These materials are © 2020 John Wiley & Sons, Inc.

Any dissemination, distribution, or unauthorized use is strictly prohibited.

 Storage
Patterns for
Kubernetes
Red Hat Special Edition

by Erin Boyd, Sage Weil,

and Karena Angell
with David Greenstein

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Storage Patterns for Kubernetes For Dummies®, Red Hat
Special Edition

Published by
John Wiley & Sons, Inc.
111 River St.
Hoboken, NJ 07030-5774
www.wiley.com
Copyright © 2020 by John Wiley & Sons, Inc.

No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or
by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as
permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without the prior written
permission of the Publisher. Requests to the Publisher for permission should be addressed to the
Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax
(201) 748-6008, or online at http://www.wiley.com/go/permissions.

Trademarks: Wiley, For Dummies, the Dummies Man logo, The Dummies Way, Dummies.com, Making
Everything Easier, and related trade dress are trademarks or registered trademarks of John Wiley & Sons,
Inc. and/or its affiliates in the United States and other countries, and may not be used without written
permission. Red Hat, Red Hat Enterprise Linux, the Shadowman logo, and JBoss are trademarks or
registered trademarks of Red Hat, Inc. or its subsidiaries in the United States and other countries. The
OpenStack Word Mark and OpenStack Logo are either registered trademarks/service marks or
trademarks/service marks of the OpenStack Foundation, in the United States and other countries and are
used with the OpenStack Foundation’s permission. Linux is the registered trademark of Linus Torvalds in
the U.S. and other countries. Java is the registered trademark of Oracle America, Inc. in the United States
and other countries. All other trademarks are the property of their respective owners. John Wiley & Sons,
Inc., is not associated with any product or vendor mentioned in this book.

LIMIT OF LIABILITY/DISCLAIMER OF WARRANTY: THE PUBLISHER AND THE AUTHOR MAKE NO

REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE ACCURACY OR COMPLETENESS OF THE
CONTENTS OF THIS WORK AND SPECIFICALLY DISCLAIM ALL WARRANTIES, INCLUDING WITHOUT
LIMITATION WARRANTIES OF FITNESS FOR A PARTICULAR PURPOSE. NO WARRANTY MAY BE CREATED
OR EXTENDED BY SALES OR PROMOTIONAL MATERIALS.  THE ADVICE AND STRATEGIES CONTAINED
HEREIN MAY NOT BE SUITABLE FOR EVERY SITUATION. THIS WORK IS SOLD WITH THE UNDERSTANDING
THAT THE PUBLISHER IS NOT ENGAGED IN RENDERING LEGAL, ACCOUNTING, OR OTHER PROFESSIONAL
SERVICES. IF PROFESSIONAL ASSISTANCE IS REQUIRED, THE SERVICES OF A COMPETENT PROFESSIONAL
PERSON SHOULD BE SOUGHT.  NEITHER THE PUBLISHER NOR THE AUTHOR SHALL BE LIABLE FOR
DAMAGES ARISING HEREFROM. THE FACT THAT AN ORGANIZATION OR WEBSITE IS REFERRED TO IN
THIS WORK AS A CITATION AND/OR A POTENTIAL SOURCE OF FURTHER INFORMATION DOES NOT MEAN
THAT THE AUTHOR OR THE PUBLISHER ENDORSES THE INFORMATION THE ORGANIZATION OR WEBSITE
MAY PROVIDE OR RECOMMENDATIONS IT MAY MAKE. FURTHER, READERS SHOULD BE AWARE THAT
INTERNET WEBSITES LISTED IN THIS WORK MAY HAVE CHANGED OR DISAPPEARED BETWEEN WHEN
THIS WORK WAS WRITTEN AND WHEN IT IS READ.

For general information on our other products and services, or how to create a custom For Dummies book
for your business or organization, please contact our Business Development Department in the U.S. at
877-409-4177, contact info@dummies.biz, or visit www.wiley.com/go/custompub. For information about
licensing the For Dummies brand for products or services, contact BrandedRights&Licenses@Wiley.com.

ISBN: 978-1-119-66632-5 (pbk); ISBN: 978-1-119-66637-0 (ebk)

Manufactured in the United States of America

10 9 8 7 6 5 4 3 2 1

Publisher’s Acknowledgments

Some of the people who helped bring this book to market include the following:
Project Editor: Editorial Manager: Rev Mengle
Carrie Burchfield-Leighton
Business Development Representative:
Acquisitions Editor: Ashley Coffey Molly Daugherty
Production Editor: Mohammed Zafar Ali

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Introduction
H
ave you run Docker containers and discovered that storage
isn’t as simple as mounting a directory? Perhaps you have
exposure to Kubernetes and have discovered volumes but
need more? The vast and flexible world of hyper-converged infra-
structure and how that can be implemented with Kubernetes can
help. Beyond understanding the pieces and parts, Kubernetes
brings it all together with concrete examples of how your applica-
tions can benefit from enhanced storage capabilities.

About This Book

In this book, we provide you with an overview of Kubernetes stor-
age and how it’s implemented. You review the common storage
mechanisms in Kubernetes and what they’re used for. Building on
the base storage capabilities, you discover the common software-
defined storage systems that can run on top of Kubernetes. A brief
look at some example architectural patterns also illustrate the
benefits of using these technologies together.

