CloudVolumeServer

Flask application to serve data via CloudVolume.

This is currently a one-trick pony! The one thing it does is provide an API endpoint to query multiple x/y/z locations at once.

Python dependencies

• CloudVolume: pip install cloud-volume
• Gunicorn: pip install gunicorn
• flask: pip install flask
• numpy: pip install numpy
• pandas: pip install pandas

Deploying

1. Clone this repository on your server

2. Install above dependencies

3. Edit config.py and minimally add the path to the volume:

   CloudVolumeKwargs = {
                        cloudpath="file:///fafbz/fafb14/segmentation/"
                       }

4. Test if Gunicorn can start the workers - from within the repository run:

   gunicorn --workers 1 --bind 0.0.0.0:6000 -m 007 wsgi:app --log-level debug

   If all works well, CTRL-C to stop.

5. Add Gunicorn to systemd:

   sudo nano /etc/systemd/system/cloudvolumeserver.service

   Then copy paste this (make sure to replace YOURUSER and PATH/TO/REPOSITORY)

   [Unit]
   Description=Gunicorn instance to serve the CloudVolumeServer
   After=network.target
   
   [Service]
   User=YOURUSER
   Group=www-data
   WorkingDirectory=PATH/TO/REPOSITORY/CloudVolumeServer
   ExecStart=/usr/local/bin/gunicorn --workers 3 --bind 0.0.0.0:6000 -m 007 wsgi:app
   
   [Install]
   WantedBy=multi-user.target
   

   Note that you can also bind Gunicorn to a sock instead of a port but that didn't work in my hands.

   You can now start the Gunicorn service by running:

   sudo systemctl start cloudvolumeserver
   sudo systemctl enable cloudvolumeserver
   

6. Setup your nginx server and add the following server block to the config file in /etc/nginx/sites-available:

   location /seg/ {
                  include proxy_params;                  
                  proxy_pass http://localhost:6000/;
                  }
   

7. Restart your nginx: sudo systemctl restart nginx

8. All done. Check if the site is running by visiting http://your-server.url/seg/.

For more details: the configuration steps roughly follow this tutorial.

Examples

import requests

# Some random locations
# Please note that we need to query pixel coordinates, not native coordinates
coords = [[58695, 28631,   696],
          [58688, 28632,   691],
          [58691, 28635,   698]]

url = 'http://your-server.url/seg/'
resp = requests.post(url, json=coords)

print('Segment IDs:', resp.json())

License

This code is published under a BSD 2-Clause License.