devopsy.kinsta.cloud

Tag: SRE

Harness Acquires ChaosNative to Meld Chaos Engineering, DevOps

Harness this week announced it acquired ChaosNative as part of a plan to more deeply integrate chaos engineering with DevOps workflows.

At the same time, the company added a Harness Service Reliability Management (SRM) and Harness Security Testing Orchestration capability to its continuous integration/continuous delivery (CI/CD) platform.

Chaos engineering, as a discipline, refers to experiments on a software system that test its resiliency in the face of an unplanned outage. The goal is to deliberately make resources unavailable to see how the application environment either adjusts or whether it fails.

ChaosNative’s LitmusChaos platform, which manages these chaos engineering tasks, was donated to the Cloud Native Computing Foundation earlier this year.

Scott Sanchez, chief marketing officer for Harness. said the company plans to shift chaos engineering further left as part of an effort to enable developers to test the resiliency of their applications before they are deployed in a production environment.

It’s still early days where adoption of chaos engineering within DevOps workflows is concerned. Many IT leaders’ performance is measured based on the stability of application environments. The idea that they should deliberately break something as part of an effort to make an application more resilient requires a shift in mindset, noted Sanchez.

However, once IT leaders start to appreciate how chaos engineering actually improves the availability of applications it’s only a matter of time before the practice becomes more widely implemented, he added.

In the meantime, Harness is looking to automate more of the lower-level tasks that today’s DevOps teams often manually manage. Harness SRM enables DevOps teams to automatically track service level objectives (SLOs) that are attached to specific pipelines within the Harness CI/CD platform.

The Harness STO module, meanwhile, makes it simpler to orchestrate security testing across multiple pipelines.

In general, there is now more focus on the productivity of DevOps teams as the number of applications that are being built simultaneously increases. The challenge, however, is not just the number of applications but also the dependencies that exist between them.

Harness is making a case for a CI/CD platform that automates DevOps tasks by employing, for example, machine learning algorithms to identify which applications are most likely to fail before a build is completed. The goal is to reduce the amount of time DevOps teams waste waiting for builds to complete only to discover that the application almost immediately failed.

The time, effort and cost of maintaining a DevOps platform will naturally vary as the overall environment becomes more complex. However, as DevOps processes become more automated, the platforms themselves become more accessible to a wider range of organizations. Today, the effort required by a DevOps to set up and then support a CI/CD environment is often considerable.

Nevertheless, as more organizations rely on custom applications to drive digital business processes, it’s only a question of when and to what degree DevOps best practices will be more widely adopted. After all, the total cost of application downtime is only going to rise in the age of digital transformation.

March 23, 2022
Secure Software Summit: Applying Chaos Engineering to Software Security
Today’s software systems are, essentially, controlled chaos—and lightly controlled chaos, at that. This makes it exceptionally challenging to model the behavior of those systems. Our systems are quickly becoming larger and larger, with more and more moving parts. It is not uncommon for enterprises to have over 1,000 microservices and millions of containers running thousands of applications containing many thousands of open source libraries.

Why are things growing more complex? Because from microservices to CI/CD pipelines, from service meshes to blue/green deployments, all these new practices and new technologies help us make our applications and infrastructure more resilient and scalable. The downside is the ever-increasing complexity, which is actually accelerating.

In addition, software has officially taken over every layer of the networking and hardware stack (which sounds like an oxymoron, but it’s containers and virtualization turtles all the way down until you hit bare metal). Software is, by definition, easy to rewrite. Thus, the ease of changing software has further increased the number of changes we make to systems. Which adds still more complexity. In fact, software only ever increases in complexity; never decreases in complexity. Software is always trying to do more and encompass more. For that reason, even when you try to simplify software, you just move the complexity around rather than eliminating it.

Introducing Turbulence

Here is how Netflix, the organization that pioneered chaos engineering, defines the practice: “Chaos engineering is the discipline of experimenting on a distributed system to build confidence in the system’s ability to withstand turbulent conditions.” Turbulence is the opposite of order, but it’s where all systems end up.

In the beginning of a system’s life cycle, there is always an orderly plan for deployment and operations. You always have a clean diagram of your resources, code repositories, containers and always a nice 3D diagram of AWS instances you will deploy to and the network topology.

In reality, our systems are never this simple. After the system goes live, that is when we start learning about how it actually operates through unforeseen events like when marketing asks to refactor the pricing module or an outage to the payments API forces you to hardcode a token or a DNS outage—well, you know about DNS outages. In the span of a year or two, you go from 300 to 800 microservices and you have to add Kubernetes clusters and a container network interface (CNI) to properly secure the networking of those clusters. And so on and so on. The bottom line? Almost all production systems evolve into a state that we do not understand and cannot easily map back to our original systems diagram.

Chaos engineering is a mechanism for dealing with this growing level of complexity. The basic idea behind chaos engineering is asking the computer or application or network (or any other environmental element) a series of questions. Consider a legacy system like a mainframe. How did our legacy systems become so stable? They are well-documented. Engineers feel confident in how the system operates. The question is, was the system always that way? Is the system the same one that they originally designed? The answer is “No.”

Legacy systems actually went through a trial-by-fire; a constant parade of unforeseen surprises. These surprises were, effectively, questions. Through these surprises, engineers learned how the system would actually operate (or wouldn’t operate) in the face of chaotic and unforeseen events. This process created iterative feedback that helped the engineers learn how to make the system continuously more stable.

The point of chaos engineering is to generate this same type of feedback loop without the pain, stress and repercussions of unforeseen (and generally unwanted) events. Thus, chaos engineering is a proactive way of introducing adverse conditions to ensure that the system can still do what it’s supposed to do when unforeseen events transpire. Because everything today is software and change is constant, the questioning must also be constant.