You also get practical information to help you apply Kubernetes

patterns to your applications, complete with pictures. And finally,
we give you a brief glimpse at some of the emerging trends in
Kubernetes storage.

Icons Used in This Book

To help guide you through the material, you find a few icons along
the way. They’re intended to grab your attention or to help direct
you through the text.

The Tip icon gives you helpful hints or pointers to something that
may assist you in understanding or implementing the technology
being discussed.

Remember icons help you recall information mentioned else-

where in the text or give you important tidbits to remember.

Introduction 1

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 The warning icon calls your attention to information that may be
a stumbling block or pitfall. As you look to discover or use bits of
info from warning sections, pay extra attention to the detail.

These icons give you a heads-up that the topic immediately fol-
lowing is likely to be deeper in the technical “weeds” than other
passages. You may find this information interesting but not com-
pletely necessary for a higher level understanding of Kubernetes
storage patterns.

Beyond the Book

This book can help you discover more about Kubernetes storage,
Rook, Ceph, and related technology, but if you want resources
beyond what this book offers, we have some insight for you:

»» Kubernetes documentation is always a good place to

start when discovering its ever growing features. Visit
kubernetes.io/docs/home.
»» The Rook Quickstart Guide can help you deploy your own
environment quickly and easily. Find details at https://
rook.io/docs/rook/v1.1/quickstart-toc.html.
»» Find helpful information and videos about NooBaa at
www.noobaa.io.

2 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
»» Discussing persistent storage for
containerized applications

»» Understanding volumes in Kubernetes

»» Using tools to help with operations for

volumes

Chapter  1
Introducing Storage
in Kubernetes

K
ubernetes has emerged as the most mature platform for
managing, configuring, and deploying containers at scale. It
provides you a declarative, API-centric way to describe
­container-based applications that can self-manage, self-heal,
self-scale, and more. In this chapter, you explore the impact of
expanding the automation and orchestration from applications
and networking to also include storage, and you look at how you
can apply these technologies and patterns to your applications.
We also discuss where storage is heading in the Kubernetes
ecosystem.

Persistent Storage in an
Ephemeral World
Containers have changed the way we build, ship, and run appli-
cations. This transformative technology affords greater velocity
and application portability but has also brought new challenges
to technology teams.

CHAPTER 1 Introducing Storage in Kubernetes 3

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Challenges of persisting data in a
­transient system
Kubernetes is the leading orchestrator for container workloads.
When a container stops for one reason or another, Kubernetes is
able to start a new instance almost immediately, making avail-
ability and recoverability something that happens in seconds as
opposed to minutes or even hours. But this new instance of the
application may not have access to the file system from the old
instance, especially in the case of local file systems, so additional
orchestration is required to ensure that the stateful component of
the application is bound to the instance at restart.

This focus on stateless applications and ephemeral systems has

prevented certain stateful and complex applications from being a
good fit for this revolutionary technology. As companies look to
bring applications into Kubernetes, the inability to persist data
through a container’s life cycle is undesirable. People need a way
of dealing with stateful applications and persisting data, all with-
out inhibiting the positive aspects of the self-healing capabilities
of the Kubernetes platform.

Preparing for enterprise workloads

To get a better sense of how to address challenges in Kubernetes,
in this section, you take a look at one of the earliest challenges
that has been successfully addressed: handling log data. The
Kubernetes community found a successful way to address log data
by sending it to an external system. By using solutions such as the
ElasticSearch, Logstash, and Kibana (ELK) Stack, you have a pat-
tern of leveraging an agent in a container alongside applications.
This process enables the developer to send logs to a standardized
end point, which makes it portable and flexible.

In the enterprise, moving storage into external systems isn’t

something new or unique. The use of a Storage Area Network
(SAN) has been a staple practice of enterprises for a number of
years, but typically it has been implemented by delivering a pre-
configured, dedicated volume to a virtual or physical machine.

So, how do you provide a stable storage solution for any type of
content, with an equally composable and portable convention as
the ELK pattern? I’m glad you asked. You don’t want to inhibit
your IT team’s abilities to backup, recover, or secure the data

4 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 either. Fortunately, that’s where Kubernetes steps in. Its features
and capabilities address this challenge, and they’re rapidly grow-
ing and maturing. Kubernetes began with storage features as part
of the Pod API and soon expanded these into standalone inter-
faces. These include

»» Volumes
»» Persistent volumes
»» Storage classes
»» Container Storage Interface (CSI)
A Kubernetes pod is the smallest schedulable resource in Kuber-
netes. It’s a logical collection of containers, which ensures the
encapsulated containers share resources and are hosted on the
same node in your cluster. Typically, a pod consists of a single
container and is an instance of an application.

Volumes in Kubernetes
The Kubernetes volume in its simplest form is a directory that’s
attached to a pod and mounted to one or more containers in the
pod. Such coupling is tight and is a foundational construct for how
data can be presented to a container.

