Cloud Native Computing Foundation KCD Berlin 2022, 18 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Opentelemetry Operator Auto-Instrumenting your NodeJS App - Yuri Oliviera Sa, Red Hat

Description

Node.js applications are fast becoming more frequent and complex through Kubernetes deployments. With that being said, developers are more concerned about generating useful logs, gathering metrics.

Website: https://www.redhat.com/de

Organized by @Microsoft @kubermatic7173 @SysEleven

Thanks to our sponsors @CapgeminiGlobal, @gardenio, @sysdig, @SUSE, @anynines, @redhat, nginx, serve-u

A

We're gonna talk about how the open Telemetry operator could helping you out instrumenting a node.js app uh a little bit about me. My name is Yuri I'm, a Brazilian guy based in Germany and currently I'm a soft engineer at Red Hat uh part of the operator, enablement team, I'm, also contributor of the open, Telemetry project, more specifically for the open, Telemetry operator. For that reason, uh I bring I'm trying to bring you an uh Rio interested uh tool that the open, Telemetry operator will be capable to help you on the application life cycle.

A

My first question is: who is familiar with the open Telemetry on the audience? Oh, that's: pretty cool, okay, uh I! Don't know if you uh uh if anybody knows uh what the open Telemetry is. Basically, uh this is the uh cool definition about the open Telemetry on the project. Open Telemetry is a collection of tools apis in the sdks, but I used to say. The open.

A

Telemetry is uh the middle of your application life cycle, because, basically, you can collect data from anywhere and send those datas processed and your target format to anywhere okay and then don't be afraid with this cool definition, because basically, the open Telemetry stays on the same on the middle of your application. Lifecycle uh chigran has defined and helped to yeah to create this otlp uh format uh with, basically is the open, Telemetry protocol and the specification you can see uh on that slide that helps the developers or uh yeah the SRE teams.

A

uh Choosing the data in the specific format and to help the open Telemetry to understand those datas and to send uh these traces logs or or spans uh properly, uh as I told the open Telemetry is in the middle. uh We have one part of this: uh the Old Tail it's the open Telemetry is the abstraction for open Telemetry.

A

The hotel collector the hotel collector, uh is in the middle that receive data from uh your microservices spread in your kubernetes cluster, or you can have uh L7 prox or a layer serving proxy which you, which the data will pass through and send to the open, Telemetry or also it's possible to collect data from the main Cloud vendors like Azure, AWS or Google Cloud, and once you have those data uh in the open Telemetry, you can process the data and change the format from uh I, don't know Json and to the otlp or otlp to Json or otlp to the Prometheus.

A

uh You, you can process the data in order to send the data uh formatted for your Target application and then on the right side. You have the trace databases. uh Anybody is familiar with the Zipkin or Jaeger and then, basically you can convert the data actually uh Prometheus understand, but sorry uh Diego understand the the hlp, but you can send to Zipkin or or Jaeger and or I, don't know any uh different uh tool that we will analyze uh the information that you sent front and the open telemetry.

A

So uh and then you can ask me: how can you configure your open Telemetry instance? uh It's basically a config map in kubernetes or a config section. When you are deploy your custom resource, The Collector resource. Then you have this four sections: receivers, processors, exporters and services, and then the receivers you can configure multiple receivers that we use. Send the data to the open, Telemetry processors again will change those data.

A

Exports will export to your destination, and then you can just organize on the yeah on the service section uh which are logs, which are traces and which are metrics and then just to give you uh yeah examples of well-known receivers like 20, Kafka or otlp. Zipking also could be a receiver in an exporter and a processor. You have a kind of batch metrics resource attributes. You can change one of them, the information that you get from your source, application and yeah.

A

Well, no also exporters, uh you have the hlp that is based in the grpc, but you have also the oclp communicating using HTTP instead and yeah, the the last one like Prometheus and Zipkin. You could also send the data, but you can ask me: okay, Yuri uh I didn't find uh any kind of uh receiver or exported that we attained my application because I don't know, I have one specific tool in my company, that is a third part or was created by B, and then you can check this repo.

A

This is the open, Telemetry collect or contrib. Let me just go there here.

A

We have a lot of different receivers. As you can see here we have a patch receiver. We have the sorry. We have the elastic search receiver and in flexdb, for example, and if you go back, for example, for the receive exporter lists, you have the fire exporter, you have also the inflex DB lucky.

A

And F5 Cloud you have many different exporter or receiver and processors also, and uh if you want to contribute creating your own receiver or uh processor or exporter, please just phrase a PR or yes. Media will be there how to raise your first PR contribution with the open Telemetry project.

A

So, and you can ask me, okay, Yuri and why uh you are talking about the open Telemetry operator uh owns the open. Telemetry can install a standalone instance and your Linux machine or your kubernetes, and for these two reasons, because when you this is indeed, when you install open Telemetry operator, you are managing automatically one collector and for the second reason we created recently possibility throughout instrument uh using yeah the annotations feature to out instrument some of applications based on python, Java or node.js foreign.

