 In this video, I'll demonstrate a few ways you can use Pixie to get immediate visibility into the health of your services without the need for manual instrumentation. Pixie is an open-source Kubernetes observability tool for developers. Pixie automatically captures all network traffic in your cluster using EBPF, a low-level Linux tracing technology. Messages of a supported protocol type, such as HTTP2 and GRPC, are parsed and paired with their responses, making latency, error, and throughput information immediately available after installing Pixie. Installing Pixie takes less than five minutes. You'll also need to install the demo microservices application to follow along with this tutorial. You can find directions for both in the tutorial linked below. When debugging issues with microservices, it helps to start at a high-level view, like a service map, and then drill down into the problem service. Let's use Pixie to see the flow of HTTP traffic between the services in the cluster. I have the live UI open, and my cluster is selected in the cluster drop-down menu at the top. I'm going to select the script drop-down menu and start typing cluster, and select the cluster script. The cluster script shows a graph of the HTTP traffic between the services in your cluster. Hover over an edge to see latency, error, and throughput stats for a particular service pair. Thicker lines indicate more traffic between services. Scroll down to the services table. This table shows latency, error, and throughput rate for all HTTP traffic. However, the inbound throughput and outbound throughput columns reflect all trace network traffic, not just HTTP for this service. To figure out which service is the slowest, click the latency column titled a sort by latency. It's good to check multiple percentiles for latency, not just the average, in order to get a better picture of the overall distribution. This script represents service latency with a box and whisker plot. To see the P50, P90, or P99 quantile values for latency, either expand the row to see the row data in JSON form, here you can see the three quantile values, or select one of the vertical quantile lines on the box and whisker plot to have the latency value updated for that value. Once we have identified a service we are interested in investigating further, we will want to drill down into its detailed latency information. Pixie's UI makes it easy to quickly navigate between Kubernetes resources. Clicking on any pod, node, service, or namespace name in the UI will open a script showing a high-level overview for that entity. From the service column in the services table, let's click on the front-end service. This will open the PX service script with the service argument prefilled with the name of the service that we selected. The service script shows the latency error and throughput over time for all HTTP requests for the service. Scroll down to the sample of slow request table and expand the rec path column. If the service handles multiple types of requests, this table can be used to identify if there is a particular request type that is much slower. This table shows individual requests, so we will see the full path with URL parameters filled in. However, Pixie makes it possible to drill down into logical endpoints. Request latency can vary greatly by endpoint, especially if one of the requests is more database-intensive. However, when there are wildcards in your request path, it can be difficult to drill down into a particular endpoint. Pixie can cluster HTTP requests by logical endpoint substituting an asterisk for the parameter in your request. Let's look at latency by logical service endpoint. Scroll up to the top. We're going to select the PX beta service endpoint script from the script drop-down menu. Select this is a beta script. Select the drop-down menu next to service and type in pxsockshop slash catalog and press enter to rerun the script. Note that Pixie displays service names in the UI in the namespace slash service format. This script shows latency error and throughput for logical endpoint for the given service. Click on the catalog slash asterisk in the endpoints table to see an overview of that individual endpoint with a sample of the slow requests. So here we can see that particular logical endpoint. This video demonstrated just a few of Pixie's community scripts. For more insight into your network traffic, check out the related scripts section of the network monitoring tutorial linked below.