Best Practices

for DevOps:
Advanced Deployment Patterns

- xebialabs.com -
 WHITE PAPER

Executive Summary
Getting new software to your users without interrupting their experience is hard; how
do you upgrade an application that’s in active use? And if you do take an application
offline for an update, what happens if something goes wrong?

It’s tempting to try to minimize downtime and reduce release risks by implementing
extensive manual processes and procedures—but pre-release work that isn’t
automated slows down delivery and makes it hard to roll out new features.

The good news is that you can avoid time-consuming manual work while reducing
release risks by following DevOps principles to automate the release and deployment
process. And as you automate, using advanced deployment patterns can help you
speed up the software delivery cycle while maintaining control over the way your
applications are deployed.

This white paper gives you insights into the DevOps best practice of advanced
deployment patterns: Blue/Green deployments, rolling updates, canary releases,
dark launches, and feature toggles. It describes how each pattern works, the
advantages and disadvantages of each one, considerations for implementing them,
and best practices when applying them.

Taking Automation to the Next Level with

Advanced Deployment Patterns
Application downtime is expensive, so most enterprises strive to minimize or, better
yet, eliminate it. Reducing downtime when you need to roll out software patches or
new features is a technical challenge. But as an organization starts to automate
software delivery pipelines and deliver applications to Production more often,
avoiding downtime with application release evolves from a “nice-to-have” into
something that’s critical for the business.

But simply automating manual activities isn’t always enough; organizations need to
ensure that DevOps teams implement automated processes in a consistent way, so
that those processes can be scaled across the enterprise. Advanced deployment
patterns provide a flexible structure that you can use when automating the release of
new software.

With deployment patterns, you can control the technical deployment of new software
versions as well as the rollout of new features to users. You can even release
features to limited user groups and test them in Production with much less risk than if
you made them generally available.

1
 WHITE PAPER

Deployment patterns enable you to:

§ Minimize application downtime and work toward the goal of zero downtime

§ Predict and contain release and deployment risks

§ Manage and resolve incidents with minimal impact on end users

§ Achieve predictable, repeatable software deployments

§ Address failed deployments in a reliable, effective way

Blue/Green Deployments
In the Blue/Green deployment pattern, a load balancer directs traffic to the active
(Blue) environment while you upgrade the standby (Green) environment. After smoke
testing the application in the Green environment and establishing that it is operating
correctly, you adjust the load balancer to direct traffic from the Blue environment to
the Green environment.

Advantages of Blue/Green Deployments

The Blue/Green deployment pattern provides a safe way to upgrade applications
without interrupting their use. Blue/Green deployments work particularly well for
monolithic applications that can take significant time to deploy, because you have full
control over the point at which users can access the new version of the software.

Disadvantages of Blue/Green Deployments

Adopting the Blue/Green deployment pattern can increase operational overhead
because you have to maintain duplicate Production environments with identical
infrastructure. Additionally, updating database schemas while following a Blue/Green
approach requires caution, as the new version of the application cannot use the
database until it has been upgraded.

2
 WHITE PAPER

Rolling Back a Blue/Green Deployment

Blue/Green deployments offer fast and straightforward
rollbacks. If the transition to the Green environment
fails, or if something goes wrong once the Green
environment is in active use, you simply adjust the load
balancer to direct traffic back to the Blue environment.

Rolling Updates
In a rolling update, each server node in the environment is taken offline and
upgraded, one by one. After a node is upgraded, manual or automatic smoke testing
determines whether the application is functioning as expected. If the application is up
and running properly, its server node is made available for user traffic, and the next
node is taken offline to be upgraded.

Advantages of Rolling Updates

A rolling update is an incremental way of achieving zero-downtime deployments.
Rolling updates work well for deployments that you expect to execute quickly, so you
can minimize the impact on users.

Disadvantages of Rolling Updates

While a rolling update is executing, the environment will be somewhat unpredictable
because different versions of the application are running in it at the same time. Also,

3
 WHITE PAPER

the application architecture must support running in cluster mode because multiple
instances of the application will access the database during the update.

Finally, taking servers offline means that there is less capacity for user traffic during
the update.