Because of this relationship with the pod, Kubernetes volumes are

designed to have the same life cycle as the pod. If an ephemeral
volume is defined by the pod, for example, its contents don’t sur-
vive a container restart. When a pod is removed from Kubernetes,
any volume relationships are broken and ephemeral volumes will
be destroyed.

Kubernetes volumes are typed, which is how you define the stor-
age mechanism backing the volume. Volume characteristics may
be defined via volume type-specific parameters. This may be a
physical disk or path on the Kubernetes node, a SAN, network
attached storage (NAS), or even a storage service from a cloud
provider — to name a few.

Volumes allow applications to have a defined storage path that

stays consistent through the redeployment of a container. This

CHAPTER 1 Introducing Storage in Kubernetes 5

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 layer of abstraction helps but on its own doesn’t solve all persist-
ent storage concerns.

Special volumes
Kubernetes makes the following special volumes available for use:

»» emptyDir
»» configMap
»» secret
When a pod is scheduled to run on a node, a volume is created
on that node and will remain there until the pod is deleted or
moved to another node. This volume is an emptyDir volume that
begins empty and is backed by the storage of the host itself. Given
the nature of this volume type, it’s not regularly backed up and
should be treated as ephemeral or temporary storage in most
cases, such as caching or working objects. Every pod can make
use of an emptyDir.

The other two types, configMap and secret, are volumes whose
content is actually stored by Kubernetes. This mechanism is often
used to pass configuration information, such as credentials, into
a container and ensure they’re available wherever the pod is
­scheduled to run.

Additional information about configMaps and secrets may be

found at kubectl.docs.kubernetes.io/pages/app_management/
secrets_and_configmaps.html.

Volumes on nodes
Much like the emptyDir (see the preceding section), you can also
define a hostPath volume or a local volume. Both volume types
provide a way for you to specify a local path or storage device on a
node and mount it in a path defined in a container.

hostPath volumes
While hostPath volumes allow you to access the file system of the
Kubernetes node it’s scheduled on, it also means your pod can
get access to other containers running on the host, certificates of
the kubelet, and other sensitive files. Some critical limitations of
hostPath volumes to keep in mind include the following:

6 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 »» The same path may be absent or contain a different version
of the referenced file from node to node, causing irregular
behavior in an application.
»» Kubernetes can’t account for the storage available for
hostPath volumes, which means your available local storage
is in a blind spot of your orchestrator.
»» Typically, only the root user has the ability to write to files
and directories in your hostPath and forces you to run your
container as a privileged container or your processes as root.
This is a bad practice because it’s at odds with the principle
of least privilege (PoLP).
»» Each node that can be targeted by a hostPath pod must be
able to support these pods.

Local volumes
A local volume varies slightly from the hostPath in two ways:

»» It’s a direct reference to a local storage device on a node —

not just a directory.
»» Local volumes provide a more portable solution because the
Kubernetes scheduler is able to calculate what node the pod
should run on based on characteristics of the volume itself.
This is in contrast to the manual process necessary for
hostPath volumes (see the preceding section).

Externally backed volumes

As the relevance and utility of local storage options are exhausted,
the need arises to leverage external storage. Individual nodes may
have storage provided by a SAN, but that’s not the same as being
able to orchestrate the external storage along with the rest of the
instructions for managing an application.

Kubernetes provides a Persistent Volume Claim (PVC), which is

a request for storage. The request defines characteristics of the
storage that are desired by your application, such as

»» Access and volume modes

»» Size of volume

CHAPTER 1 Introducing Storage in Kubernetes 7

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 »» Selectors
»» Storage class
Find details about the PVC elements at https://kubernetes.io/
docs/concepts/storage/persistent-volumes/#persistent
volumeclaims.

In conjunction with PVCs, Kubernetes provides a Persistent Vol-

ume (PV) resource. A PV has historically been a preprovisioned
storage volume but more commonly is provisioned from a Storage
Class dynamically and made available to the cluster for use by a
pod.

In a Kubernetes cluster, when you create a storage class, you

define

»» Provisioners: The utility or component responsible for

interacting with the backend storage system
»» Parameters: Settings or values that you can pass to the
provisioner
»» reclaimPolicy: Defines how the volume should be pro-
cessed when a pod no longer needs it

The reclaimPolicy instructs Kubernetes to either retain or delete

the volume when it’s released. If your volume is retained, you
may need to perform your own volume cleanup and maintenance.

Extending volume support

To make storage system support more extensible, plugin-based
storage driver mechanisms are needed. Instead of only developing
and shipping storage drivers as part of Kubernetes itself, storage
vendors and users wanted a way to provide and use the drivers
they need, and to omit the ones they don’t. So along came CSI
and FlexVolume, which are the out-of-tree extension points for
storage drivers.

Both plugin types allow for administrators to deploy storage driv-

ers that can be loaded dynamically by Kubernetes nodes, and both
provide the ability to provision volumes of any number of storage
types. FlexVolume was introduced first but CSI is emerging as a
more dynamic and comprehensive specification.

8 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 FlexVolume is included for completeness; however, the conven-
tion is being deprecated. Only use it if you must.

The CSI specification outlines a number of storage operations

in a way that can extend to additional container orchestration
platforms.

