.Net Foundation .Net Community Standups, 20 Oct 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: ASP.NET Community Standup - Distributed Tracing in .NET

Description

Sourabh Shirhatti shows how to use distributed tracing in your cloud native .NET applications.

Featuring: Sourabh Shirhatti (@sshirhatti)

Community Links: https://www.theurlist.com/aspnet-standup-2020-10-20

#aspnet #dotnet #OpenTelemetry

A

A

A

All right, hello, welcome to the asp.net community stand-up, I'm john galloway, I'm a pm on the.net community team and today we're joined by saurabh srihadi, and so you work on a lot of cool stuff. um But I I think some of the the stuff we've talked to you recently about is is very like it's server oriented, and can you describe your.

B

Yeah, so one of the things that I work on now is diagnostics and observability of the dotnet platform, all up so irregardless of the application model you use you're, probably interested in figuring out, what's happening. What's going wrong, you know maybe you're curious in like performance insights, why something is taking so long. So generally, you know that encompasses uh diagnostics and observability, and you know, we've been in the dot-net space. We've been working all the way from you know like logs and metrics that everyone uses to deep diagnostics and dump analysis.

B

So there's a lot of good stuff happening in that area, but today I kind of wanted to you know hop on the show. Obviously, hang out with you john and talk about distributed tracing.

A

Awesome, and just you know, I think suspense is good, and you were talking to me before the call you've got a very cool webcam, going that you said it's blue screen to your computer a few times recently, so just yeah. So this is that could be like you know. Just if you know if anyone loves suspenseful tv shows, this could be that so.

B

Fingers crossed, let's see what happens.

A

We'll see how it goes. Okay, I will jump right into the community. Links got some fun interesting ones today, um as always, I share these at the end, actually I'll I'll, throw them in the, um and so you can follow along community links and we'll see. If I can type there we go so I threw those out in the chat all right. So here we go first of all, one really exciting one james uh just posted about this last week. So this is the dot net live tv and what's really cool with this?

A

Is we started with the asp.net community stand up like back when dot net core was still being kind of designed and since then over time, james and I worked together and we started the thursday shows up and those include like desktop tooling.

A

um You know all kinds of like mobile, all the different kind of areas and then since then, we've had it on wednesday shows, and so we have ef and machine learning, and so especially this year with everyone, you know working from home and- and um you know we felt like these shows- are really important, so we built out and and james ran this- it's amazing.

A

So this is this whole.net live tv, so it includes not just the community stand-ups, but it also includes things like jeff fritz doing the c sharp um there's this c-sharp corner with insta-fluff, there's the.net docs show, so this is kind of your one place for watching all the shows. um So this is neat one. One thing I want to call out here: that's really cool we've got the blazer focus show starting up soon, so safiya abdullah is going to be running that, and so this is basically.

A

This is where you can find all the stuff um so really excited about this, and this site is built using blazer, which is really neat. So I've uh I've been able to like watch as james and the blazer teams, doing code reviews and helping um get it all set up and everything. So it's it's. It's really. um It's a it's a cool site and I'm really excited about it.

B

I hear you had an interesting outage on the new site last week. Oh.

A

Let's worry about.

B

A

No actually, well I mean it's been up and down I for oh yeah that was certs right. There was a key or a.

B

Shirt rotation.

A

Issue yeah, so there's there's that there's been all kinds of interesting things. I've been learning as far as how they do um loading- and you know um some of the things were originally like this. Some of the site was originally the code that we wrote was razer pages and then they did blazer server and now it's blazer webassembly, and um so it's really pretty cutting edge stuff, um how they're doing it so yeah, okay. This is an interesting one I saw come across.

A

This is from alex and he's writing about uh he's asking some interesting questions. You're saying why are web apps here he wants to. You know, look at, you know buying something on amazon and it's it's. You know it's a huge.

A

It takes like several seconds to load and there's a ton of data being being sent over the wire and stuff, and then he compares that to um you know something like doom, and you know he says like what's the frames per second that we get in, you know when you're when you're just like running a doom app or you know what I mean running- a game- why? Why are these web applications so slow, so he actually does uh somewhere towards the beginning. I'm I'm missing it here, but he does talk about the like.

A

Why are they not that quick and he's got an interesting system that he's written up here, and this does it's dependency tracking um for everything in the page and he actually it's it's. um He uses blazer on the front end and he's he's using web websockets and he's actually got this library called fusion, and so it's doing doing this.

A

um Actually, it's not just tied to blazer he's got a sample using blazer, but he's got a lot of different ones, but the idea here is where he's doing change tracking and allowing for very rapid updates, um based on uh tracking just the things that need to be updated so, instead of like, if you look at something like one of these page loads, you're, reloading tons of stuff and you're, sending tons information over the wire and what he's building is examples that are just doing very um tight change tracking.

A

So it's very interesting he's got a lot of interesting um samples that you can dig into in his repos. So I I'd love to hear comments on this. How people are um you know how, uh if, if people could look into this and see see what they think, I think there's some very interesting ideas there all right, here's an interesting one from madhu and he's looking at. Can I write a blazer component with no no html and no events, and so basically just write a class and it's kind of a just.

A

Did you know you could do this sort of thing? So here he's creating a class inheriting from my component and then he's got some handlers.

A

So, for instance, he's got a rendered delegate and he's got set parameters async and he's got this eye component attached and using that this is a simple self-contained class that renders, in this case this um this component. So you know he's a lot of. It is just interesting to see how it's, how it's wired up and how the kind of blazer internals work. I do think this could be interesting for distributing uh classes and stuff. It's just very you know tight single uh c sharp file so neat stuff.

A

All right, brian laguna, is writing about using npm packages in blazer. So there's a lot of great libraries out there, javascript libraries that are available in npm- and you know, especially at this point in where we're at with blazer blazer itself being relatively new and they're a huge there's, not always a one-to-one mapping of everything you want that's available as a blazer component, there's more stuff out there in npm, so he's showing how to set that up. So he does a walkthrough.

A

He shows you know using setting up npm directly in your app uh doing installing and then like with webpack cli as a developer dependency, setting up the build script for it and then integrating it into your blazer application. And then here it shows you know, npm build run and you can set up your cs proj to actually run those npm tasks, so there's definitely different ways to do this. I've seen also people just taking javascript and wrapping that using the blazer.

A

um You know using blazer, interop and creating a light wrapper component, but this is interesting to see the full kind of through using npm all right, andrew lock again with an amazing series, so we've been showing a few of these here, he's showing on running with kubernetes, and here he's looking specifically at running database migration. So he's actually got two here.

A

He's got one on running when deploying and then this specific run one is on using jobs and init containers, and so here he talks through uh how he's how he's running this- and this is always you know, database migrations- are, are tricky and there's different ways to do it um a lot of time. I've heard from different teams. Some are it's only.

A

You know, that's something that the dbas run, and you know we don't do anything with that or because it can be a little tricky how you handle it, um but so here's the example using using that setting that up using the init containers he talks through some of the different options. You know you can use ef core migrations from global tool. You can execute them manually, calling database migrate.

A

You can use some great libraries out there or uh using some sort of different tools, so here he's got a console application as and he's showing using a migrator doing that. So um so you know a nice kind of walk through setting that up um curious in the in the chat as well. If you've got a system that you use for doing migrations so uh and then of course uh he ends, I love how his posts end up.

A

He he always, you know, gives his his link to a source, and then he talks about um you know he sums it up and talks about how he's he's got that set up and um using a helm chart installation. So this is just something that applies for everybody. This is really cool. This is uh code navigation for c sharp uh repositories in github, so this was just announced yesterday. So now, as you are browsing your c-sharp repositories, you actually get this nice semantic code navigation.

A

I want to shout out specifically a lot of people, but I know damien guard put a ton of work on this. This goes back to this issue was opened june 2019, and it's taken a lot of work. Updating these um um tree sitters and everything so they're called out in the um announcement post here on github, but I just a big thank you to the the folks that set that up. That's gonna make net coding easier.

B

That's definitely been a godsend in terms of activity.

A

Yep exactly being able directly in the in the github interface to be able to search through so yeah uh cool. This is a cool one from shahed, so he's writing about debugging, multiple.net, core projects and vs code. um So he answered this question um from an individual and so might as well just write it up as a blog post, and so he shows how to set up a launch. Configuration has launched json.

A

um This is again, like you know, regularly you'll have like a front-end app and a back-end api, or maybe a few different apis and so running and debugging. Those all uh can be a pain, and so this is nice to just set up a launch configuration and then also this is nice. With the launch panel and being able to have with the terminal, you can have multiple terminals as well. So that's nice to be able to do that.

A

So that's that all right christos. So this is something that they've been coding up on their 425 show. They were on our show um gosh about a month ago now and talking about integration integrating with. uh In that case they were doing microsoft.identity.web.

A

So here's here they're showing accessing microsoft graph from grpc using azure ad.

A

So uh there's a lot here he's you know, talks about benefits of grpc talks about all the integration, a lot of screenshots here, of course, because we've got to go through and set up the azure id integration.

A

So he does all the app registration stuff and updating grpc service et cetera, but then once you're done and you're running it, it's wonderful! You do the azure id authentication, it works everyone's happy, so he shares out the code. And then, if you want to see more about it, you can interact directly with them on tuesday mornings. They have. This 4 25 show all right yarp. So this is cool yarp just released 100 preview, 6 and some neat things here.

A

There's support for uh http helper for query, transforms header base, routing which, which um I understand was an important user request and so cool to see as they continue rolling this out and they have support for 500rc2. So that's good!

A

Okay, here's another blazer one! uh This is neil swimberger and he's talking about a real-time application with blazer server and firestore. So I actually have two clustered together here: firestore is a google technology um on google cloud platform and he's showing integrating that so there's firebase and firestore, and he does the walkthrough creating the application, uh creating the firestore backend, um integrating that in with a.net console application, and then we also see uh later integrating with the real-time real-time ui, with blazer server in firestore.

A

So this is cool. So it's neat uh because blazer is you know it's web technologies? It's all! It's all. uh You know open standards and all that can integrate with whatever backend so cool, to see that and then so, there's, google and here's a cool one from aws and so aws has this open source portingassistant4.net.

A

So this is net in general. Any up, you know updating any net framework to net core. um What's they'd previously announced this in july, and what they are announcing here is that it's open source they're, open sourcing it. So this is cool. This is something that'll analyze, dotnet framework, dot, net framework applications and help. You understand um what you'll need to do in order to move to dot net core, as well as give some recommendations.

A

So, for instance, he gives an example here of configuring, a recommendation for like system data, sql, client, um and talking about you, know, here's here's! What your recommended update will be, you know, add a reference to this, so these are all these rules. They have pre-configured ones and you can extend them as well. So you can extend using these json things. You can also they have support for um rosalind analyzers, and so he talks about there. There's this uh here, he's talking about in the codalyzer repository and using rosalind based code analyzer for that.

A

So this looks like a really. um You know: help helpful utility as you're analyzing your applications, there's of course other things you see, um porting assistant and there's things in the docs where we make recommendations, but this this is a really nice kind of automated thing that you can point at your source code.

A

All right, just a couple more um so this is um actually. This is the last one, because this one here is a dupe. So this is a neat one. This is a community roadmap.

A

um There's been a lot of this has been you know, built up over um there's been years of contributions on this, and this is neat. This is a asp.net core developer in 2021.

A

So looking ahead, here's kind of a map of things to learn, so you know talking about the general development skills, c-sharp, solid asp.net basics and then getting into more advanced things, so like task scheduling or microservices- and this is up to date, including things like dapper and tai, so that's kind of the map up at the top and then links to all these different places where you can learn about it. So this is always cool to see the community building, these sorts of things out so wow a lot of good stuff.

A

As always, I'm always thinking like. Oh do we have any links today and then I'm frantically cutting them out because there's just too much goodness. So that is all my links. I've got today.

A

All right, so what do you got for us? Rob.

B

So I wanted to talk about you know before I get into demos and show you what we have sort of give some historical basis of how we landed up here right. So we've had this library, uh in.net called system diagnostics diagnostic source. Specifically, there was a like a type of interest called activity and activity was basically the primitive used for you know tracking, distributed tracing information, and I think back in was it 3-1?

B

We did some work where you know, if you, uh in the context of an asp.net application, we created a new activity for every incoming request and then on outbound requests, we sort of propagated this information, and you know we used known http, headers and so on the other side, if you had another asp.net application, it would understand the distributed racing information we sent over the wire. It would interpret it and ergo you could.

B

You know propagate this information from app to app and what we had done is we invented this hierarchical format and the way we did it was we looked internally. We saw what a couple of themes in azure were doing and we thought hey. This is a good way to represent it. At the same time, in the community there were a couple of projects. There was uh open, tracing and open census and they sort of merged together to form this new project called open, telemetry and open telemetry is now part of the cncf.