Rolling Back a Rolling Update

To roll back a rolling update, you apply the same
strategy of taking servers offline one-by-one to be
downgraded. This approach ensures that a version of
the application is always available for users.

Canary Releases
In a canary release, you upgrade an application on a subset of the infrastructure and
allow a limited set of users to access the new version. This approach allows you to
test the new software under a Production-like load, evaluate how well it meets users’
needs, and assess whether new features are profitable.

A canary release is similar to a rolling update because it involves releasing a new

feature in a staged manner. However, when you perform a rolling update, you only
verify that the application is technically stable and functional before moving on to the
next stage of the rollout. With a canary release, you evaluate users’ reactions to a
new feature or functionality before deciding whether to release it more widely.

4
 WHITE PAPER

Advantages of Canary Releases

Canary releases are a good way to release experimental or beta features to users
and gather their feedback. For example, you can use a canary release to roll out a
new feature to users in a specific geographic region. If those users like the feature,
you can deploy a canary to another region and test the user response there; or, you
could choose to upgrade all remaining nodes in the environment, possibly by
applying another release pattern to minimize application downtime.

Disadvantages of Canary Releases

Canary releases pose similar risks to rolling updates because multiple versions of the
same application run in the environment while the canary release is deployed. For
that reason, the environment will be somewhat unpredictable during the canary
timeframe, and the application architecture must support running in cluster mode so
that multiple instances can access the database.

Rolling Back a Canary Release

Rolling back a canary release is relatively easy
because you only need to roll back the application
version on servers where the new application version is
deployed. User traffic can be redirected to nodes
running the older version of the application, so users
will not experience application downtime.

Dark Launches and Feature Toggles

Dark launching a feature means releasing an early version to Production so you can
test it with realistic user traffic and usage patterns without actually exposing the
feature to users. Dark launching is particularly effective for testing the back end of a
feature because you can run back-end functionality in the background without the
user being aware of it. When it’s time to give users access to the feature you simply
deploy its front-end interface.

For example, you could implement the back end of a product recommendation
feature on an ecommerce website and test its performance without showing the
recommendations to site visitors. When you’re satisfied with the feature’s
performance, you can deploy the front-end recommendation interface on the website.

Feature toggles—also known as feature flags—allow you to further decouple the

deployment of different software versions from the release of features to users. You
can deploy new versions of an application as often as needed, with certain features
disabled: releasing a feature to users is simply a matter of toggling it “on”.

5
 WHITE PAPER

Advantages of Dark Launches and Feature Toggles

Like canary releases, dark launching and toggling features allow you to roll out new
features in a controlled way. However, dark launches and feature toggles do not
require you to run multiple versions of an application in an environment
simultaneously, which you must do for a canary release.

Disadvantages of a Dark Launch or Feature Toggle

Dark launches and feature toggles can be difficult to adopt because they require you
to change the application that you want to deploy. Development teams must build the
capability to dark launch or toggle a feature into the application at the code level, but
doing so can be a challenge for mature applications with a large code base.

Rolling Back a Dark Launch or

Feature Toggle
Rolling back a feature that has been dark launched or
controlled by a feature toggle is relatively simple. In
the case of a dark launch, you simply remove the
interface that allows users to access the feature. And
in the case of a feature toggle, you disable access to
the feature.

6
 WHITE PAPER

Considerations for Adopting Advanced

Deployment Patterns
While advanced deployment patterns can help you take release and deployment
automation to the next level, keep the following considerations in mind before you
start applying them.

1. You might need to change your applications. Minimizing or eliminating

downtime often requires multiple versions of the application to be active in
Production at the same time, which might require you to change your
applications at the code level.

2. Running multiple application versions at the same time means that the
database must be compatible with all active versions. You must
decouple database updates from application updates. In addition,
Development teams must design database changes to be backward
compatible. In practice, maintaining compatibility has consequences for
both database changes and application logic.

3. You might need to change your infrastructure. Some deployment

patterns require you to create and maintain multiple environments that end
users can access. Many DevOps teams achieve a multi-environment setup
by using virtual infrastructure in the form of cloud-based environments.
They might also containerize applications so that they can more easily stop,
start, scale up, scale down, and load balance them using container
management and orchestration tools.

