Engineering
What’s Wrong with Feature Flags?
Engineering and Growth teams aren't speaking the same language
Learn more
What are Rolling Deployments & How do They Work?
Discover how rolling deployments streamline software updates, reduce risk, and keep your app online. Learn best practices and how to simplify the process.
Rolling out new features in SaaS can be tricky. Often, it involves server downtime, risks of compatibility issues, and a large resource investment across your entire development team.
However, there’s a way to avoid all of these obstacles. The key is implementing rolling deployments.
In this primer, we’ll examine this software update strategy and show you why and how it works.
We’ll go over:
A quick definition of rolling deployment as a strategy
An example scenario of rolling deployment in action
A step-by-step guide showing you how rolling deployments work
A list of benefits of using rolling deployments
Some considerations you should have during implementation
Tips and best practices to keep in mind
Let’s get started.
What is a rolling deployment?
Rolling deployments are a software update strategy designed to minimize downtime and risk when releasing new versions of an application.
Rolling deployments gradually replace individual components instead of replacing the entire system at once. A small subset of servers or instances is updated, tested, and then another subset is updated, with this process repeating until the entire system is running the new version.
Why use rolling deployments?
Zero downtime: Users can continue to access the application without interruptions.
Safer updates: Problems with the new version can be quickly identified, and the rollout can be paused or reversed if necessary.
Smoother on resources: Rolling deployments avoid the need to duplicate the entire production environment.
Keep in mind that rolling deployments rely on backward compatibility, meaning the new code must function correctly with older components that haven't been updated yet.
A rolling deployment example scenario
Imagine a popular online photo-sharing platform running in the cloud. This platform relies on multiple services: a web frontend for users, an image processing service, and a database to store all the data.
The development team has a major update ready — new image filters, faster load times, and more. Let's see how a rolling deployment strategy makes this happen with minimal disruption:
One by one: The team starts by spinning up a single instance of the image processing service with the new code. This instance runs alongside the existing ones.
Health checks: Before directing any real user traffic to the updated instance, they run careful tests to make sure everything is working as intended.
Switch and repeat: Once the new instance is verified, they take one of the old image-processing instances offline. Now, some users will start getting their images processed by the new code. This step will keep repeating — update, check, switch — until all image processing instances are running the new version.
Database updates: If the new version involves database changes, these need to be made in a way that doesn't break the old code still running on other instances.
Frontend update: Lastly, the frontend gets updated using the same rolling process to make sure it works seamlessly with the updated backend.
Here are some key points to highlight from this example:
Throughout the process, the platform stayed online. Only small portions were updated at a time.
The team could quickly pause or reverse the rollout if any problems showed up, minimizing risks.
Old and new versions of the code needed to work together temporarily. This ensured a smooth transition for users.
How rolling deployments work
Rolling deployments are all about a careful, step-by-step approach to minimize downtime and risk when updating applications. Let's break down the key stages:
Step 1: Divide the environment into manageable groups or batches
Planning is indispensable. Think about your system's structure and choose a batch size that makes sense.
You might group things based on individual servers, containers, or even logical parts of your application. Smaller batches offer more fine-grained control and safety during the update process.
Step 2: Deploy the update to the first batch
This is where the actual deployment happens. New instances running the updated code are brought online alongside the existing ones. Your load balancing needs to be carefully adjusted to start directing a portion of user traffic toward the updated batch.
Step 3: Monitor the deployment for stability and errors
Now, it's time to be extra attentive. Use monitoring tools, check logs, and consider user feedback to assess the updated batch's health.
Look for any performance hiccups, errors, or anything that doesn't seem right. This could point to issues in the new code or potential problems interacting with the rest of your system.
Step 4: Move to the next batch after ensuring the prior update worked
If everything looks good, you can move on with confidence. But, if you see any problems, it's time to pause the rollout.
The great thing about rolling deployments is that you can investigate and fix issues before they snowball. You might even decide to roll back that updated batch to the previous version if needed.
Step 5: Move to the next batch after ensuring the prior update worked
Keep moving forward, repeating this pattern of updating, monitoring, and verifying. Each successful batch builds trust in your new version. Eventually, the entire system will be running smoothly on the updated code, and your users will hardly have noticed.
Why use a rolling deployment strategy?
Rolling deployments offer significant advantages for updating applications in a way that's both safe and user-friendly. Let's dive into the key strengths:
Minimizes downtime: Perhaps the biggest advantage of rolling deployments is keeping your application online for users during the entire update process.
Instead of everything going offline for maintenance, updates happen piece by piece. Your users might not even notice that things are changing behind the scenes.
