Blog |

Meet Continuous Code Improvement

Meet Continuous Code Improvement
Table of Contents

If your experience as a developer is anything like mine, the best moments are those known as the "flow state.” When distractions drift into the background and all your energy is going in the creative direction of solving the problem at hand. Your brain is directly connected to your users through your code. Months of progress happens in hours. Unfortunately, those moments are rare.

But it’s not emails or Slack messages that are the biggest distractions. The biggest headache is when things should work but don't. We spend too much time and effort banging our heads against the wall just trying to keep things from breaking. It’s not sustainable. Especially in 2020, when every business is now a software business.

I believe that “flow state” is achievable. The entire Rollbar team and I are inspired by the challenge to make it a reality for every developer, every day. But we need to move the focus from fixing to building. And, after 8 years with more than 5,000 businesses relying on Rollbar to build software quickly and painlessly, I’m excited to share our vision for the future.

Today, we’re announcing the next evolution of Rollbar as The Continuous Code Improvement Platform. In support of this evolution, we’re introducing a new brand identity—including a new logo—and we’re releasing unique features to enable a new way to code. We call them AI-Assisted Workflows and Automation-Grade Grouping. Now, it’s possible to deal with errors proactively, instead of being reactive.

Introducing Continuous Code Improvement

Look, errors will always be a part of software development. But Cory Virok and I knew there was a better way to code by dealing with errors proactively. To let developers focus on innovating while giving their customers an error-free experience at the same time. We simply wanted every developer to feel that sense of accomplishment and joy. So, we started Rollbar in 2012 to deliver on that promise.

What’s new about Rollbar’s Continuous Code Improvement Platform? First off, we’re still the same Rollbar that our customers already know and love. But we’re also setting a new vision to help developers realize their “flow state.” We’re moving past simple error monitoring and tracking to actually reduce the time developers spend finding and remediating errors.

We want developers to be:

  • Continuously proactive, fixing critical errors sooner with real-time alerts and automated workflows, minimizing the impact on end users.
  • Continuously learning, with a grouping engine that’s powered by machine learning trained with our database of billions of errors.
  • Continuously predictive, reducing deployment risk with predictive code analytics at each step, from dev to staging to release.

Start Being Proactive, Not Reactive

Software in the last decade has changed. Today’s applications are updated continuously, are made of microservices, and are deployed on cloud infrastructure. But the way developers find and fix errors in their code hasn’t kept up.

There’s still too much reliance on logs, infrastructure monitoring, and application performance monitoring (APM). These tools are great for understanding infrastructure because they help you make sure that everything’s working today, just like it did yesterday. But code isn’t supposed to work the same way it did yesterday. It’s supposed to change. You can’t deal with changing code the same way you deal with stable infrastructure. If development teams don’t design their process to expect the unexpected, they’ll always be one step behind errors.

That’s why Rollbar focuses directly on the errors in the code, unlike traditional monitoring solutions. Our continuous code improvement platform helps developers proactively discover, predict, and remediate errors faster, before users report issues.

The features we’re releasing today help to do just that.

Reducing Noise and Automating Error Response

While Rollbar has offered error response workflows for some time, customers tended to revert to manually resolving issues because the workflows just weren’t smart enough to trigger actions they could trust. But now, because we’re introducing a smarter machine learning-based grouping engine (which I elaborate on below), our workflows are smarter too. We call these AI-Assisted Workflows.

Our AI-Assisted Workflows help to:

  • Proactively stop errors from persisting in production environments.
  • Increase developer productivity by reducing the manual efforts associated with fixing bugs.
  • Lower mean time to awareness (MTTA) and mean time to resolution (MTTR) by handling issues faster.
  • Maintain high quality customer experiences by resolving errors before they impact users.

Today, we’re introducing three AI-assisted workflows:

  1. Rollbar Toolkit for Kubernetes. Our new toolkit features a workflow that integrates with Prometheus, Kubernetes, and Weaveworks Flagger, giving development teams the ability to automatically halt, sustain, or expand progressive deployments using Rollbar error data.
  2. Automated Feature Flag Triggers. Our integration with LaunchDarkly, this workflow enables development teams to instantly turn off any feature flag, at any time, when a new, unique error happens behind that feature flag.
  3. Automated Issue Tracking. Now, users can generate tickets in their favorite issue tracking tool (such as Jira or GitHub) automatically for any new or critical error based on predetermined rules. While this has already been possible in Rollbar, it’s now recommended because of the improvement we’ve made in our grouping engine.

Increasing Error Grouping Accuracy Enables Automation

The reason error response automation is not already the standard is because developers haven’t been able to trust the error signals they receive. Other solutions don’t have the technology to accurately identify unique errors, meaning they get noisy false alarms (“undergrouping”) or missed alarms (“overgrouping”). Other error grouping engines use hard-coded algorithms that remain static, while logging and APM tools are only able to group on rudimentary aspects like severity level or endpoint name.

As a result, those solutions can’t identify unique bugs, so developers have to dig through logs by hand to do it themselves. That’s reactive.

Rollbar’s Automation-Grade Grouping is the next generation of our grouping engine, and it changes everything. Automation-Grade Grouping uses machine learning trained with our database of billions of errors, analyzing frequently occurring error types to accurately identify unique errors. We model this problem as the binary decision of whether two errors are the same or different—and then scale it in real-time. This forms the foundation for our AI-Assisted Workflows.

Continuous Code Improvement Starts Today

Too many developers spend their days monitoring, testing, and fixing bugs. It’s painful for them and it hurts their business too. When businesses don’t pay enough attention to their code, they pay for it with unhappy customers, unhappy developers, failed projects, and an inability to innovate.

At Rollbar, we’re changing that. We’re giving control back to developers, and businesses, to focus on building new features and functionality that can drive customer engagement and revenue goals. I’m excited about how Rollbar is helping developers continuously improve their code. The best part is that you can start using Rollbar for free today.