For example, chances are that in the time since we originally designed any given system, it has changed a lot. This means some or all the mitigation and resilience we programmed in—retry logic, failovers, circuit breakers—are old news, designed for an older system. When triggers for those conditions actually occur, those mechanisms of resilience we designed may not perform as we had intended.

Chaos engineers proactively ask the question, “Under the conditions in which you are designed to perform, do you do what you are supposed to?” We introduce the conditions themselves in a controlled manner with a limited blast radius to put the system to the test and verify that our assumptions line up with the intended reality. The practice of chaos engineering is intended to be performed on live, running systems without the risk of taking down critical production assets or ambushing unsuspecting applications and their associated DevOps and development teams. This helps engineering teams reduce uncertainty by building confidence in their understanding of their systems by providing them with enhanced context. It turns out that, when engineers are presented with better context, it considerably improves their ability to make smarter, well-informed decisions where resilience is concerned.

Chaos Engineering for Cybersecurity

Cybersecurity is a context-dependent discipline, dependent on how well you understand the systems you are protecting. Software engineers are constantly changing the system. And they must have the flexibility to change it because their primary task is to constantly move towards business value. Yet to effectively apply cybersecurity practices, we must know what we are trying to secure. From that knowledge, we then can know what we most need to secure and how to do it on a given system in its most current state.

Before an engineer can understand how to build robust security measures that provide effective cybersecurity, the engineer must first understand the system. This reliance on the “system first” is a foundational driver for why our security systems and controls—even the immutable and ephemeral ones—experience considerable engineering drift over time. We are continuously burdened by the dependence on a stateful understanding of the system.

Without having some sort of testing or instrumentation feedback loop in our post-deployment world, we don’t know there is a problem with our security until there is an actual problem. Finding out that the room is on fire after the house has been burning for hours is too late. The result is the self-defeating series of reactive incident fire drills, war rooms and outages that we find ourselves in today. If we do not change this reactive crisis loop of chasing the tiger’s tail, it inevitably leaves us taking two steps back for every step we take forward with engineering teams.

When this drift goes undetected, we are left with our wounds exposed and attackers salivating at the meal to come. Engineers always prefer to find evidence and sources of drift through some internal engineering failure or controlled test rather than experiencing a successful real-world attack. Over time, there is a constant drift between the system and security; engineers are constantly changing the system. As a result, there is frequently a lack of recalibration between systems and the security designed to protect them. The way to achieve this recalibration is through continuous testing or experimentation.

Testing is not chaos engineering. Testing is a verification and validation of something you know as either true or false. This could be vulnerability to a specific CVE or exposure to an attack pattern. In experimentation, you are trying to derive new information and explore the unknown unknowns. From this exploration, you seek to create a better understanding of how a system reacts and then share that with the people who are designing the system.

Live security incidents are not effective measures of detection because, by then, it is already too late in the life cycle to effectively address the problem. Worse, incidents cause stress, fear and irrational behaviors. People freak out. Not surprisingly, a security incident is not a good learning environment. Yet security incidents serve the same purpose: Asking questions about a system or systems and helping teams see how to improve them. Chaos engineering offers a better way to create these questions under conditions where it is easier to learn and observe system behaviors, providing clues to improving security. In other words, the purpose of security chaos engineering is to help us proactively understand our systems and where the security gaps are before an adversary has a chance to take advantage of them.

There are multiple use cases for security chaos engineering including:
- Incident response: Testing efficacy of run books, response plans, and determining whether data collection and alerting systems work as intended
- Security control validation: Introducing conditions that should induce behavior from one or more controls and observing behavior
- Security observability: Introducing unexpected attacks or breaches to see if your systems report and observe the incident accurately
- Compliance monitoring: Applicable across all chaos engineering events by documenting that systems do what they say they are supposed to do
A Practical Example

A team I headed wrote an open source application, ChaoSlingr, designed to generate chaos engineering events for security experiments. The application consisted of three serverless functions to generate the experiment, apply it to infrastructure and then monitor the resulting impact on the system.

One example of how we used ChaoSlingr was an exercise where we induced an open port. Enterprises and engineers inadvertently open ports in firewalls all the time. Sometimes this is a result of misconfiguration. It may also be the result of a software upgrade after which the developer or AppSec engineer forgot to reset the port to the proper security settings. In theory, firewalls should automatically detect an unauthorized open port.

For the experiment, we selected a set of resources that had opted-in to our experiments, selected a security group and opened the port. At the time we were relatively new to Amazon Web Services. We were surprised to find that the firewall only detected an unauthorized port opening 60% of the time. We discovered clear evidence of configuration and security tool drift.

We also noticed that a cloud-native configuration and management tool we had deployed was identifying and alerting the unauthorized port opening almost 100% of the time. We then realized that correlating and responding to incident events of an unauthorized port opening was complicated; it was hard to map the incident back to the log data describing how our systems responded to the event. We added a metadata pointer to each alert, allowing us to more easily correlate the information and save valuable time finding the affected security group and port.

In short, the exercise gave us excellent context on where our systems were drifting and failing to keep up, which of our systems were the most reliable and whether we could quickly and easily respond to these incidents by chasing down the right log files.

Security Chaos Engineering Programs Build Trust

In any environment where we are managing complex systems, trust is critical. You must trust that your systems will handle unforeseen events gracefully and not cost your company money or reputational currency. You must trust that the people designing and building your systems are constantly trying to improve them. You must trust that the security teams protecting your systems are up to the task. As systems grow more and more complex, the number of unknown unknowns multiplies. As humans, with our limited processing capacity, we cannot solve for unknown unknowns at scale, continuously. This is the task of security chaos engineering and the processes around it: To create a programmatic way to experiment against and find answers to unknown unknowns. Your systems, developers, customers and your CEO will thank you for it.

