Debugging DevOps Using ValueStream and LightStep

dm03514
dm03514
Jul 21, 2019 · 5 min read

The popularity of DevOps has skyrocketed since its practices were first outlined in Continuous Delivery. Even though it has been nearly 9 years the analysis of DevOps success and efficacy is still dark arts among senior leaders. ValueStream brings DevOps monitoring up to CloudNative speeds. This article shows how ValueStream, backed by LightStep can be used to help decision makers debug their delivery pipelines and help them quickly arrive at delivery root-cause hypotheses.

Consider you’re working for an organization and you would like to ship faster? How would you go about doing it? How do you visualize and begin to understand and map the delivery process of each team in your org? How do you start to analyze candidates for improvement? Where do you begin to focus your efforts?

This has traditionally been a difficult process. Just building a mental model of delivery for a single team (let alone multiple teams) may involved consulting multiple different systems for metrics, interviews, embedding, inferring metrics through proxies, relying on experience or just plain guessing and experimentation. ValueStream, backed by LightStepHQ, completely changes this paradigm. Delivery metrics are stored in a single system providing a uniform view into software delivery (DevOps) performance.

Background

This article assumes that Github issues are used to track software work. Each issue created can generate one or more deployments, and there are 3 deployment types: container, mutable and ami.


NOTE: The data in the examples below was manually generated by submitting Github and Jenkins events to ValueStream. The data is completely valid and representative of the real data that Github and Jenkins webhooks generate with the sole EXCEPTION of the event durations. Since this was automatically generated the event durations in this tutorial are in milliseconds. In real life they would most likely be in hours or days.


Issue Latency (Lead Time)

ValueStream and LightStep offer a birds eye view of an organizations software delivery system. This allows seeing a topdown approach to the amount of time an issue spends open and each major step involved in delivering the issue:

In the issue above the total lead time is represented by the amount of time an issue is open.LightStep correlations (a built in LightStep feature) is then able to analyze what attributes latent operations (issues) have in common. The image below shows and example of LightStep correlations applied to the issue event. LightStep is able to automatically determine that the longest issues are much more likely to have ami based builds:

The graph above shows that the property build.parameter.deploy_type:ami is being analyzed. This is notable because it is a property of the deploy build and not on the issue itself. Since tracing maintains a graph of relationships, it is able to associate events within a trace. In this case, deploys are a build event within an issue:

LightStep is able to associate the issue with a deploy because they model the issue and build as a Directed Acyclic Graph. This allows LightStep to establish a strong relationship between the events in a trace and to leverage properties of events taking place within an Issue in order to provide context into why some issues take longer to complete then others. LightStep is able to automatically show ami based builds are present in the longest issues:

Imagine how long building a technical inventory and understanding all projects, delivery methods and teams would take. ValueStream and OpenTracing (LightStep) has all this information, automatically collected, automatically analyzable in single place.

Inversely LightStep is also able to automatically identify that containerized deploys are associated with the shortest issues:

By leveraging web hooks from popular services (Github, Jenkins) and the OpenTracing data-model, ValueStream is able to collect and surface metrics in a highly actionable easy to operate way.

Deploy Failure Practices

This next example will focus on deubgging deployment latencies and errors. This example will start by showing a more “traditional” debugging approach which filters deployments and aggregates them based on their deployment type:

The aggregate above shows that failures and latencies are most associated with mutable deployment type. In traditional debugging it’s usually the responsibility of an operator to know which aggregates and properties should be used. LightStep is able to automatically determine that mutable build are responsible for the most latent builds:

This same automatic analysis can be applied when filtering on errors (in this case errored deploys):

LightStep is automatically able to determine that failed (errored) builds are more likely to occur during “mutable” type deployments.


This post shows how ValueStream and LightStep can be used together to collect and analyze software delivery (DevOps) metrics. ValueStream is currently available and supports collecting Github Issues and Pull Requests, and Jenkins build events.

References

ValueStream by Operational Analytics, Inc.

ValueStream — DevOps Accelerate Metrics

dm03514

Written by

dm03514

https://stackoverflow.com/users/594589/dm03514

ValueStream by Operational Analytics, Inc.

ValueStream — DevOps Accelerate Metrics

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade