Cloud Native Computing Foundation KubeCon + CloudNativeCon Europe 2021, 15 May 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Sponsored Session: Dynatrace - Extending OpenTelemetry to Observability

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon North America 2021 in Los Angeles, CA from October 12-15. Learn more at https://kubecon.io The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Sponsored Session: Dynatrace - Extending OpenTelemetry to Observability

Speakers: Daniel Kaar

OpenTelemetry has delivered significant advancements in application instrumentation. It also powers new capabilities like observability which enables developers and operations teams to get more complete insights into applications and infrastructure. Despite OpenTelemetry’s advantages, it is a piece of the larger observability puzzle. This session will discuss how you can augment OpenTelemetry with automation, analytics and AI to identify and resolve problems faster. The session will center on a demo where we will highlight a real-world case study and focus on the challenges associated with large scale and complex environments. We will show how you can gain deep insights into root cause analysis with a few clicks using OpenTelemetry data.

A

Welcome to cubecon cloudnativecon europe welcome to my session and I'm really excited to present to you what dynatrix can do for you around open telemetry, and some of you might find our session title for today a little bit confusing because you know isn't open, telemetry and observability the same. Are they? Are they not so technically? Open telemetry is a framework to observability right and there is lots of advocacy around that and all the people around. It really think that observability is more than just you know the collection of tools um and the standard data types.

A

You know logs metrics and traces and they they share, or you know the common perception here that it matters what data you collect and how you collect it and then what you do with the data right and here is where open telemetry comes into the play. um It's a way of collecting observability data right and dyno trace can help at both angles, collecting the data and helping to analyze it at scale and on the left side. You can see here the dinotrace1 agent that massively automates and simplifies data collection um and on the right side.

A

uh You know that the brain represents dino traces, ai, which we call davis and you'll you'll, see the power of davis a little bit later in the live demo, um and it helps really you know, processing and analyzing data at scale.

A

Now, if we look at the the full-blown technology stack at standard trace here, which I'm bringing up now on the next slide, um you know on on the bottom left corner in the bottom left corner. You see that dynamite release goes way beyond collecting traces, metrics and logs. You know the weaker look to porch information, which is a key ingredient for for analyzing data at large scale way beyond that.

A

Looking then into user behavior data deeply into code execution flows, um collecting additional metadata for for other purposes and so forth, and then also being very open about they have tons of apis to for allowing you to bring in any data that you want and then, of course, also the open telemetry standard for for ingesting additional data into the dangerous ecosystem into the dietary software intelligence platform.

A

Now, if you go uh one level up, then those those you know key innovations that donates did over the last decade, starting with the one agent, as said before, massively automates data collection and simplifies that, alongside with the purepass, the most advanced tracing to thought technology for more than a decade out there in the market, smart skip topology information that allows you to have. You know everything in context all the time, a real time to purging model of your deployments.

A

That also is the key foundation for the days yeah. It helps you know to analyze the data with true cause and effect relationship and then last but not least doing that at enterprise scale um providing you you know with with uh really high. You know high skillability and failover. You know data collection and so forth.

A

Now, on top of it, we have six solutions tailored to very important use cases around observability and then also the diameter is harp to the far right that helps you to further customize and- and you know, extend the diameter's ecosystem to your needs.

A

Now, specifically today, I want to highlight three use cases that we also replicated in the demo environment from a real. You know scenario with with one of our customers and they are ingesting automatic, ingesting open, telemetry spans through the one agent.

A

So whenever an application or a piece of code is monitored by the one agent dynatrace automatically picks up uh open, telemetry uh spans, then metrics and spam import through a open, telemetry exporter, and uh you know bringing that data into the hana trace through that way, which is your use case number two and three now those isolated and very simply looking use cases. Let me bring that into a much more complex distributed application, and here we can really now demonstrate the power of dyna trace. So let's say we have.

A

You know a little bit of a more complex situation where we have the end user to the far left, using a mobile application or a web browser to interact with an application here with a front-end.

A

Then you know serverless components and then a front and a back end, which is not typically the case, but you, you would typically have a um a very you know: microservice based architecture, these days and then databases and others in the end.

A

Now, if, for instance, your backend now loads, the library that comes pre-instrumented with open telemetry or if you as developers, decide to instrument parts of your code with open telemetry to capture further context, the one agent that already is, they are monitoring this component, this micro service or back end or front and or whatever that is, will automatically pick those.

A

um You know that instrumentation up and invest it into the pure path, and we will see that in the live demo shortly and the other scenario really is that you loads or you know, a container image is loaded or in your container spins up somewhere, um and you know when dynatrace monitors that, as just mentioned, you will automatically get the full stack uh com. You know information here um and, if you're not able to put it on a 361 agent there or you don't want to.

A

um You also can then ingest that data through an exporter and open telemetry export into dynagrace, and this is exactly what we're going to look at now at the live demo. So let me actually go over here.

A

uh We can see here uh in my front-end service just a second ago. We have a lot of failed dynamic requests. So let's investigate that and drill into, and I'd like to check out the elevated failure rate here. So my demo environment there's a little problem here.