CHAPTER 1 Introducing Storage in Kubernetes 9

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
»» Discovering real-world Kubernetes
storage examples

»» Moving storage solutions into your

Kubernetes cluster

»» Seeing the practical side of converged

storage systems for your applications

Chapter  2
Looking at the
Convergence of Storage
and Applications

I
n Chapter  1, you discover the numerous capabilities in
Kubernetes to address both ephemeral and persistent storage.
In this chapter, you find out about systems and patterns that
can be added to Kubernetes to solve even more complex storage
needs and architectural challenges.

Facing Storage Concerns in the

Real World
When you think about adopting a new application architecture,
such as cloud native, it seems logical to do it with a new appli-
cation. More times than not, however, you don’t find yourself
only developing new applications. If you’re like most folks, you
have to blend the traditional needs of applications and the func-
tions they serve, while applying emerging application architec-
tures to improve speed, adaptability, and costs. At this point,

CHAPTER 2 Looking at the Convergence of Storage and Applications 11

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Kubernetes and the real world collide, and storage is a big part of
this transformation.

Containers need storage

You build business applications in containers. They implement
business logic, automate decisions, and set out to make the lives
of your users easier. Your applications also produce data — it may
be in the form of files, data tables, reports, and so on. Building
small functional bits of logic into containers is a good decision,
but what do you do with the data?

Without storing data in a persistent volume, the file system of a

container is also ephemeral. When a container restarts, the file
system is reset to whatever the file system state is in the con-
tainer’s image and files not stored in a persistent volume don’t
survive the restart.

You need a sustainable and supportable way to store data and

objects in flight or as output from your applications, while tak-
ing advantage of the portability, scalability, and recoverability of
containers. Instead of fighting the natural tendency of contain-
ers to be stateless and lightweight, the pattern looks to augment
Kubernetes to facilitate the behaviors we want.

If your Kubernetes cluster has access to or is in a public cloud

provider, such as AWS, you have platform services that can pro-
vide the following:

»» Block devices: Elastic Block Store (EBS)

»» File system as a service: Elastic File System (EFS)
»» Databases as a service: Relational Database Service (RDS)
»» Object storage: Simple Storage Service (S3)
Each of these offerings has individual performance characteris-
tics, pros, and cons. You must align your application and work-
load with the proper type of storage for its needs. This alignment
requires multiple unique methods and processes for orchestrating
the various storage mechanisms, and not all are aligned with the
way the application is orchestrated by Kubernetes.

12 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Software-defined systems
include storage
Orchestrating stateless, containerized applications requires adopt-
ing a declarative model of describing your application deployment.
The desired state of your deployment is described as YAML code,
and Kubernetes works to make the deployed system match that
desired state.

So logically, you leverage Kubernetes to provide a consistent

interface between applications and compute, achieving portability
of the application without needing to know the nuances of indi-
vidual hardware. The same can be said for storage services from
the cloud service provider (CSP).

Let’s compare S3 and Google Cloud Storage. Both have a lot of

similar characteristics and capabilities, such as providing means
to create storage buckets, store and list objects, and so on. The API
for interacting with each (the methods and properties), however,
aren’t identical. An application would need to be modified if it
was moved between the two offerings. Likewise, the provisioning
and security conventions of the two offerings are also different.
This is in contrast to the use of block or file-based storage from
Kubernetes, which uses the familiar file system interface provided
by Linux and containers.

GCP, AWS, and Azure all have storage drivers in Kubernetes that
help offload the concerns of managing and provisioning the plat-
form storage services and provide a common interface in PVCs
and Storage Classes. But how do you achieve parity for clusters
running inside corporate data centers, or for storage services you
wish to use when a consistent experience in Kubernetes doesn’t
exist yet? One solution, which is to host your own storage system,
is covered in the next section.

Bringing Storage Orchestration Solutions

into Kubernetes
Capabilities on-premises and across different clouds are different
enough for storage and application orchestration that it can leave
you overwhelmed. Converged infrastructure is an architecture

CHAPTER 2 Looking at the Convergence of Storage and Applications 13

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 where you orchestrate your compute and networking in one plat-
form. In this section, you look at bringing storage into the mix as
well and how doing so with Kubernetes can help you run the same
platform on premise and across multiple cloud providers.

Hyper-converged storage systems

Hyper-converged infrastructure (HCI) is the combination of com-
pute, networking, and storage, all implemented as software, and
all deployed on the same nodes. This software-defined approach
introduces immense flexibility in the underlying hardware, elimi-
nating dependence on specific vendors and specialized compo-
nents. It also affords IT teams the ability to achieve portability
and consistency of management and configuration.

Rook
Rook is a storage orchestrator for Kubernetes. It uses the declara-
tive nature of Kubernetes to make storage services self-healing,
self-scaling, and managed with similar conventions as your con-
tainerized applications. Rook uses a plugin-extensible architec-
ture to help you automate storage administrative tasks such as
deployment, bootstrapping, configuring, scaling, provisioning,
and monitoring.

What does this mean to you? Rook allows you to use a declara-
tive style of management that empowers you to achieve “as code”
capabilities for a multitude of storage types. It implements the
new necessary Kubernetes objects so you have a common inter-
face to provision storage software that provides object, block, and
file system storage types.

