---
layout: default
navsection: installguide
title: Install Keep-web server
...

The Keep-web server provides read-only HTTP access to files stored in Keep. It serves public data to unauthenticated clients, and serves private data to clients that supply Arvados API tokens. It can be installed anywhere with access to Keep services, typically behind a web proxy that provides SSL support. See the "godoc page":http://godoc.org/github.com/curoverse/arvados/services/keep-web for more detail.

By convention, we use the following hostnames for the Keep-web service:

<notextile>
<pre><code>download.<span class="userinput">uuid_prefix</span>.your.domain
collections.<span class="userinput">uuid_prefix</span>.your.domain
*.collections.<span class="userinput">uuid_prefix</span>.your.domain
</code></pre>
</notextile>

The above hostnames should resolve from anywhere on the internet.

h2. Install Keep-web

Typically Keep-web runs on the same host as Keepproxy.

On Debian-based systems:

<notextile>
<pre><code>~$ <span class="userinput">sudo apt-get install keep-web</span>
</code></pre>
</notextile>

On Red Hat-based systems:

<notextile>
<pre><code>~$ <span class="userinput">sudo yum install keep-web</span>
</code></pre>
</notextile>

Verify that @Keep-web@ is functional:

<notextile>
<pre><code>~$ <span class="userinput">keep-web -h</span>
Usage of keep-web:
  -allow-anonymous
        Serve public data to anonymous clients. Try the token supplied in the ARVADOS_API_TOKEN environment variable when none of the tokens provided in an HTTP request succeed in reading the desired collection. (default false)
  -attachment-only-host string
        Accept credentials, and add "Content-Disposition: attachment" response headers, for requests at this hostname:port. Prohibiting inline display makes it possible to serve untrusted and non-public content from a single origin, i.e., without wildcard DNS or SSL.
  -listen string
        Address to listen on: "host:port", or ":port" to listen on all interfaces. (default ":80")
  -trust-all-content
        Serve non-public content from a single origin. Dangerous: read docs before using!
</code></pre>
</notextile>

{% assign railscmd = "bundle exec ./script/get_anonymous_user_token.rb" %}
{% assign railsout = "zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz" %}
If you intend to use Keep-web to serve public data to anonymous clients, configure it with an anonymous token. You can use the same one you used when you set up your Keepproxy server, or use the following command on the <strong>API server</strong> to create another. {% include 'install_rails_command' %}

We recommend running Keep-web under "runit":https://packages.debian.org/search?keywords=runit or a similar supervisor. The basic command to start Keep-web is:

<notextile>
<pre><code>export ARVADOS_API_HOST=<span class="userinput">uuid_prefix</span>.your.domain
export ARVADOS_API_TOKEN="<span class="userinput">{{railsout}}</span>"
exec sudo -u nobody keep-web \
 -listen=<span class="userinput">:9002</span> \
 -attachment-only-host=<span class="userinput">download.uuid_prefix.your.domain</span> \
 -allow-anonymous \
 2&gt;&amp;1
</code></pre>
</notextile>

Omit the @-allow-anonymous@ argument if you do not want to serve public data.

Set @ARVADOS_API_HOST_INSECURE=1@ if your API server's SSL certificate is not signed by a recognized CA.

h3. Set up a reverse proxy with SSL support

The Keep-web service will be accessible from anywhere on the internet, so we recommend using SSL for transport encryption.

This is best achieved by putting a reverse proxy with SSL support in front of Keep-web, running on port 443 and passing requests to Keep-web on port 9002 (or whatever port you chose in your run script).

Note: A wildcard SSL certificate is required in order to support a full-featured secure Keep-web service. Without it, Keep-web can offer file downloads for all Keep data; however, in order to avoid cross-site scripting vulnerabilities, Keep-web refuses to serve private data as web content except when it is accessed using a "secret link" share. With a wildcard SSL certificate and DNS configured appropriately, all data can be served as web content.

For example, using Nginx:

<notextile><pre>
upstream keep-web {
  server                127.0.0.1:<span class="userinput">9002</span>;
}

server {
  listen                <span class="userinput">[your public IP address]</span>:443 ssl;
  server_name           download.<span class="userinput">uuid_prefix</span>.your.domain
                        collections.<span class="userinput">uuid_prefix</span>.your.domain
                        *.collections.<span class="userinput">uuid_prefix</span>.your.domain
                        ~.*--collections.<span class="userinput">uuid_prefix</span>.your.domain;

  proxy_connect_timeout 90s;
  proxy_read_timeout    300s;

  ssl                   on;
  ssl_certificate       <span class="userinput"/>YOUR/PATH/TO/cert.pem</span>;
  ssl_certificate_key   <span class="userinput"/>YOUR/PATH/TO/cert.key</span>;

  location / {
    proxy_pass          http://keep-web;
    proxy_set_header    Host            $host;
    proxy_set_header    X-Forwarded-For $proxy_add_x_forwarded_for;
  }
}
</pre></notextile>

h3. Configure DNS

Configure your DNS servers so the following names resolve to your Nginx proxy's public IP address.
* @download.uuid_prefix.your.domain@
* @collections.uuid_prefix.your.domain@
* @*--collections.uuid_prefix.your.domain@, if you have a wildcard SSL certificate valid for @*.uuid_prefix.your.domain@ and your DNS server allows this without interfering with other DNS names.
* @*.collections.uuid_prefix.your.domain@, if you have a wildcard SSL certificate valid for these names.

If neither of the above wildcard options is feasible, only unauthenticated requests (public data and collection sharing links) will be served as web content at @collections.uuid_prefix.your.domain@. The @download@ name will be used to serve authenticated content, but only as file downloads.

h3. Tell Workbench about the Keep-web service

Workbench has features like "download file from collection" and "show image" which work better if the content is served by Keep-web rather than Workbench itself. We recommend using the two different hostnames ("download" and "collections" above) for file downloads and inline content respectively.

Add the following entry to your Workbench configuration file (@/etc/arvados/workbench/application.yml@). This URL will be used for file downloads.

<notextile>
<pre><code>keep_web_download_url: https://download.<span class="userinput">uuid_prefix</span>.your.domain/c=%{uuid_or_pdh}
</code></pre>
</notextile>

Additionally, add *one* of the following entries to your Workbench configuration file, depending on your DNS setup. This URL will be used to serve user content that can be displayed in the browser, like image previews and static HTML pages.

<notextile>
<pre><code>keep_web_url: https://%{uuid_or_pdh}--collections.<span class="userinput">uuid_prefix</span>.your.domain
keep_web_url: https://%{uuid_or_pdh}.collections.<span class="userinput">uuid_prefix</span>.your.domain
keep_web_url: https://collections.<span class="userinput">uuid_prefix</span>.your.domain/c=%{uuid_or_pdh}
</code></pre>
</notextile>