Reduces risk: With rolling deployments, problems introduced by the new version only affect a small portion of users. This means if something goes wrong, the impact is contained.
You can quickly catch issues, pause the rollout, diagnose the problem, and even revert to the older version if needed. This is very different from a "big bang" deployment, where a single glitch could bring down the whole system.
Incremental testing: Rolling deployments allow for testing in a real production environment. As you update each batch, monitor it closely for performance, errors, or unexpected behavior.
This real-world feedback helps you catch potential issues early before they affect your entire user base. Your team can make adjustments on the fly or pause the rollout if needed.
Challenges you could face when using rolling deployments
While rolling deployments offer many benefits, it's important to be mindful of the potential complexities.
One potential drawback of rolling deployments is that the update can take longer to complete than methods where you replace the entire system at once. This is because you're updating things batch by batch.
The size of your batches plays a big role in determining the overall speed of the deployment.
Another consideration is the need for excellent monitoring capabilities. With old and new code versions running simultaneously, you need a way to keep a close eye on the health of individual batches and the overall application performance.
This might mean investing in new monitoring tools or expanding your existing setup to track the metrics that matter most during the rollout.
Perhaps the most crucial challenge is ensuring compatibility between different versions of your software components. For a smooth rollout, your new code must be able to work correctly with older parts of your system that haven't been updated yet.
This is known as backward compatibility and demands careful planning and code design.
It's important to weigh these considerations against the benefits that rolling deployments bring. For many applications, the increased safety, control, and minimal downtime will outweigh these challenges.
Tips and best practices when implementing rolling deployments
Rolling deployments can be a powerful way to update your application, but they work best when approached with careful planning and a focus on monitoring and adaptability. Let's explore some key practices:
Thorough planning: A well-thought-out plan is critical. Determine batch sizes, rollout order, and avoid scheduling updates during peak traffic hours. A staggered approach with sufficient time for testing and monitoring each batch gives you the best chance of catching issues and minimizing user impact.
Monitoring and logging: Visibility is key. Set up detailed monitoring systems for both individual batches and the health of your overall application. This should include performance metrics, error tracking, and log analysis tools. Consider setting up alerts to notify you quickly if there's unexpected behavior in the updated instances.
Embrace automation: Automation tools can be a lifesaver with rolling deployments. Look for tools that can automatically handle instance creation, traffic switching between versions, and even basic health checks. This reduces the risk of human error and lets you focus on the analysis of your updated batches.
Feedback loops matter: Think of each completed batch as a learning opportunity! Gather feedback through monitoring, user reports, and even direct testing. Use this information to improve the rollout process for the next batch or make adjustments if issues are spotted. Each successful deployment should refine your overall process.
Next steps
You should now understand the concept of rolling deployments — updating your application incrementally to ensure stability and reduce risk.
Now, the challenge is in the execution. Implementing rolling deployments effectively requires robust tools for granular control, continuous monitoring, and the ability to make data-driven adjustments quickly.
This is where Eppo comes into the mix.
Eppo is an all-encompassing experimentation and feature management platform that streamlines the process of rolling out new features and updates.
Eppo empowers you to effortlessly manage rolling deployments with feature flagging, robust analysis tools, and a focus on trust and control.
Designed for data-driven teams seeking to minimize risk and maximize efficiency, Eppo enables you to test, analyze, and safely roll out updates across your application.
Here's how Eppo helps with your rolling deployment implementation:
Control and flexibility: Eppo's feature flagging allows precise control over traffic allocation for each batch. You can gradually expose updates to users to monitor for issues.
Real-time insights: Eppo's warehouse-native architecture leverages your existing data infrastructure. Monitor performance metrics as each batch rolls out, and Eppo's diagnostics alert you immediately to any potential problems.
Trusted data: Eppo prioritizes rigor and auditability. Advanced statistical methods ensure reliable results, giving you confidence in every deployment phase.
Minimized risk: If issues arise, Eppo's feature flags and kill switches allow you to quickly pause or roll back an update, minimizing any negative impact on your customers’ user experience.
Culture of experimentation: Eppo's accessibility encourages safe testing practices. Experiment on a finer scale during your rolling deployment phases, leading to continuous improvement.
Discover how rolling deployments streamline software updates, reduce risk, and keep your app online. Learn best practices and how to simplify the process.
Table of contents
Ready to go from knowledge to action?
Talk to our team of experts and see why companies like Twitch, DraftKings, and Perplexity use Eppo to power experimentation for every team.
Get a demo