Rook goes beyond primitive storage types and also enables you to
orchestrate databases and other storage types in a claims-based
fashion. This common way of dealing with volumes and now other
storage types in Kubernetes is one way it is growing to orchestrate
all storage types.

On its own, Rook isn’t a storage system; it must be coupled with a

portable storage subsystem or service. Rook has operators to sup-
port the following storage backends:

»» Cassandra
»» Ceph

14 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 »» CockroachDB
»» EdgeFS
»» Minio
»» NFS
»» YugabyteDB
Ceph and EdgeFS are the most mature and are actively supported
by the Rook community. The other operators are mentioned for
completeness and are in various stages of development.

Ceph
Ceph is a high-performance, distributed storage platform designed
to bring similar capabilities found in GCP and AWS to any cloud
platform, such as Kubernetes. It provides object storage, block
storage, and distributed file systems backed by a single, reliable
storage cluster running on commodity server hardware.

Ceph is designed to be extremely scalable. It’s based on the Reli-

able Autonomic Distributed Object Store (RADOS), a self-healing
system that distributes and replicates data safely across storage
nodes. It then layers a distributed file system (CephFS), block
storage service (RBD, or RADOS Block Device), and S3-compatible
object storage service (RGW, or RADOS Gateway) on top of RADOS.

Because Ceph is open source and purely software, it can provide

its rich set of storage capabilities and features when deployed
across almost any server hardware and either hard drives, SSDs,
or any combination of the two.

Ceph is widely used in a range of environments, especially pri-

vate cloud deployments, but it can be complicated to properly
­provision and connect a Ceph system to Kubernetes. Rook sim-
plifies the deployment and management of Ceph itself and also
provides an extremely simple and Kubernetes-native way to pro-
vision storage resources (persistent volumes for consumption by
containers). If you want to learn more about Ceph and its compo-
nents, visit https://ceph.io.

Longhorn
Longhorn is another software-defined storage platform, pro-
viding Kubernetes with a storage subsystem which rounds out
the HCI capabilities. Unlike Rook-Ceph, it is a single solution

CHAPTER 2 Looking at the Convergence of Storage and Applications 15

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 encompassing features of both of those projects, but only imple-
ments a distributed block storage solution, which means volumes
can only be consumed by a single pod at a time.

Architectural patterns
If you’ve been reading this chapter up to this point, it may be
helpful to see how Rook-Ceph fits into your Kubernetes cluster.
Take a look at Figure 2-1. You see the relationship of Rook, Ceph,
and Kubernetes components.

FIGURE 2-1: HCI with Kubernetes + Rook-Ceph.

Rook deploys an operator for Ceph, which is responsible for man-

aging the state of the Ceph storage cluster. It will start, stop, or
schedule the necessary Ceph services and components to provide
the desired storage types.

Running your storage ecosystem inside your Kubernetes cluster

means the same technology used to orchestrate your applications
is now being used to orchestrate and ensure the availability of
your storage systems for your applications. Native Kubernetes
objects will be used to represent and manage the storage pro-
vider for your applications, such as Custom Resource Definitions
(CRDs) to represent storage pools, object stores, file systems and
other objects which are all part of the Rook-Ceph system — now
in your cluster.

CRDs are portable objects in the Kubernetes API that describe

user-defined components without modifying Kubernetes itself.

16 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Practical Application of Converged
Storage Patterns
Armed with an understanding of Kubernetes storage and how
Rook-Ceph extends those capabilities, this section helps you
understand how you can make good use of the combined tech-
nologies. Putting HCI to work for you and your applications can
make you more nimble and resilient.

Highlighting critical use cases

In this section, you take a look at actual use cases:

»» Using a cloud provider storage service to externalize data

from the container
»» Using a software-defined storage system to provide a
consistent storage service across different cloud environ-
ments and on-premises

Starting in Figure 2-2, you see the implementation of the appli-

cation and storage patterns in a progressive fashion. You can use
each of these application and storage patterns in the HCI system;
we discuss this in the earlier section “Bringing Storage Orches-
tration Solutions into Kubernetes.” In Figure  2-2, you see an
application that consumes storage via a persistent volume (PV).
If you move from AWS to a different provider such as Google, you
can provision the application in the same way (although your data
won’t move with it — see Chapter 3). However, if you’re running
Kubernetes on-premises, there’s no preexisting storage service
to consume.

FIGURE 2-2: An application attached to a PV backed by Amazon EBS.

CHAPTER 2 Looking at the Convergence of Storage and Applications 17

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 In Figure 2-3, the application can use storage provided by Rook-
Ceph (or another software-defined storage solution), which is in
turn consuming compute and storage resources from any cloud
provider.

FIGURE 2-3: An application that consumes PVs provided by Rook-Ceph in any

environment.

Finally, Figure 2-4 shows your application deployed to a hyper-

converged cluster. The diagram illustrates your ability to make
your applications portable across similar HCI clusters hosted
in two different providers. It’s critical to point out that this is
accomplished with only one set of orchestrations to manage
applications and storage thanks to the common interfaces and
consistency provided by Kubernetes + Rook-Ceph. You can move
your application as-is between providers now.