About Secure Software Summit
The Secure Software Summit brought together the world’s leading innovators, practitioners and academics of secure software development to share and teach the latest methods and breakthroughs on secure coding and deployment practices. If you are about developing, releasing and securing software, delivering new features fast and building things right from the start, click on the link below to get access to all of the sessions from this summit.

https://go.shiftleft.io/secure-software-summit-2022-replay
March 18, 2022
What Does AIOps Mean for SREs?
It seems SREs are of two minds when it comes to AIOps. On one hand, AIOps’ potential is pretty exciting. By automating complex workflows and troubleshooting processes, AIOps could make your life as an SRE much easier.

But on the other hand, some SREs may choose to view AIOps with disdain and distrust. They might think of AIOps as just another trendy buzzword that doesn’t live up to the hype and that can become a distraction from the SRE tools that really matter.

Which perspective is right? Should SREs embrace AIOps with open arms, or should they resist marketers’ efforts to position AIOps as the latest, greatest tooling innovation in the IT industry?

Those are subjective questions that we can’t answer definitively, but let’s at least gain some perspective by examining what AIOps means for SREs.

What is AIOps?

If you keep up with IT buzzwords, you’ve probably at least heard the term AIOps—which is short for artificial intelligence for IT operations—and that it refers to the use of AI and machine learning (ML) to help automate IT operations workflows.

The big idea behind AIOps is that, by using AI and ML to perform advanced analysis of large volumes of data from IT systems, IT and SRE teams can solve complex problems more efficiently than they could using a traditional, manual approach.

AIOps can, for example, help to surface the root cause of a performance issue in a complex, multilayered environment like Kubernetes. Or it could make recommendations about how best to resolve an incident.

AIOps entered the IT lexicon in 2016, when Gartner coined the term. Now, six years later, it’s a relatively well-established tool domain.

How SREs View AIOps

Despite the fact that AIOps has been around for some time, it doesn’t yet appear that many SREs have bought into the AIOps revolution. Catchpoint found in a 2021 survey that just 7.5% of SREs reported that AIOps tools delivered high value to their organizations.

It’s unclear exactly why there are such low rates of excitement about AIOps, but we’d speculate that there are a few key factors at play:
- AIOps is a new term for an old idea: Many monitoring and observability tools have included at least basic AI and ML analytics features for a long time, even before tool vendors slapped the AIOps label on their products. SREs probably realize this and, to some extent, view AIOps as an effort by marketers to rebrand functionality that is not actually new.
- AIOps is hard to implement: Setting up an AIOps tool requires integrating it with diverse data sources and customizing it to fit your workflows and environment. It’s possible some SREs view this setup work as requiring more effort than it’s worth.
- AIOps can’t replace human insight: While AIOps tools may be useful to a point, it would be unwise to place blind trust in AIOps-based analyses or recommendations. For this reason, some SREs may believe that AIOps encourages organizations to rely too heavily on automated tools at the cost of the expert analysis and perspective that only SREs can provide. (This is similar to the fact that it sometimes makes sense to prioritize human intuition and expertise over playbooks.)
From an SRE’s perspective, then, AIOps may appear over-hyped, overly complicated and underperforming compared to traditional approaches to SRE.

What SREs can Gain From AIOps

SREs’ wariness toward AIOps is valid—but only to a point. It’s important not to let suspicions about the limitations of AIOps turn into excuses not to use AIOps at all. AIOps has some value to offer to SREs, even if it’s not perfect.

For example, AIOps can play a role in reducing toil. To the extent that AIOps tools can recognize complex patterns or discover interrelated data sets more quickly than human engineers, AIOps reduces the time SREs have to spend manually troubleshooting problems or poring over complicated information.

AIOps also helps to enable a more proactive approach to monitoring and incident management. If AIOps tools can alert SREs to emerging issues before they would otherwise recognize them, AIOps can help the SREs to get in front of the problems before they turn into true incidents. That’s better for SREs and end users alike.

There is also an argument to be made that AIOps can help SREs do more with fewer engineering resources. If you can use AI to automate some aspects of monitoring and incident response, you can maintain the same levels of availability and performance with fewer human engineers on hand.

Conclusion: AI Won’t Replace SREs—But it can Help

None of the above is to say that AIOps can replace SREs, or that it magically solves every problem SREs face. Anyone who believes AIOps is a silver bullet has bought into the marketing hype to an unhealthy degree.

Nonetheless, AIOps tools do offer value to SREs by making their jobs easier in some respects and improving reliability outcomes.

So, while it’s wise to maintain a healthy perspective about the limitations of AIOps, SREs shouldn’t rule out AIOps tools as one way to improve reliability engineering.
March 15, 2022
Analyzing SRE Job Postings
You can find plenty of high-level definitions out on the internet about what site reliability engineering means and what site reliability engineers do.

But if you want to understand what it’s actually like to work as a site reliability engineer, there is perhaps no better source than job descriptions. SRE job ads explain what real companies are looking for and what they expect on a day-to-day basis.

To provide insight on what SRE looks like in practice, we’ll look at some job ads from several leading companies. As we’ll see, each company has a slightly different take on what the site reliability engineer role entails, but it’s possible to identify some core themes that underline what businesses expect.

To understand the content of SRE job descriptions, let’s examine examples from four leading companies.

GitLab’s Job Ad

GitLab’s SRE job description contains most of what you would expect from a typical site reliability engineer job ad. It highlights the importance of having a mix of technical skills and collaboration skills. It also emphasizes expertise in areas like Unix/Linux, infrastructure-as-code and Kubernetes.

What may be slightly surprising, however, is the emphasis that GitLab’s job description places on programming skills. GitLab expects SREs to know how to code in specific languages, like Ruby and Go. The takeaway is that SREs need to know more than just how to use code to solve IT operations problems. They must also be able to develop applications or contribute code to application platforms (the job description mentions “contributing to code in GitLab” as a required SRE technical skill).

