minimal fuss and cost.

Save Data/Source and Delete Datalab

1. Commit changes to Your Cloud Git Repository. As long as the source changes are committed, they'll pop backup when you re-create google cloud. Keep your data in buckets gs://... or BigQuery. Commit all changed Notebooks.

Your Cloud Git Repository

2. Delete the datalab module.

https://console.cloud.google.com/projectselector/appengine/versions?moduleId=datalab enter image description here

Resuming Your Work:

When you're ready to resume your work, goto launch cloud datalab sign-in, and deploy. It will take 6 or 7 minutes. Once it's deployed, up, and running, you can go directly to your tree and resume your work. You're probably talking 10 minutes tops.