Keeping abstractions aligned

Normalizing how your applications (and your IT teams) interact
with the entire stack enables you to achieve this goal. With a sin-
gle set of instructions, you can run a hyper-converged Kubernetes
cluster in any cloud or on-premises — yet not change any of your
application manifests. Kubernetes is the single, common compo-
nent that enables you to interact the same way with all providers.

18 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 FIGURE 2-4: Application and storage deployment that is portable between
HCI platforms on-premises and different cloud environments.

Having Kubernetes as the common interface solves challenges

such as

»» Normalizing configuration logic

»» Using the same declarative code to deploy everywhere
»» Providing a consistent security and compliance model across
boundaries or providers

CHAPTER 2 Looking at the Convergence of Storage and Applications 19

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
»» Discovering persistent storage for your
applications

»» Planning for and averting disaster in your

storage platforms

»» Migrating applications and their storage

between platforms

Chapter  3
Multi-Cloud and Hybrid
Cloud Considerations

I
n this chapter, you examine different patterns for dealing with
the complexity of running stateless apps that produce and use
stateful data.

Portability Concerns around Block

Storage Services
The Kubernetes Persistent Volume claim pattern provides a pow-
erful storage abstraction that isolates your application from the
details of the storage system — but only to a point. In practice,
capabilities, performance, cost, and security vary among cloud
storage services, traditional storage providers, and software-
defined solutions.

For example, if you want to run Kubernetes in AWS, you may be

using EBS volumes for block devices. A few limitations in using
that technology include the following:

»» Provisioning and attaching EBS volumes dynamically can be

slow (for example, minutes), depending on the size of the

CHAPTER 3 Multi-Cloud and Hybrid Cloud Considerations 21

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 volume. This has a direct impact on your application start-up
and fail over times.
»» Performance of EBS volumes is directly tied to the size of the
volume. Such a relationship can lead to over-provisioning of
EBS, which has a proportionate financial increase to size (and
speed).
»» EBS is built in a way that it’s scoped to exist in a specific
availability zone (AZ), inhibiting your ability to automatically
fail over to another zone.

Software-defined storage platforms like Rook-Ceph implement

a block device service whose capabilities are determined by the
way the software is deployed and configured, and not by which
server hardware or cloud platforms they’re deployed on. It may
seem strange to deploy a system like Ceph on top of a public cloud
when analogous services are provided by the cloud provider, but
it may be necessary to avoid limitations in the native service, like
those for EBS.

Perhaps more importantly, software-defined solutions can pro-

vide consistency across cloud providers and on-premises deploy-
ments, so operators and application developers don’t need to
worry about inconsistencies. Furthermore, newer multi-cluster
disaster recovery and replication features  — such as asynchro-
nous replication of volumes across data centers and between dif-
ferent cloud providers — may be available.

Looking at Shared File Services

Shared file system services are generally more demanding on the
infrastructure provider because of the complexity of providing a
consistent view of your data from multiple containers or nodes.
For this reason, shared file services (for example, those that pro-
vide NFS-based storage) aren’t always available from all cloud
providers, and when they are, they tend to be very expensive.

It is also worth noting that capabilities between shared file ser-

vices tend to have more variance in behavior and performance
than block-based storage. The consistency model for NFS, for
example, when data is accessed between multiple nodes is much
weaker than a local file system or many other distributed file sys-
tems like CephFS or Lustre.

22 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 For these reasons, it’s often smart to provision your own
­software-defined file solution by using a system like Rook-Ceph
and CephFS.

Object Storage
If your application stores a lot of mostly static content, it may be
best suited to use object storage services as opposed to block or
file storage.

The advent of the public cloud, and in particular Amazon’s S3,

have popularized the use of object storage. Today, object storage
is a critical part of storage infrastructure for most cloud-native
applications. Historically, Kubernetes hasn’t addressed object
storage in any meaningful way  — it’s up to each application to
manage object storage resources and handle connectivity to object
storage services.

Today, Google, Azure, and AWS all provide object storage as a ser-
vice. However, these services all implement slightly different APIs
and security models, and provide slightly different management
capabilities.

The complexity of your application, if it needed to be able to dis-

cover how and when to use each potential API, would be unsus-
tainable. Solving this challenge is precisely what the storage
abstractions in Kubernetes are designed to improve for develop-
ment teams.

ObjectBuckets and ObjectBucketClaims

The Rook and lib-bucket-provisioner projects have recently
brought the same simplicity of on-demand persistent volume
provisioning to object storage. With persistent volumes, an appli-
cation’s declared state can specify a claim for storage that’s
matched against an existing storage class, and the storage class
provisioner handles the platform-specific details of managing the
life cycle and attachment of the volume to a pod.

The same pattern allows applications to request object storage

with an ObjectBucketClaim (OBC) object that’s matched to a stor-
age class. That storage class may represent S3 on AWS, Blob stor-
age on Azure, or object storage from Rook-Ceph  — whatever it

CHAPTER 3 Multi-Cloud and Hybrid Cloud Considerations 23

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 is, a bucket is provisioned, and access credentials and URLs are
passed to the pod so it knows how to connect to the storage.