Also interesting is the fact that GitLab expects site reliability engineers to be able to write blog posts. Blogging may not be one of the first skills that come to mind when you think of a site reliability engineer’s responsibilities, but at GitLab, at least, it’s part of the job.

LinkedIn’s Job Ad Template

LinkedIn provides an SRE job ad template that businesses can use when posting their own SRE jobs on the site.

LinkedIn’s ad is pretty high-level (which is unsurprising, given that it’s a template rather than an ad for a particular job). For the most part, it doesn’t mention specific tools—although the main exception is programming languages, where it says SREs should know how to code in languages like C, Python, JavaScript and Ruby. Like GitLab, then, LinkedIn clearly sees development as a core SRE job responsibility.
The LinkedIn job description also places a heavy emphasis on the ability to “gather and analyze metrics.” It doesn’t use the word “observability,” but that seems to be what LinkedIn is getting at.

An SRE Job at Microsoft

Continuing the theme of development-centric SRE work, a recent job ad for a specific SRE opening at Microsoft reads a lot like a DevOps engineer job description. The ability to “design, write and deliver software” is the top responsibility listed in the ad, which also mentions familiarity with CI/CD tools.

The Microsoft job description also mentions the importance of written and verbal communication skills. Unlike GitLab’s ad, this one doesn’t say SREs will need to write blog posts; still, it’s clear that companies expect people in this role to be creative communicators in addition to technical experts.

It’s worth noting, too, that Microsoft wants site reliability engineers who know both Windows and Unix. If you need further proof that Unix skills are absolutely central to SRE work, look no further than this job ad from the company that created Windows.

Amazon’s Site Reliability Engineer Ad

Even Amazon—a company that is more of a platform provider than a software vendor—places major emphasis on the ability of SREs to code in one of its recent SRE job ads.

Interestingly, the Amazon job also highlights networking skills as a major qualification. That’s understandable given the complexity of networking in the AWS cloud. But it’s a bit surprising because most other site reliability engineer job ads don’t call out networking expertise as a key skill—despite the fact that network engineering is critical to SRE work.

The Amazon job description mentions the importance of an applicant’s willingness to be on-call, too. We suspect that’s critical in virtually any SRE job, but at least Amazon is upfront about the fact that, at time, you’ll be woken up at 4 a.m. to manage incidents.

Conclusion: Takeaways

There are a few key takeaways from these various SRE job descriptions.

Development skills are absolutely critical in order to be an SRE. Scripting for administration or IT Ops purposes is not enough on its own.

The ability to communicate clearly—including both verbally and in writing—is more important to SRE work than you may realize.

Network engineering skills, too, are a crucial part of SRE jobs, at least at some companies.

In practice, SRE job descriptions sound more similar to DevOps job ads than you might expect.
This is not to say that all SRE roles are identical or that all companies expect the same things from their SREs, of course. But the jobs ads or templates we’ve surveyed do highlight how the high-level understanding of SRE translates to on-the-ground reality at actual companies.
February 17, 2022
The Evolution of Incident Management
Have you ever thought about the history of incident management?

If you’re an SRE, you might be so caught up in the day-to-day work of managing reliability and responding to incidents that you never take time to step back and reflect on the evolution of your role and your responsibilities. And that’s a shame because SREs didn’t invent incident management concepts and strategies on their own.

On the contrary, the way SREs think about incident response, structure incident management teams and rank the priority of incidents owes much to incident management strategies developed in the offline world decades ago. To understand fully what it means to be an SRE today, you have to appreciate this deep history of incident management and response.

So, let’s take a look at that history and examine where modern incident response concepts originated.

Historical Problems in Incident Management

Societies have always had incidents, of course. Fires, floods, infrastructure breakdowns and similar crises have been happening for millennia.

For most of history, however, humans lacked an efficient, purposeful way to manage these sorts of incidents. Response efforts were ad hoc, and their effectiveness owed more than a little of their success to sheer luck.

Particular challenges included:
- Lack of effective, consistent communication between stakeholders.
- Varying organizational structures which made it hard to identify leaders, coordinate response efforts and delegate tasks.
- Inconsistent response strategies.
- Differing approaches to assessing the priority of incidents.
Historically, organizations may have been able to handle incidents well enough if the incidents required response from only one, small group. But the more stakeholders involved, the harder it was to respond quickly and effectively.

Putting Out Fires: The Birth of the ICS

Things began to change for the better when stakeholders started thinking about better ways to put out fires—literally.

By the 1960s, fire chiefs in California realized that they were struggling to respond effectively to the wildfires that broke out every summer. Each year brought worse blazes than the last, with more land burned and more buildings lost. The Laguna fire of 1970 brought matters to a head and was the catalyst for a new approach to incident response for fire agencies.

After assessing what was going wrong, the fire chiefs determined that it wasn’t a lack of equipment or personnel. It was poor coordination among the various firefighting agencies that responded to blazes. Lacking a clear chain of command and a systematic approach to firefighting, the agencies struggled to deploy their resources rapidly and, thus, effectively.

To fix the problem, California fire chiefs developed what became known as the Incident Command System, or ICS. The ICS defined a hierarchy for incident response with an incident commander at the top. It also defined several categories of incident response processes including operations, planning, logistics and finance. And it established a consistent set of terms that stakeholders can use to describe their actions during incident response, which makes it easier to communicate clearly.

Although the ICS was initially conceived to fight fires, it became the de facto standard for organizing incident response strategies of all types.

From ICS to NIMS

The history of incident response doesn’t end with the ICS. A new chapter began in the early 2000s when the U.S. federal government developed an even more comprehensive approach to incident management called the National Incident Management System, or NIMS.