A

I I don't get a load at the moment, so I'm looking at historic data from the you know past couple of days and as you can see, there is no recent data coming in uh because you know I don't have any load on the system right now, but we still can see. There is a lots of failed requests coming in and let's look uh uh at the details and analyze those right.

A

So I see a couple of product details that fail at 100 rate, uh but more important here is this checkout card service that also fails with 100 for currency. So let me drill in and analyze all these pure pulse, which is essentially all those different traces that come in here.

A

So I need again extend my time frame here to let's say the last 24 hours to pick up some traces and then, let's just look at the top one here now, I'm now looking at a single execution, a single trace within the front-end server, and what I can see here is the checkout card transaction. So if I click on that and then go over here to code level analysis, I can see here uh this method or the details info.

A

uh Where additional you know span attributes have been captured and they come to life here when I hover over it. So essentially, the developer decided here to capture uh detailed information about what user. Essentially, this checkout card transaction triggered, and in that case that comes in actually quite beneficial, because it helps me to sort of reach out to the eos and apologize for. You know that failed transaction, and maybe you know, give him a little voucher that he comes back and does more business with me.

A

um If we go further down this code execution, we actually also can see that the place order fails, which is then leads me further down in into that. Tray and part of the checkout card transaction is actually a currency converter and if I click here on sort of the call, so the application actually runs on the google account and I'm sort of calling over how to component hosting on being hosted on aws.

A

I can see that call you know comes back with a 500 and if I didn't have the you know additional instrumentation through open telemetry. um I would be blind to the rest here and I could only tell there was a 500 coming back, but remember- and this was the top branch on the previous slide- a foreign component.

A

Where I don't have any me, you know any capabilities to deploy one agent to, but the code luckily was instrumented previously with open telemetry data, and this is essentially where I can go in into the conversion request that goes over to that exchange rate api, and that tells me um I've run into a rate limit right. So I now get to the core of the problem, which means essentially, I'm you know, calling them specific api too often, so a good way to cope with it would be actually implementing.

A

uh You know some sort of caching, so that brings me to use case number three right. So we we talked about instrumenting um or you know, or spanning just through the exporter. We talked about the one engine automatically picking up open, telemetry data, which was up here and use case number three, as well as looking at metrics, and what we also set up here is that dynatrace collects and again I'm going back here into the last seven days for a little bit, a more beautiful chart and more interesting chart. Let me put it that way.

A

um So dynatrace also collects the currency from all those checkout cards transactions through an exporter, and if we actually go into the details here.

A

We can play a little bit with that specific chart and the details, and um I'm I'm essentially looking at the currency here, and you know I could could further. So let me actually go in into we're looking at hipster currency and uh in my case I've split the buy currency, but I also could look in you know. I only want to look at successful transactions or unsuccessful transaction. In my case, all the transaction fails right. I've implemented a bug on purpose here that we have to look at something.

A

So if I'm only looking at you know, split by currency and all the failing transactions, I'm ending up uh should end up actually with the with the same dashboard. Here in my case, let's go back to the dashboard and then look at the ei component that dynatrace brings to the table and before doing that, I would like to briefly make you familiar with the smartscape topology.

A

So let me drill in here and the smartscape topology is really you know the key foundation for all the advanced analytics that dynatrace does with the with with the ai component. So, as I said before, the smart skip is essentially a real. uh You know: up-to-date life model of the application deployment and now, when an anomaly occurs, standard trace can take advantage of that to really understand what is the road cause and the impact of that specific situation.

A

So I've prepared a problem over here um that that we ran into before all the failing uh check out or check out card transactions and down here in my visual resolution path, now, dynatrace visualizes only those entities that are you know related to that specific problem, and if I go up here into the root cause, you know to those components where I have an increased failure rate.

A

I actually come back to what we looked at before right. So I see that 500 uh error, http 500 error. uh We precisely run into that before then, whenever we did sort of walk through the pure paths, and if we go to the details down here, we can also see the checkout card transaction with 46 failed requests, and I could you know, follow the same pattern as before: drill into that and realize what the you know.

A

What the details of those errors were to actually understand, I'm hitting the rate limit of the currency conversion, and that already brings me to the end of the demo. So let's go back and conclude our session here, real quick.

A

If you want to see more of dynatrace go or if you want to try yourself to be more precise, go to dynatrace.com trial and it's really easy to get started to roll out the agent on any specific host. It's as easy as one to three essentially download the agent installer, execute it or you know, run the signal or verify the signature and then install it. And if you want to use an exporter, go to github and find our repository there.

A

The link is available- and this is you know, a way of getting spams and metrics into dynatrace. If you don't want to work with the agent or can't work with the agent and then, of course, there is also you know, tailored instrumentation, available and rollout for kubernetes and other platforms through an operator um once you've roll out the agents.

A

It's really a matter of seconds to minutes that donald truce automatically builds the smartscape model for you and, as you remember, that's the key ingredient for our daily ci to really provide you with uh automated analysis based with the dynatrace purpose, build ai, and that brings me to the end of the session thanks for watching.

A

uh Remember if you like what you saw, go to dynatrace.com trial and enjoy the rest of the conference. Talk to you soon, bye, bye,.