OBCs solve the provisioning part of the portability problem, but

now the API part: Each cloud provider still implements its own
object storage API.  However, most have a mode that’s compat-
ible with S3, at least for the most commonly used operations.
More importantly, virtually all object storage solutions used on
­premises — including Ceph — implement the S3 API.

NooBaa is an up-and-coming project that implements the S3

bucket protocol, using the Object Bucket (OB) and OB claim pat-
tern. It normalizes how your applications interact with object
storage, while affording you the ability to use multiple cloud pro-
viders on the back end. On the front end, it presents to applica-
tions and consumers an S3-compatible API. In addition, NooBaa
has developed an operator for Rook making deployment a breeze.
By providing such storage abstraction, NooBaa is able so solve
many data portability challenges for hybrid and multi-cloud.
Lastly, NooBaa exposes a method for expressing data policy that,
as we move into complex hybrid cloud and multi cloud manage-
ment, will be critical at scale. Discover more about NooBaa at
www.noobaa.io.

When you save an object in NooBaa, the file is fragmented into

chunks. The chunks are then processed to deduplicate, compress,
and encrypt. Finally, the processed data chunks are stored in your
configured backend storage systems.

lib-bucket-provisioner
lib-bucket-provisioner is a library for Kubernetes that intro-
duces the OB and OBC, OB/OBC objects, and associated workflows,
interfaces, and so on. Discover the details about this project at
github.com/kube-object-storage/lib-bucket-provisioner.

There are examples of the OB/OBC pattern implemented in the

library, the AWS S3 Operator, and Rook-Ceph. Members of the
Kubernetes community who are working on these patterns and
technologies are seeking broader alignment and consensus on
how and when to implement in a more native way.

Challenges faced in implementing the OB/OBC pattern in a more

generic fashion are how to address data policy management in an
appropriate way. As these challenges and others are being looked

24 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 at, you should recognize that providing a native way to deal with
object storage rounds out the most common storage scenarios
directly in the platform. Being able to describe and dynamically
orchestrate, in an API-first fashion, block devices, file systems,
and object stores solve a great number of challenges for develop-
ers and administrators alike.

Concerning Disaster Recovery

You don’t want to think about it, but failure and disaster does
occur. Kubernetes provides a great deal of application and work-
load resiliency, self-healing, and management improvements,
but you need more than just that. As you implement Kubernetes
and grow its usage in your environment, you need to address con-
tinuity at the application, cluster, and even provider level. This
section describes the capabilities and tools available and how you
can use them.

Application stack configuration state

If your application were completely ephemeral, there would only
be the need to extract the application configuration from the old
cluster and apply to the new cluster; however, we know a lot of
business applications also require storage.

Store or stash your Kubernetes objects in a system external to

your cluster, such as a Git repository. When coupled with CI/CD as
part of a DevOps or GitOps methodology, your pipeline becomes
an invaluable tool for deploying your app stack to your platform
of choice.

If you wanted to make your Kubernetes objects even easier to

manage and reuse, you’d leverage Helm charts or Operators. Their
inherent templating capabilities make installing your application
easier in any target environment, modifying only the environ-
ment specific values as necessary.

Learn more about Helm charts at https://helm.sh or Operators

at https://operatorhub.io/what-is-an-operator.

CHAPTER 3 Multi-Cloud and Hybrid Cloud Considerations 25

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Coupling of application stack and
­storage state
Traditionally, disaster recovery (DR) is implemented in one of the
following patterns:

»» Back up and restore, where you reconstruct your environ-

ment in a different data center, provider, or region from
backups
»» Active-passive, in which case you have a “cold” or offline set
of components that need to be brought to the same data
state as your current cluster before you start using it
»» Active-active, which relies on a parallel set of components,
but state and data are replicated so you may fail over
immediately

In the absence of Kubernetes, managing data and applications

across multiple platforms is difficult to do efficiently as the
requirements for each are unique. Hyper-converged storage pat-
terns and consistent platforms provide a means of standardizing
the strategies for DR.  By maintaining your data and application
stacks as Kubernetes native elements, you reduce the complexity
and thus the pressure in the event of a disaster.

In terms of offline backup strategies, Velero is a tool to help you

back up and restore cluster objects in Kubernetes, including your
persistent volumes. It plugs into the underlying storage systems
snapshot capabilities to keep a point-in-time backup of content
and objects for you.

By leveraging Velero’s capabilities, you can

»» Back up and restore your applications and associated data.

(It’s always a good idea to back up before an upgrade or
major release.)
»» Migrate whole sets of workloads between clusters.
»» Replicate your cluster for testing, troubleshooting, and so on.
Velero consists of a server and a command line interface (CLI).
Coupling Velero with Rook-Ceph or NooBaa can be an all-­
inclusive buffet of storage capabilities that will be self-hosted in

26 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 your Kubernetes cluster. Velero triggers snapshots of your vol-
umes, captures the objects from the Kubernetes API, and stores
them all together in its own space in the storage provider.

Migrating Applications and Their State

Moving your apps, their configurations, and associated content
between providers and platforms is a complex challenge that
doesn’t yet have an elegant solution in Kubernetes itself. Work-
load migration commonly happens as a manual process that
includes several steps. This process typically forces an outage to
the application and represents a fair amount of risk in the mar-
gins of error.