NIMS was born in the wake of the September 11, 2001, terrorist attacks, which underlined the importance of efficient communication not just between different agencies of the same type (like fire departments), but of entirely separate organizations. To achieve this, NIMS expanded upon the principles of the ICS.

In addition to adopting most of the incident command principles and practices included in the ICS, NIMS included standards for coordinating the distribution of resources. It also embraced the concept of the emergency operations center, which is in some ways similar to a network operations center in the digital world.

In some respects, NIMS resembled a compliance framework (although to be clear, that’s not what it is). It includes fourteen management principles, similar to compliance controls, that organizations must implement to manage incidents using a NIMS approach.

Incident Management Today

Obviously, putting out forest fires and responding to terrorist attacks is pretty different from dealing with data center failures or a buggy application deployment. ICS and NIMS weren’t designed for site reliability engineering or IT teams specifically.

Still, the influence of ICS and NIMS on the way SREs think is clear enough. Terminology like “incident commander” comes from these frameworks. So do concepts like shared accountability for incident response processes and the importance of involving all stakeholders—not just technical teams—in incident response.

ICS and NIMS may not be acronyms familiar to most SREs. But they should be, because they are the historical sources of the incident management philosophies that form the foundation for SRE work today and offer valuable lessons for any SRE on the job today.
February 8, 2022
Site Reliability Engineering (SRE) Comes of Age in 2022

The site reliability engineer (SRE) role is still gathering steam across organizations. In January 2022, LinkedIn listed SRE as the 21st job with the highest global demand throughout the past five years. That’s pretty high for such a specific tech role. And, looking to the future, it appears the SRE practice will only continue to gain adoption as a method to support high availability, reliability and improved digital customer experiences. The SRE approach is also essential to meet service-level agreements (SLAs) and internal service-level objectives (SLOs).

I recently met with Kurt Anderson, SRE architect, and Emily Arnott, content writer from Blameless to get their take on the top trends and predictions for the SRE practice in 2022. According to the Blameless team, 2022 will likely see increased adoption of the SRE role throughout companies and across internal departments.

Just as manufacturing sets high standards for its quality assurance, digital applications now require the same high reliability to meet inflated user expectations. To address these new realities, SREs have their work cut out for them—they must confront how to unify disparate DevOps tools and must drill farther down to unearth specific user experience qualms—all without adding to alert fatigue.

Increasing SRE Adoption

As mentioned above, the SRE role has seen tremendous growth in recent years and doesn’t show signs of slowing. SRE as a practice was first popularized by Google in 2015. Today, they define it simply as “what happens when you ask a software engineer to solve an operational problem.” This equates to maintaining high levels of reliability for digital services by creating errors budgets, designing SLAs/SLOs and optimizing process automation over time.

To Arnott, we’re still seeing SRE adoption take off at an exponential rate as more and more companies embrace Google’s SRE doctrine. “During the pandemic, more and more companies were forced to adopt a digital strategy. Reliability has to be fundamental to that strategy.” Arnott foresees SRE becoming a fruitful role at startups, which may initially have been intimidated by the prospect.

“The practice of SRE is a scale issue,” Anderson explained. It can be difficult for time-pressed companies to enact an SRE program when their attention is already so divided. However, as new requirements force warp-speed innovation, we’re seeing greater DevOps adoption. And the SRE role, consequentially, is becoming more of an imperative.

More Unification of Multi-Tool Condition

Most DevOps professionals would agree—we live in a multi-tool world. For example, multi-cluster management concerns are bringing a plethora of new tools to the fold. And computing environments are just as diverse, often combining multiple clouds and composed of hybrid estates that house various clusters of microservices. Even within the same organizations, different teams may opt for differing DevOps tooling. Anderson believes that SREs can bring “more unification to a fractured DevOps adoption across a single company.”

As companies adopt more software, there’s inevitably some degree of tool sprawl, Anderson said. For example, there may be competing methods to monitor the use of digital services. The platform team may be using Datadog, an app team may be using New Relic and another team may be using Prometheus. These tools don’t speak the same language and generate different logs, creating silos.

For Arnott, the future involves accepting the innate complexity of the microservices era and using an agnostic layer that can receive data from anywhere to identify meaningful incidents.

Observability Gets a Bigger UX Emphasis

The SRE role is definitely a trend that has been growing and will continue to grow in importance, Anderson said. And as more organizations turn digital, it’s becoming necessary; with this shift comes enormous pressure to exceed user experience expectations. “Competition and people’s expectations are so high right now,” said Arnott. “If you are unreliable, it’s unacceptable these days.”

Whereas DevOps focuses more on getting applications into production, SRE takes a more UX-focused approach. Take Netflix, for example, which has invested heavily in a core SRE team that ensures robust and reliable user-facing services. “A key underpinning to keeping our customers happy and streaming is a strong focus on reliability,” according to the Netflix Technology Blog.

To cater to this new paradigm, teams will require a deeper understanding of observability based on particular cohorts of users, said Arnott. SREs require precise user experience contexts, such as an uptick of lags or increased error responses. But they also must discover the contributing factors behind these bugs. “SLOs and error budgets are like taking your temperature—they tell you if something is wrong. They don’t tell you why you have a fever.”

SRE Takes Cues From Manufacturing

As reliability becomes more and more central to a company’s brand presence, it’s only natural that something that important would be elevated into the executive suite, Anderson explained. As a reference example, he pointed to the fact that manufacturing industries often appoint a chief reliability officer. In traditional manufacturing, product lines undergo a statistical process control with rigorous testing to ensure quality. Analogous positions exist across industries, such as chief quality officers in financial services.

The takeaway is that as reliability becomes more of defining competitive metric, tech companies may require a similarly high-level C-suite position to align on that idea and to direct teams to implement SRE practices. Of course, the necessity for such a title will entirely depend on organizational size and makeup.