However, shifting from on-premises infrastructure and traditional

middleware platforms to containers and the cloud cannot be done overnight.
Most applications require architecture changes to run effectively in such
environments.

Best Practices When Using Advanced

Deployment Patterns
While advanced deployment patterns can help improve your DevOps performance,
each pattern involves a certain level of risk. However, there are several best
practices that you can follow to reduce the risk of adopting advanced deployment
patterns.

1. Limit the number of application versions that are live at the same time
to two. This approach reduces the complexity of deployments,
infrastructure management, and database compatibility.

7
 WHITE PAPER

2. Deploy in small increments, only releasing a few features at a time.

This approach, which is one of the core principles of DevOps, reduces
deployment risk because the more frequently you release software, the
more reliable your release process becomes, as teams have a greater
incentive to fix problems and improve the process.

Deploying small increments also reduces lead time, which is the time that
elapses between the start of feature development until that feature is
available to end users. Reduced lead time is one of the key goals of
Continuous Delivery, as it enables you to deliver value to users and to the
market faster.

3. Manage risk by refactoring monolithic applications into microservices

that you can deploy independently. Microservices can give you more
control over the impact of changes to the environment. They can also
increase development throughput and enable faster development cycles, as
new features are typically delivered as one or more microservices and don’t
require changing the entire application.

Get the Most Out of Advanced Deployment

Patterns with Application Release Orchestration
In an enterprise environment, standardizing and automating releases and
deployments is key for optimizing software delivery. Application Release
Orchestration (ARO), which encompasses both release orchestration and
deployment automation, enables you to standardize and automate release and
deployment by orchestrating the build, test, provisioning, configuration management,
and deployment tools in your software delivery pipeline.

ARO functionality is also sometimes called Application Release Automation (ARA),

Continuous Delivery Management, or Continuous Delivery and Release Automation
(CDRA).

You can use Application Release Orchestration to implement advanced deployment

patterns in a consistent way across applications (from monoliths to microservices)
and infrastructure (from on-premises to the cloud). ARO can also help you combine
deployment patterns in sophisticated ways, so you can maximize the benefits of the
patterns you’re using.

The fastest way to get started with advanced deployment patterns is to adopt an
Application Release Orchestration solution that supports them out of the box, so you
don’t have to write scripts or build workflows to create the pattern you want. The right
ARO solution ensures patterns are executed consistently while still supporting
flexibility. An ARO solution that comes with pre-defined deployment pattern

8
 WHITE PAPER

templates allows DevOps teams to try out different patterns and see what works best
for their applications and infrastructure.

The XebiaLabs DevOps Platform is an ARO solution that supports automation and
orchestration for your entire software delivery pipeline. With XebiaLabs, you can
create scalable release and deployment processes that DevOps teams can reuse
across applications and environments—all without writing scripts or building
workflows by hand.

The XebiaLabs DevOps Platform includes fully integrated, out-of-the-box support for
Blue/Green deployments, rolling updates, canary releases, dark launches, and
feature toggles. It lets you take control of release and deployment processes and at
the same time, gives your teams additional flexibility to best meet their applications’
unique needs.

The XebiaLabs DevOps Platform makes release and deployment

practices across the enterprise reusable, scalable, and secure
through:

§ Proven templates that work for Production deployments at

enterprise scale

§ Integrated support for both automated and manual release activities

§ Visibility into the status of running releases and deployments

§ Proactive notification of delays and bottlenecks in the process

§ Detailed error analysis for troubleshooting failed releases

and deployments

§ Fully auditable releases and deployments, with detailed reports

and analytics

§ Granular role-based access control over all applications

and environments

9
 WHITE PAPER

About XebiaLabs
XebiaLabs develops enterprise-scale Continuous Delivery and DevOps software,
providing companies with the visibility, automation and control to deliver software
faster and with less risk. Global market leaders rely on XebiaLabs to meet the
increasing demand for accelerated and more reliable software releases.

The XebiaLabs DevOps Platform for

Continuous Delivery at Enterprise Scale

Release Deployment DevOps

Orchestration Automation Intelligence
Orchestrate, automate Automate Get unprecedented
and get and standardize insight and decision
visibility into release complex application support for your software
pipelines deployments delivery process

For more information and a free trial please visit www.xebialabs.com.

10