Most web applications require or do load balance between servers/services in the same data center. However, if you are running mission-critical applications where uptime is needed all the time across the world, then you need a cloud load balancer. Not just uptime but could be many other factors. Ex:
Active-passive or active-active data center requirementDisaster recovery planTaking advantage of multiple data centers to serve requests from the nearest locationCompliance
Cloudflare offers local and global load balancing options, which help route traffic to multiple data centers. Some of the Cloudflare LB features are:
Health checks are inbuilt, so you can quickly take down a faulty server.Trigger failover when health check failsReduce latency by routing traffic to the closest serverDNS level and support HTTP(S), TCP, and UDPSession stickiness to ensure request goes to the same server
You can configure entire things either through the Cloudflare dashboard or API. Technically, the following instructions apply to load balance on any cloud platform like Azure, DigitalOcean, Alibaba, etc. But I’ve chosen GCP and AWS for the demonstration.
AWS & GCP Setup Details
I’ve provisioned one sever in GCP and AWS platform with the following.
Installed NginxAdded index.html with custom text to show the page is being served from the respective serverStarted Nginx, and the page is accessible from both servers
Let’s move to Cloudflare to implement LB.
Activating Cloudflare Load Balancer
I’ve one available domain (bloggerflare.com), which I will use for this lab. Note: Load balancer by Cloudflare is not FREE, and pricing starts at $5 per month. I assume you already have an account with Cloudflare; if not, you can create and add the domain, as I explained in my previous post.
Log in to Cloudflare and select the domain where you want to enable balancingGo to the traffic tab and enable load balancing.
Configure the features based on the requirement. I am proceeding with minimal configuration.
If you want requests to be redirected to the nearest location, then you got to enable Geo Routing.
Confirm the subscription and enable it.
As you can see, you can get it started from $5 a month with two origin servers and a one-minute health check interval. This indicates Cloudflare LB is activated and ready to be configured.
Creating Cloudflare LB
It will take a few seconds to confirm the subscription and take you back to the Traffic page.
Click Create Load Balancer
Enter the domain where you want to setup balancing.Expand Session Affinity and select By Cloudflare Cookie if you need to enable session stickiness
Enter a pool name, and it’s the origin (server where traffic should be redirected to)
Next, you can configure a health check.
A health check is essential. Cloudflare will stop sending traffic to the faulty origin when a health check fails.
If your origin is listening on port 80, then you can select HTTP or https for 443 port.
Cloudflare also lets you configure advanced health check settings such as:
GET or HEAD methodExpected HTTP status codeContent validation in the response bodyNumber of attempts before they consider healthy or unhealthyHeader name validation
And finally, save the configuration and deploy
LB will perform a health check, and in a few seconds, you will see the status is healthy.
Great, this concludes Cloudflare load balancer is ready to accept traffic from the world and forward it to the configured origin servers.
Testing Load Balancer
Let’s perform a basic test to see if it works.
First, try to access the domain.
Awesome! LB forwarded a request to Google Cloud VM and got the response. I can see the request in Nginx.
Let me shut down Nginx on GCP and access the page again.
And here you go. A requested page is served from AWS.
I can see LB took GCP pool member down. By default, Cloudflare IP will be shown in Nginx access logs, and if you need to restore client IP, then you can check out this guide.
Conclusion
Implementing a Cloudflare load balancer is straightforward and can get it started in less than 15 minutes. If you are looking for high availability between multiple data centers or origin servers, try it to see how it works.