How to Run Varnish Cache on Fly

Varnish Cache is a popular open-source web application accelerator. When placed in front of a web page, Varnish will request the page from once to cache the page - making a copy of its resources (including images, styles, and scripts) in memory. Varnish then handles all future requests for the same page by serving the cached copy.

Varnish acts like a reverse proxy that sits between your end-users and your web server. It can be configured to filter requests based on rules written in Varnish Configuration Language.

You can configure Varnish to store your pages' content and serve the cached pages whenever it receives new requests. Thus, Varnish speeds up your web application and reduces the load on the webserver. According to their site: “Varnish typically speeds up delivery with a factor of 300 - 1000x, depending on your architecture.”

Fly.io runs your application in Docker containers across its edge network. This can make your application run 80% faster and increase reliability. By running Varnish on Fly, you can give your application an acceleration boost to make pages load even more quickly.

How to Configure Varnish to Run on Fly

In this tutorial, you’ll see how to configure and run Varnish on Fly. You’ll create a new Fly application, customize your Varnish .vcl file, create a Dockerfile, and check that responses are successfully cached as you would expect.

Prerequisites

The Flyctl command line tool.

Creating a New Fly Application

First, you'll need to use flyctl to create a new application. If you haven't already, install the appropriate version of flyctl for your operating system using the instructions at Installing flyctl.

After that, sign up or sign in to your Fly account from your console:

# Sign up 
flyctl auth signup

# Or sign in 
flyctl auth login

Running those commands opens a web page that allows you to log in by using your GitHub account or your email and password.

Create a new directory for the project and then use flyctl init to create a new application and generate the Fly configuration file.

mkdir varnish-on-fly && cd varnish-on-fly
flyctl init

For App Name and Select Organization prompts, press Enter to use an auto-generated name and your default organization.
For the Select builder prompt, press Enter to select the (Do not set a builder) option.
For the Select Internal Port prompt, enter 80 as the port number and press Enter.

Every Fly application makes use of the fly.toml file to manage deployments. After completing the steps above, you'll see the message Wrote config file fly.toml.

Deploying Varnish

To deploy Varnish to Fly, you can start from the official Docker image available on Docker Hub.

Varnish uses a domain-specific language called Varnish Configuration Language (VCL). You can configure Varnish by specifying the configuration in a default.vcl file. Varnish then uses the specified configuration for request handling and document caching policies for Varnish Cache.

Create a new file called default.vcl with the following content:

vcl 4.0;
backend default { 
    .host = "flygreeting.fly.dev";
    .port = "80";
}

The value of .host is a fully qualified hostname or IP address (typically a web server). This example uses the Flygreeting example app. The value of .port is the listening port of the Varnish backend, that is, the server providing the content that Varnish accelerates.

Next, create a new file called Dockerfile and add the following:

FROM varnish:6.4
COPY default.vcl /etc/varnish/
CMD ["/usr/sbin/varnishd", "-F", "-f", "/etc/varnish/default.vcl", "-T", "none"]

The Dockerfile uses the official Varnish Docker image with the configuration specified in the default.vcl file. By default, Varnish sets a management interface on your application, but because you shouldn't publicly expose this interface, you should update the CMD statement to specify the -T none flag. This disables the management interface. Read more about the varnishd command-line parameters here if you want to customize your run command further.

Now that your Dockerfile is set up, you're ready to deploy Varnish. Run flyctl deploy to deploy your application to Fly.io.

To check that Varnish is working, run the following command to get the response headers from your application:

curl -Is <your-app-address>/v1/countries/
HTTP/2 200
content-type: application/json; charset=utf-8
date: Sun, 30 Aug 2020 10:24:56 GMT
content-length: 1260
server: Fly/f8f635b (2020-08-24)
via: 1.1 fly.io, 2 fly.io
x-varnish: 32793 22
age: 53
accept-ranges: bytes

The presence of an x-varnish response header confirms that Varnish is intercepting web requests at this address.

Verifying That Caching Works

To check that Varnish is properly caching the Flygreeting web application, you can configure Varnish to update the response headers if the request hits or misses the cache. To do this, replace the default.vcl file with the following:

vcl 4.0;

backend default {
  .host = "flygreeting.fly.dev";
  .port = "80";
}

sub vcl_recv {
    unset req.http.x-cache;
}

sub vcl_hit {
    set req.http.x-cache = "hit";
}

sub vcl_miss {
    set req.http.x-cache = "miss";
}

sub vcl_pass {
    set req.http.x-cache = "pass";
}

sub vcl_pipe {
    set req.http.x-cache = "pipe uncacheable";
}

sub vcl_synth {
    set req.http.x-cache = "synth synth";
    set resp.http.x-cache = req.http.x-cache;
}

sub vcl_deliver {
    if (obj.uncacheable) {
        set req.http.x-cache = req.http.x-cache + " uncacheable" ;
    } else {
        set req.http.x-cache = req.http.x-cache + " cached" ;
    }
    set resp.http.x-cache = req.http.x-cache;
}

This configuration adds an additional x-cache response header that returns hit when the response is delivered from the cache or miss when the response comes from the backend.

Deploy the updated version on Fly by running flyctl deploy again. Send a request to your web application using curl to inspect its headers:

curl -Is <your-app-address>/v1/countries/ | grep 'x-cache'
x-cache: miss cached

The miss cached value denotes that the response is being delivered from the Flygreeting backend. Now send the same request again:

curl -Is <your-app-address>/v1/countries/ | grep 'x-cache'
x-cache: hit cached

The hit cached value denotes that the response was delivered from Varnish's cache. Your Varnish installation is now successfully deployed to Fly's global edge hosting network and is ready to improve your website's speed from anywhere in the world.

Conclusion

In this tutorial, you've seen how to deploy Varnish on Fly to take advantage of the application acceleration features that Varnish offers. You also learned how to configure Varnish to include caching status in a response header and how to verify that responses are being cached.

Varnish is a great way to improve the speed of your static web pages and assets, and deploying it to Fly is a great way to minimize the latency of your cache by distributing it globally. You can customize the way Varnish handles requests and caching using the Varnish Configuration Language, and for more information about setting up Varnish, see the Reference Manual.

If you have questions about using Varnish with Fly.io, reach out to support@fly.io, and we'll help you get started.

mciparelli / fly-varnish