Run Tensorboard on Google App Engine

Google App Engine can provide an easy way to get a persistent Tensorboard server with authentication for a small cost.

In addition if you find yourself pulling a large amount of data from GCS when starting up Tensorboard servers, you may actually pay less for a persistent GAE server, since you don't pay for data egress between GCS and GAE.

Setup

Setup assumes you have installed the Google Cloud SDK and have a billing enabled, Google Cloud Platform project.

1. Write a configuration file for the OAuth2 proxy that sits in front of your tensorboard server. See the oauth2_proxy docs for instructions. This may require creating an OAuth2 client ID for your chosen provider. oauth2_proxy supports authentication with Google, Github, Facebook, Azure, GitLab, and MyUSA.

2. Upload your configuration file to GCS:

   gsutil cp oauth2_proxy.cfg gs://$BUCKET/oauth2_proxy.cfg
   

3. Write an app.yaml file with runtime: custom to configure your application. Full reference here. In the env_variables section you must specify 3 environment variables:

• GCS_PROXY_CONFIG The fully qualified path to the configuration file you uploaded in step 2.
• EVENT_FILE_PATH The fully qualified GCS Path for your Tensorboard summary files.
• RELOAD_INTERVAL Frequency (in seconds) to poll GCS for new summary files.

Note that reloading data directly from GCS might cause a lot of API requests and incur high costs, so do not use the strategy of directly reading from GCS if you use it for more than a few hundreds of log files. This is due to inefficient way GCS direct access is working currently as described in tensorflow/tensorboard#158

For more frequent reloads/bigger number of log files, syncing the GCS data to local folder using gsutil rsync and loading the data from local folder seems to be a better strategy.

4. Deploy your GAE app.

   gcloud app deploy app.yaml --image-url=gcr.io/google-samples/appengine-tensorboard:latest