As we outline in the preceding section, assume you can freely

move the application itself around — what do you do about the
storage? One strategy is

1. Stop your application where it’s currently running.

2. Move your data manually, either cross region or cross
providers through backup or copy operations.
3. Update your application configuration and cluster to
appropriately associate the new storage to the target
cluster.
4. Redeploy your application in the target environment and
validate its operation.

If your storage is more of a fixed point in the equation, such as

when you’re using NooBaa or any common external storage, you
may take a more active-active migration approach. Stand up your
application in the target environment while the original applica-
tion is still in service, and simply reconfigure your entry point to
direct users to the new environment or cluster.

Your applications and environmental decisions dictate the best

approach, so plan as far ahead as you can for what exit strategy
and continuity requirements you must implement.

CHAPTER 3 Multi-Cloud and Hybrid Cloud Considerations 27

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Managing Storage in a Multi-Cluster/
Hybrid Cloud Environment
Many of the patterns we discuss in this book center on meth-
ods for managing your deployments for a relatively small num-
ber of clusters. As we scale, the complexity increases not only for
ephemeral applications but also even more so for applications
that create and utilize storage.

We discuss in Chapter 2 how software-defined storage can facili-

tate creating a consistent storage interface across clouds and on-
premises, but what technologies are being developed to interact
with the API to make deploying and managing systems easier?
A new project, Crossplane, looks at addressing these challenges.

Crossplane improves multi-cluster environments. More than just

Kubernetes, the entirety of the Crossplane project encompasses
four main feature areas that may be used independently:

»» Crossplane services: Provision managed services from

kubectl
»» Crossplane stacks: Extends Crossplane with new
functionality
»» Crossplane workloads: Defines complete applications and
schedules across clusters, regions, and clouds
»» Crossplane clusters: Manages multiple Kubernetes clusters
from a single control plane

This combination of capabilities points toward being able to

orchestrate and manage your applications and associated storage
systems, including native provider services, from a common con-
trol plane. Find out more at https://crossplane.io.

28 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
»» Getting involved in the Kubernetes
community

»» Keeping an eye on the storage drivers

»» Maintaining your learning curve

Chapter  4
Ten Items for Your
Kubernetes Storage
Checklist

I n this chapter, we give you ten critical items to keep in mind as

you tackle storage in your Kubernetes clusters:

»» Get involved in the Kubernetes community. If you haven’t

already become involved in the vibrant Kubernetes commu-
nity, you should. Open-source software is made and tested
by its community members. Ensure your interests are
represented and join in the community. Learn more at
kubernetes.io/community.
»» Consider using a converged storage platform. If you want
to run your application platform on any cloud, complete
your capabilities by including storage in your cluster. If you
think about portability as applying to individual workloads,
embrace the most portable and commonly supported
capabilities to achieve your goals.
»» Make good use of PersistentVolumeClaims (PVC) in your
application bundles. This is a critical abstraction layer that’s
highly portable but allows developers to describe critical

CHAPTER 4 Ten Items for Your Kubernetes Storage Checklist 29

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 aspects of their volumes such as read/write and size.
Expecting each cluster to be able to satisfy your PVC via a
storage class isn’t an unreasonable assumption.
»» Embrace the ObjectBucketClaim model for provisioning
object storage. Using the OB/OBC claim-based model
insulates you from the different object storage provider
backends much like PVCs do.
»» Understand the limits of emptyDir volume in your
applications. Leveraging ephemeral volumes still provides
persistence beyond the container lifetime and is useful when
you need ramdisks and the like.
»» When possible, be API driven. With the use of both the
Kubernetes API and Custom Resource Definitions (CRDs)
users can create a consistent model for defining their
resources that make their deployments more portable in
the future.
»» Keep an eye on the storage drivers you use in your
environments. As more storage drivers are converting to
Container Storage Interface (CSI), you want to embrace the
go-forward pattern as soon as you’re able to.
»» Mind your Kubernetes versions. As you look at branching
out into multi-cloud deployments, mixed versions will have
varying levels of support for the same features. For example,
volume snapshotting is still an Alpha feature and was
introduced in Kubernetes version 1.12. Alpha features aren’t
enabled by default, so not all 1.12+ clusters will have this
available to you.
»» Get running quickly with operators and Helm charts for
Rook-Ceph, NooBaa, and so on. You can gain a lot of
capability for less effort than deploying these solutions
manually. Critical tools are available at the following
websites:

• operatorhub.io/operator/noobaa-operator
• operatorhub.io/operator/rook-ceph
• github.com/helm/charts/tree/master/stable/velero
»» Never stop learning. Kubernetes is one of the best technolo-
gies we’ve ever worked with. It’s rapidly evolving, with quar-
terly releases. As the ecosystem continues to become more
plugin-centric, the components and tools will continue to
change and grow exponentially. You can’t stop reading and
learning to stay on top of the rate of change to remain current.

30 Storage Patterns for Kubernetes For Dummies, Red Hat Special Edition

These materials are © 2020 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
WILEY END USER LICENSE AGREEMENT
Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.