Using arv-web

arv-web enables you to run a custom web service from the contents of an Arvados collection.

Note:

This tutorial assumes that you have installed the Arvados Command line SDK and Python SDK on your workstation and have a working environment.

Usage

arv-web enables you to set up a web service based on the most recent collection in a project. An arv-web application is a reproducible, immutable application bundle where the web app is packaged with both the code to run and the data to serve. Because Arvados Collections can be updated with minimum duplication, it is efficient to produce a new application bundle when the code or data needs to be updated; retaining old application bundles makes it easy to go back and run older versions of your web app.

$ cd $HOME/arvados/services/arv-web
usage: arv-web.py [-h] --project-uuid PROJECT_UUID [--port PORT]
                  [--image IMAGE]

optional arguments:
  -h, --help            show this help message and exit
  --project-uuid PROJECT_UUID
                        Project uuid to watch
  --port PORT           Host port to listen on (default 8080)
  --image IMAGE         Docker image to run

At startup, arv-web queries an Arvados project and mounts the most recently modified collection into a temporary directory. It then runs a Docker image with the collection bound to /mnt inside the container. When a new collection is added to the project, or an existing project is updated, it will stop the running Docker container, unmount the old collection, mount the new most recently modified collection, and restart the Docker container with the new mount.

Docker container

The Dockerfile in arvados/docker/arv-web builds a Docker image that runs Apache with /mnt as the DocumentRoot. It is configured to run web applications which use Python WSGI, Ruby Rack, or CGI; to serve static HTML; or browse the contents of the public subdirectory of the collection using default Apache index pages.

To build the Docker image:

~$ cd arvados/docker
~/arvados/docker$ docker build -t arvados/arv-web arv-web

Running sample applications

First, in Arvados Workbench, create a new project. Copy the project UUID from the URL bar (this is the part of the URL after projects/...).

Now upload a collection containing a Python WSGI web app:

~$ cd arvados/services/arv-web
~/arvados/services/arv-web$ arv-put --project [zzzzz-j7d0g-yourprojectuuid] --name sample-wsgi-app sample-wsgi-app
0M / 0M 100.0%
Collection saved as 'sample-wsgi-app'
zzzzz-4zz18-ebohzfbzh82qmqy
~/arvados/services/arv-web$ ./arv-web.py --project [zzzzz-j7d0g-yourprojectuuid] --port 8888
2015-01-30 11:21:00 arvados.arv-web[4897] INFO: Mounting zzzzz-4zz18-ebohzfbzh82qmqy
2015-01-30 11:21:01 arvados.arv-web[4897] INFO: Starting Docker container arvados/arv-web
2015-01-30 11:21:02 arvados.arv-web[4897] INFO: Container id e79e70558d585a3e038e4bfbc97e5c511f21b6101443b29a8017bdf3d84689a3
2015-01-30 11:21:03 arvados.arv-web[4897] INFO: Waiting for events

The sample application will be available at http://localhost:8888.

Updating the application

If you upload a new collection to the same project, arv-web will restart the web service and serve the new collection. For example, uploading a collection containing a Ruby Rack web app:

~$ cd arvados/services/arv-web
~/arvados/services/arv-web$ arv-put --project [zzzzz-j7d0g-yourprojectuuid] --name sample-rack-app sample-rack-app
0M / 0M 100.0%
Collection saved as 'sample-rack-app'
zzzzz-4zz18-dhhm0ay8k8cqkvg

arv-web will automatically notice the change, load a new container, and send an update signal (SIGHUP) to the service:

2015-01-30 11:21:03 arvados.arv-web[4897] INFO:Waiting for events
2015-01-30 11:21:04 arvados.arv-web[4897] INFO:create zzzzz-4zz18-dhhm0ay8k8cqkvg
2015-01-30 11:21:05 arvados.arv-web[4897] INFO:Mounting zzzzz-4zz18-dhhm0ay8k8cqkvg
2015-01-30 11:21:06 arvados.arv-web[4897] INFO:Sending refresh signal to container
2015-01-30 11:21:07 arvados.arv-web[4897] INFO:Waiting for events

Writing your own applications

The arvados/arv-web image serves Python and Ruby applications using Phusion Passenger and Apache mod_passenger. See Phusion Passenger users guide for Apache for details, and look at the sample apps arvados/services/arv-web/sample-wsgi-app and arvados/services/arv-web/sample-rack-app.

You can serve CGI applications using standard Apache CGI support. See Apache Tutorial: Dynamic Content with CGI for details, and look at the sample app arvados/services/arv-web/sample-cgi-app.

You can also serve static content from the public directory of the collection. Look at arvados/services/arv-web/sample-static-page for an example. If no index.html is found in public/, it will render default Apache index pages, permitting simple browsing of the collection contents.

Custom images

You can provide your own Docker image. The Docker image that will be used create the web application container is specified in the docker_image file in the root of the collection. You can also specify --image on the command arv-web line to choose the docker image (this will override the contents of docker_image).

Reloading the web service

Stopping the Docker container and starting it again can result in a small amount of downtime. When the collection containing a new or updated web application uses the same Docker image as the currently running web application, it is possible to avoid this downtime by keeping the existing container and only reloading the web server. This is accomplished by providing a file called reload in the root of the collection, which should contain the commands necessary to reload the web server inside the container.


Previous: Concurrent Crunch tasks Next: run-command reference

The content of this documentation is licensed under the Creative Commons Attribution-Share Alike 3.0 United States licence.
Code samples in this documentation are licensed under the Apache License, Version 2.0.