Now that you have your deployment marker created, you know you’ll be able to quickly see if a regression occurs after a deployment. You can view the marker from the Deployments tab of the Service Health view.
Let’s say you just deployed the
inventory service and want to check how performance is looking.
Click on Service Directory from the navigation bar.
The Deployments tab of the Service Directory page shows the current health of key operations on your services. In this image, you can see the key operations for the
inventoryservice, as well as the deployment marker showing when your deploy happened.
Hovering over the Latency chart, you can see that the 90 and 99 percentiles seem to be experiencing more latency (the blue lines) in the
update-inventoryoperation than before the deploy (the gray lines).
You know that sometimes latency might go up after a deploy, especially when only a portion of the traffic is going to the new service, as you can see in the callout above. Lightstep Observability let’s you compare performance of one deploy against another to root out true latency from expected deployment noise.
Use the Compare to dropdown to select the previous deploy, so you can compare performance of the two and see if this latency is expected.
Lightstep Observability compares “time-shifted” views of performance. That is, it shows how the previous deploy performed for the same amount of time after its deploy, and not how it’s performing during the same time period of the current deploy.
Looks like this latency in the
update-inventoryoperation is not normal. In the magnified view, you can see that a previous deploy (the gray lines) are at a low level of latency.
Had the latency been similar, it would have looked something like this.
In the next step, you’ll learn how to investigate the latency.
What did we learn?
- Lightstep Observability displays deployment markers on the Service Health view, along with information about the traffic going to the recent deployment.
- You can compare performance between two deploys to verify when there truly is an issue.