LightStep’s Service Health for Deployments feature helps you understand your services, their deployments, and related performance problems so you can deploy with confidence. Let me show you how it works!
Let’s pretend I’m responsible for a backend service called “inventory,” and my team’s just deployed a new version. LightStep automatically detects deployments based on the performance telemetry the application is reporting and points them out on the service’s latency diagram.
Judging by the graph, looks like a period of high latency occurred immediately after this deployment. Let’s investigate it by comparing it against a period of normal performance.
LightStep analyzes and compares thousands of traces so you don’t have to compare raw data manually. Automatic insights are generated based on what is slow for each individual request.
This histogram compares the latency for both the baseline and regression. It highlights this cluster of slow requests that are particular to the regression.
With the Compare Tags analysis, I can understand the context of my bad performance. Before this deployment, none of the requests had a large batch. However, those requests processing a large batch since the deployment are over a second slower than those that didn’t.
Using the dynamic operation diagram, I can better understand where latency is being generated in the stack. Here, it appears that the write-cache operation is slowing down our requests more than any other operation.
This is confirmed by the Compare Operations feature, which shows that the write-cache operation’s latency is much higher in the regression dataset.
Service Health for Deployments not only allows you to assess service health after a deploy but points you towards the cause of any resulting regressions.