Greater Reliance on SLOs to Avoid Broken SLAs

A service-level agreement (SLA) is typically a formal contract built around the reliability of a digital service with an external partner, which, if breached, may result in financial penalties or other forms of restitution. This agreement could exist between a software provider and an end user but is more strictly enforced in a B2B context.

On the other hand, a service-level objective (SLO) is more of an internal target. SLOs are often set at a lower threshold than SLAs. For example, perhaps an SLA ensures a user login fails no more than 10 out of one million attempts, but the SLO is set at two out of one million.

Setting internal SLOs helps keep engineers abreast of modest customer experience issues before they become a major incident that results in fines. “Some companies treat SLOs as just another alert; they should be representative of user experience,” said Anderson. It makes sense that as SRE initiatives begin to mature, companies will slowly start to adopt even more competitive goals to deliver superior user experiences and avoid unsatisfied partners.

Final Thought: Don’t Play The Blame Game

A big part of the SRE doctrine is the concept of blameless postmortems. Because when you’re in the middle of an incident, the worst thing to do is point fingers. “Blaming triggers a defense mechanism and when you’re in a defensive mode, you can’t respond well,” said Anderson.

Instead, it’s better to keep an open mind when performing these retrospectives—accept that what occurred happened for a reason, and seek to discover that reason to make positive improvements. “What can we change at a systemic level to become more resilient?” inquires Arnott. “What is the origin of that mistake? What information was lacking? What communication should have happened to prevent this?”

Typically, accidents don’t occur intentionally or due to incompetence but rather due to holes in the process. By not playing the blame game, companies are better positioned to fix issues as they progressively release new features. Then, hopefully, with the right SRE approach in place, teams can advance their digital services to meet the high stakes of today’s modern user experience expectations.

January 25, 2022
CNCF Takes LitmusChaos Platform to the Incubation Level

The technical oversight committee (TOC) for the Cloud Native Computing Foundation (CNCF) announced today it is elevating the open source LitmusChaos application testing platform to the incubation level.

LitmusChaos is a chaos engineering platform donated to the CNCF by ChaosNative in 2020. Since then, it has seen adoption within production environments by more than 25 organizations, including Intuit, Lenskart, Orange, Red Hat and VMware.

ChaosNative CEO Uma Mukkara says LitmusChaos was developed to provide DevOps teams with a chaos engineering platform built using containers that makes it easier to scale up and down than existing monolithic alternatives. The platform itself, however, can be used to test both monolithic and microservices-based applications.

The platform consists of a Chaos Operator based on a software development kit (SDK); a ChaosHub for hosting and sharing experiments; Litmus Workflows that declaratively chains experiments (either in sequence or parallel) to build a chaos scenario; a centralized control plane, dubbed ChaosCenter, to design, schedule and monitor Litmus Workflows; Litmus Probes for creating chaos scenarios that automate steady-state validation and remediation actions and a Chaos Observability tool to export Prometheus metrics.

In total, there are now more than 400 contributors to the LitmusChaos project and more than 4,000 pull requests have been made. Mukkara said since the beginning of this year, Litmus operator installations have grown to more than 2,000 per day. Going forward, the goal for 2022 is to significantly increase the number of LitmusChaos integrations there are for a wider range of DevOps platforms, added Mukkara. Additional sets of experiments for Kubernetes and non-Kubernetes targets, improved observability and integration with other platforms—via the open source OpenTelemetry agent software also being advanced by the CNCF—is also planned.

In general, it’s still early days as far as the adoption of chaos engineering with DevOps workflows is concerned. However, Mukkara said that as more organizations embrace observability to better understand IT processes, the number of organizations that also use chaos engineering to test IT resiliency should also increase. In fact, as IT environments become more complex—thanks mainly to the rise of microservices-based applications—chaos engineering approaches to testing may someday be required, noted Mukkara.

In the meantime, there are still a lot of IT professionals that are uncomfortable with chaos engineering approaches to testing that deliberately break underlying services to test the resiliency of an application. The core idea, however, is that no application should have a single point of failure that could lead to it becoming unavailable. In theory, a microservices-based application won’t fail because when a service becomes unavailable calls are automatically rerouted to another service. The performance of that application will degrade, but it won’t go offline. The challenge is that building an application that achieves that level of resiliency using microservices requires a significant amount of DevOps expertise.

Regardless of how an application is constructed, the one thing that is clear is how dependent organizations are on applications to drive revenue. As such, the tolerance for application failures—or performance degradation—has never been lower.

January 11, 2022
Best of 2021 – Hiring When You’re Not FAANG

As we close out 2021, we at devopsy.kinsta.cloud wanted to highlight the most popular articles of the year. Following is the twentieth in our series of the Best of 2021.

Most of us are not FAANG, and we shouldn’t act like we are. Instead of using that, I’m going to use the term we used to use to refer to them: “Web Monsters”, which isn’t 100% accurate, but includes more than just the five FAANG members (Facebook, Amazon, Apple, Netflix and Google) for this article. At some point, we are going to need a new term. Web Monsters doesn’t clearly include Apple or Citi, but FAANG clearly excludes Citi and eBay. But this blog is about something else, so please accept my apologies for that lack of clarity. Assume that ‘Web Monster’ means “Big, popular high-tech company that pays well and looks good on a résumé.”

There are several angles to “Not Web Monsters” that we can address, but I’m mostly going to focus on DevOps hiring, with a nod to irrelevant sales/marketing techniques. Hiring is an easy target because Web Monsters offer resume-building, pay and benefits that are super attractive in the industry, so they can (right now, at least) reject 99% of applicants and still have a good selection of candidates to choose from. Most organizations cannot compete in even one of those areas, though a few can when it comes to salary and/or benefits.