A

Besides, that you have the possibility not to only Auto instrument using uh the the operator, the the open Telemetry operator, but you can also create a manual instrumentation for your applications, and this is the compatibility like C, plus plus go.net. You can use all of them. We have native libraries for them and your developers, sorry, your developers will be able uh to create the otlp format to yeah, to give to the open television to do Telemetry data to the hotel instance.

A

um Okay and what is the requirement uh to deploy the open Telemetry operator and the out instrumentation for your node.js application? First of all, uh when you have the operating Stout, you have to create a resource. A custom resource called open, Telemetry collector that it will be the instance which will collect and Export the data, and we have different modes like deployment or sidecar.

A

In that case that you want to out instrument your application, then we should set the mode as a sidecar, the second one we have created a new custom resource called instrumentation, and then you have to create a new one and the annotations also that we that I will present on the next slides.

A

We have this uh four annotations that are mandatory to deploy your alt instrumentation. If you are deploying for Java, you have to include this inject Dash Java equals true, node.js, Python and so on and so forth, and this injection equals true that will enable The Collector to inject a sidecar container in your deployment.

A

So uh this is a code snippet about the instrumentation customer custom resource. uh You have on the spec section, like the exporter propagators in sample in simpler, sorry and the the inch Point. Basically, you have to put the address, uh which is the address of your collector, because this open Telemetry can be in another kubernetes cluster, not in the same and the propagators. Also, you have this uh these different propagators. That is a basically the header format defined by the w3c or the B3.

A

You have you could Define different formats, different header formats, uh for your data and how they spans will be collected. You can Define on the type of the simpler, simpler section.

A

uh Basically, this is that was uh overview about the yeah, the open, the open Telemetry or what open Telemetry is uh I I'll, bring a demo just to demonstrate you. The open, Telemetry action and I have here a node.js a basic example of node.js I, just reload, node.js and eaeger to demonstrate all the traces the open Telemetry collect for this simple application sent to to the Jager 2.

A

A

I have on the namespace called dashing apps one, my deployment with Sidecar, okay and on that I will check that on emo format.

A

I, have these three annotations: I have the first one as I told you the container names it will be possible so because, if your application has more than one container inside at the same part, you can Define for the open Telemetry which containers will be instrumented. And then, in that case, uh I decided to instrument both containers and uh the inject node.js is true, because I'm trying to out instrument a node.js application and the sidecar will be injected by the by the open, Telemetry.

A

Okay and besides, that, I will demonstrate you, the uh the open Telemetry here just a second.

A

Sorry is on the app foreign I have created that and I will show you.

A

For example, in my case, I decided to Define receivers uh as a otlp and the Jaeger and I put both um uh both receivers so glp and Jaeger on the service section code here, eager to Saint traces and the receiver to get traces from the otlp.

A

Okay, it's pretty straightforward because you have just tuned form uh and Define any form to the open, telemeter. Okay, you are getting data from the hlp and you are sending data to the Jaeger.

A

uh I've also deployed an Jaeger uh here. Let me try to yeah, send to some data here, just open the port and do a curve.

A

So two or three: it's not it's not too much just to generate data, and here on the Jaeger UI.

A

If I, for example, try to the last hour try to find the the traces you can see here, all the operations that uh were called by the my app application and then we have the DNS lookup, get slash because I just did a curve with the slash, because it suggests reload word application and if you see, for example, here on the get find traces you can enter on the traces and get for example, process and eager just demonstrate you, okay and this Trace came from and the my app and the my deployment with Sidecar the plan deployment name and the Telemetry.

A

This is a node.js application. The process ID was the 25 and, and so is for um yeah. You can enter in a different, also https get sorry.

A

Middle quality like that and then you see, then you're gonna see the process also, and then it's a it's really helpful, because imagine that you have a situation in your development team that that your application has no uh your develop developers has no visibility uh over your application. Then, if you use uh the open Telemetry with the out instrumentation for different languages like python, Java or node.js, you don't need to write any kind of line of code to instrument or to get this overview about your uh application.

A

And then it's in my point of view, it's really helpful uh to at least debug your web application. uh Besides, that.

A

Besides that, the takeaways of this pretty straightforward talk is again: if you have no visibility uh over your application, please uh using use open Telemetry because it will help uh to see what is the the current status. If uh yeah, you have any kind of method or call that is getting stuck uh yeah in your database or somehow communicating with another microservices you're gonna see this visibility.

A

uh The out instrumental so will help the yeah the development, the development team and to to leverage uh the quality of data and prove and or improve uh the application. Somehow, and as I uh mentioned the beginning of this presentation, uh if you can see a new receiver or export or if you are facing uh any kind of problem, then I encourage you yeah to start contributing with this amazing project, uh sorry to be biased, because it will yeah help as a soft Engineers or the SRA team to improve our applications.

A