We in IT have long had a bad habit of asking our applicants to be more qualified than the average employee. It’s pretty ridiculous at times; things like asking for a candidate to have more years of experience with a technology than the technology itself has actually existed. What happens when you ask for the impossible? Honest, qualified applicants self-select out, and only those who don’t realize the technology hasn’t been around that long and liars end up applying. That’s just bad hiring policy, even if you are a Web Monster. But at smaller organizations where each hire has a larger impact on the organization, it can be devastating.

It seems worse in DevOps. In fact, all of the hiring items listed here seem worse in DevOps because DevOps is actually a nightmare mish-mash of skill sets. No two people define the job the same because “the job” covers everything from cradle to grave for an application. There is no part of app dev that “DevOps” couldn’t touch. That’s problem number one. Define the responsibilities of the role clearly, don’t ask for everything that it could be. As a rule, an Ansible specialist is not going to be a DevSecOps expert, and you shouldn’t expect them to be. I think, at a minimum, we need to split hiring along my traditional emphasis lines… DevOps and DevOps. If you are writing code and using Checkmarx while GitLab is building the results and you configured it to roll out to test … you’re not a DevOps engineer, even though there’s a bit of ops in there. Your strengths need to be in development and app configuration. Your job description needs to focus on that.

Could there be roles that need it all? Yes, particularly in the SRE or chaos engineering space; I could see that. But they are certainly not entry-level roles, or even close. Knowledge in the breadth of technologies required to touch the entire DevOps toolchain is developed over years and with a healthy curiosity. And, of course, the person you are hiring should have a similar skill set to those doing the job today. Which is a good sanity check against the “Throw in the kitchen sink” syndrome.

Next up is take-home tests and shared coding during interviews. We know that you have a vested interest in landing the best people, but if you are doing coding exercises with people who are primarily DevOps, you are wasting everyone’s time. Most DevOps jobs are done with PowerScript and/or bash with YAML files. Other languages used occasionally have no more place in a core job interview than other occasional bits. “It’s about how you think” is a common refrain. I would argue that an assignment that takes an applicant hours to complete (or in several cases I know of, days) is not an efficient way to evaluate the way a prospective employee thinks. And, in at least one case, I’m pretty certain the interview solved an actual problem that employees didn’t—or couldn’t—solve. Ask the candidate questions about how they’d solve problem X. Ask them convoluted questions. They don’t need to get to code level to show you how they think.

And finally, my nod to bad marketing: Unless you actually are doing the volume of one of the Web Monsters, the fact that Facebook is a customer of vendor X is totally irrelevant. In fact, like the government, most vendors can mark one or more of the Web Monsters as their customers, just because Web Monsters are so big they try a lot of different tools across many different teams. So having them as a customer should be irrelevant, but if it isn’t considered irrelevant, should be expected.

The team you have today is rocking it. Hire more people like them. You know what works and what you need. Get back to customizing hiring processes to deliver more of what you need and stop over-asking. There are more great people out there looking for roles; don’t scare them off by asking for irrelevant experience. Stick to the core. What are they going to do daily? Require that, then probe for other useful items at interview time. And get more good peeps in the seats!

December 29, 2021
Prioritizing Scalability, Reliability and Security in Engineering
As digital products and services are more deeply embedded in critical industries and infrastructure and the implications of problems grow in scope, engineering organizations are renewing their focus on building platforms that are scalable, reliable and secure. More importantly, they are recognizing that this isn’t the responsibility of an SRE or CSO, it’s an ethos that must be carried each day by every engineer on the team. This is an important shift—from teams and cultures that valued speed of delivery above all else to teams and cultures that make high-quality systems the foundation of everything they do.

This isn’t easy, and big cultural changes don’t happen overnight. There are many things that can go wrong when the team is operating in the legacy mode of prioritizing quick time-to-market but there are also some strategies and best practices for beginning the process of cultural change across your engineering organization.

The Old Way of Doing Things

You’re the engineer on call, and you get woken up at 3:00 a.m. by an alert that tells you an app is down. Groggy and annoyed, you look at the message and have no idea what service it’s referring to or where to begin.

After nearly an hour, you can’t figure out who owns the service or how to get it up and running. So, you call that senior engineer on your team—you know the one—the person who gets called every time there’s a tricky question or problem. Thankfully, they answer your call in the middle of the night and help you bring the app back up. But in a month, the same scenario happens again; only this time, when you go to call that senior engineer, they’re no longer there. They quit to go work at a company that cares about reliability.

Best Practices for Today

How can you avoid this vicious cycle? Here are some best practices you can implement to begin shifting focus away from speed and toward scalability, reliability and security.

1. Evangelize the Narrative That the Systems you Work on Matter

At the end of the day, your systems are there to serve your customers and users. Leadership and teammates who embody these values are incredibly important. People should worry about letting their customers down, not about being yelled at by their manager or receiving a bad performance review.

2. Set Explicit Expectations about Uptime

When you say that your app needs to have 99.99% uptime, it doesn’t have the same impact as saying 52.6 minutes of downtime a year. Translate uptime numbers so they’re tangible—and on people’s minds every time there’s an outage.

3. Don’t Sweep Problems Under the Rug

After the dust settles from an incident, it’s important to get together and talk about it. Be transparent about what happened, which person or team made the error and what the implications to users were. The idea isn’t to point fingers or assign blame but to bring the team together so that you can learn, evolve and figure out how to prevent this from happening in the future.

Here are some questions you can use:
1. Who wrote the code?
2. How was it tested?
3. Should our automated testing have caught this?
4. Who reviewed this code?
5. Were they the right people to review the code?
6. Why wasn’t it caught in the staging environment?
7. Who released it and who went through the release validation checklist?
Formalizing this process lets everyone on the team know that mistakes are part of the job and that there’s a standard way that problems are handled. This is a great way to help engineers feel free to do their best work without fear or avoidance.

4. Define Individual Owners for Every Piece of the System

If there’s any confusion about who owns what within your engineering org, you need to address that right away; it’s a massive problem. Not only do you need an owner for every part of the system, but they must also be empowered to fix issues and know that they’re the ones on the hook.

5. Invest in Documentation, Runbooks and Logs

The owner of each piece of the application is also responsible for making sure that documentation is up to date, runbooks are clear and easy to follow and logs are readily accessible. When you’re getting paged at 3:00 a.m. about a service you’ve never heard of before, these resources are key.

6. Prioritize the Delivery Pipeline Like it’s a Key Feature

You should always be thinking about how to build safeguards around your workflows, like automated testing or new tools. This may mean you need to go toe-to-toe with product managers to make sure the work required to build a robust delivery system isn’t getting kicked down to the backlog.

7. Get Business Buy-In by Using Terms They Understand

Unreliable code is technical debt. And debt is something all leaders understand, regardless of their technical abilities. Stripe published a survey in 2018 that found the average developer spends more than 17 hours a week dealing with maintenance issues, costing $300 billion per year globally. If you’re pushing to invest in SRE practices, those numbers might help you make your case.

Culture comes first, but the right tools can reinforce the culture and drive accountability over time. Tools that help teams keep track of critical information such as service owners, documentation and runbooks can help tremendously. Additionally, having easily available dependency maps of the architecture are key to quickly understanding complex systems. Lastly, setting organization-wide standards for best practices and providing templates for spinning up new services that follow these can help ensure that these things are not forgotten.

This article was based on a talk given by the author at Conf42: SRE 2021.
November 12, 2021
Defining Availability, Maintainability and Reliability in SRE

In the world of reliability engineering, you’ll frequently encounter the three “-ability” words: Availability, maintainability and reliability. They sound similar and have similar meanings. In fact, these words may seem so similar that it can be tempting to use them interchangeably.

That would be a mistake. Availability, maintainability and reliability all have distinct—if related—meanings, and they each play different roles in reliability operations.

Definitions of Availability, Maintainability and Reliability

Let’s start by succinctly defining each of the terms.

Availability, Defined

Availability is the extent to which an IT resource is ready to perform a task when requested.

Thus, an application or server that is responding to requests is available (even if it takes longer to respond than desired or otherwise operates suboptimally, in which case it may have a reliability issue but not an availability issue).

Availability is usually measured as a percentage. A resource that has 99% availability, for example, is one that is up and responding 99% of the time.

Maintainability, Defined

Maintainability is a measure of how quickly and easily a resource can be fixed when something goes wrong.

If a buggy application release can be quickly fixed by rolling back to a stable version, the application would have a high degree of maintainability. On the other hand, if you have a server that needs to be rebuilt manually after it fails, it’s not very maintainable.

Reliability, Defined

Reliability is the extent to which a resource functions as required upon request (as opposed to simply being available).

For example, if your users require an application to process each transaction within one second, and it does this 99%of the time while also maintaining a 99% availability level, that would be a relatively reliable application. In contrast, an application that responds to almost all requests but that suffers from high latency or error rates would not be very reliable, although it might be highly available.

Differences Between Availability, Maintainability and Reliability

The simplest way to spell out the differences between availability, maintainability and reliability is to highlight what’s unique about each concept.

Availability

Unlike maintainability and reliability, availability is an essentially binary metric, in the sense that a system is either available or it’s not. Although availability status can change over time, there is no such thing as varying degrees of availability.

Availability also stands out because it is often the most important metric in defining SLAs and SLOs. Contracts typically specify that resources will achieve a certain level of availability, defined in percentage terms. They may sometimes also specify metrics like response times, which are a reflection of reliability, but availability is more likely to be the most important metric within service agreements.

Maintainability

Maintainability is unique in that it’s a pretty subjective concept. A system that one SRE considers easy to maintain could seem difficult to maintain to another SRE. The methodologies that engineers leverage to optimize maintainability could vary, too; for instance, someone from a DevOps background may be more likely to focus on optimizing maintainability within the software delivery chain than someone trained in classic site reliability engineering, which focuses on automating IT operations through code more than on the software delivery process.

That said, automation tools help virtually every team maximize maintainability, no matter their preferences or background.

Reliability

Reliability stands out because it reflects how well a system performs. Here again, this is a somewhat subjective metric, because performance requirements may vary from one user to another. Nonetheless, reliability is the most effective means of measuring whether a system meets the performance levels it needs to, even if those requirements change over time.

Why do availability, maintainability and reliability matter?

Because availability, maintainability and reliability each measure different aspects of a system’s status, putting them together is a useful means of gaining insight into the overall reliability of a system.

In order to be reliable, a system requires both availability and maintainability. A system can’t be reliable if it’s not available. It’s also unlikely to be highly reliable if it takes a long time to fix issues due to low maintainability.

At the same time, by focusing on availability, maintainability and reliability individually, you can drill down into specific issues within the IT resources you manage. For instance, you might have systems that are high-performing when they’re available, but that have low reliability rates because of availability issues. In that case, you’ll know that an investment in increased availability is likely to yield the greatest reward for increasing overall reliability.

The Bottom Line

The bottom line: While availability, maintainability and reliability are all reflections of the quality of an IT resource, they measure quality in different ways. Tracking each category separately is important for ensuring that you know where your weakest links exist within overall system performance and health.

At the same time, however, it’s important to compare and correlate availability, maintainability and reliability data so that you can achieve continuous insight into a resource’s status. When you can track each item separately but also combine them together to gain a complete picture of your system, you are in the best position to optimize reliability operations.

October 25, 2021