B

uh For folks who are not familiar, that's the you know cloud native computing foundation, which is part of the linux foundation right and what they're trying to do is they are trying to now define a vendor neutral specification for publishing and capturing observability data and when I say, observability data, there's sort of this consensus in the industry that there are three pillars of observability, so we have tracing which is distributed.

B

Tracing logging and metrics, so open telemetry is trying to grow into the role to solve all those needs, but the focus of the project- primarily I mean at least thus far- has been on distributed racing. So, while we're looking at investments, especially in the dot net six time frame to align the metric story, in.net with open telemetry today, let's focus on the distributed tracing aspect of it.

A

B

Let me go ahead and pull up this link, I'm always worried when we're on this stream, because one of the tabs is our stream. Yet can you see my screen john.

A

uh It's up there now.

B

Okay, so this is first thing. I want to point folks to there's this. uh Like I mentioned right, open telemetry aims to define a vendor neutral specification, so this repository sort of has the specification.

B

So this is a language agnostic specification and then it's really up to the open telemetry implementations in each of the language stacks to implement the specification. So what we did in net is when we spoke to customers. The thing that we found out is uh folks are wary of taking uh external dependencies when they're building libraries like it's a little different if you're an end user right, but as a library we've all you know, we know the horrors of binding redirects of bus.

A

B

Conflicting versions of transit dependencies, so we said, is hey. If this specification is stable, let's make sure the primitives that exist in dot net can capture the same fidelity of information, or at least everything that we think makes sense. So the open telemetry.

B

uh So there's a project called opentelemetry.net, which, understandably you can guess from the name, is the net implementation of the open telemetry project, and this is supported on um you, know.net four, five and above and uh dot net core, all the supported versions, of.net core! So it's you know it's pretty widely applicable and what we said- and I I won't read through the entire document, but the basic guidance is rather than directly relying on the open telemetry apis. You don't in your.

B

If you're a library, author, you don't need to take a dependency on this, you can just use. uh You know like uh types in system diagnostics, diagnostic source, which is actually part of the bcl. It also ships as a new package for, like you, know, framework, and you can create these types and now because we have the same projection of data that open telemetry spans.

B

Do when you integrate your project with open telemetry, like, if you add in their exporter apis, for example, you use open telemetry to export this data to zipkin and if that doesn't make any sense, no worries I'll cover it in a demo. Very soon you get that information without having you know, taken a dependency on open telemetry to synthesize the data.

B

So there's really, you know the the biggest call to action I have today is if you're a library author now is a seriously good time to look at adding support for open telemetry right. So let's do this. Let me hop over into visual studio and I'll show. A couple like I'll show a simple demo and then we'll go from there right.

B

So sorry, john, did you say something that just sounds.

A

B

We have this, you know like I mentioned, we have this new uh uh activity api and what we've done is we sort of made it uh as part of the work to you know, make it more compatible with open telemetry. We also made it a more ergonomic and easy to use api. So now, if you were a library author, this is sort of the pattern you'd be using to create an activity. So I have an activity source here.

B

It's a named activity source and whenever I have my operation of interest, I can use the uh activity source to start a new activity, and you know it in it handles certain things for me like it makes sure. Like you know, you obviously don't want to allocate create an activity if no one's listening, right or based on uh the sampling policy, you might, this activity might not be observed. You might only observe ten percent of all.

B

Like all uh you know, traces going through the system, so it sort of, is smart there and helps you save on things, but once you're you know in the context in in this context, and you have an activity created, you can do stuff like annotated with information. You obviously have you know, tags or set of key value pairs. Events are certain, you know operational time stamps, and then there are certain. You know.

B

Other key value pairs you can add like status, is something that's defined as part of the open. Telemetry span specification, something we don't quite have in dot net yet, but you can use um you know, I think, there's some. uh What's the what I was looking for semantic guidance on use, a well-known uh tag, I'm actually trying.

A

To look at it here, but.

B

I would say you know this would be a good link for folks to check out it's on the open, telemetry side, and this is the guidance so yeah for things that aren't necessarily supported you can set. You know, well-known tags, and things like that so anyway, now that I've created the activity, uh there are two ways to consume it right. You could create an activity listener directly and that's sort of the raw you know bcl type or what you could.

B

What you would most likely be doing is you'd, be using open, telemetry right and what I've done here is. I have configured a tracer provider and in this case I've added a console exporter.

B

So this is again, you know this is a simple demo, but you know in a more real-world application. You might add a zipkin exporter, and you know you probably add this zipkin exporter to both your front-end back-end or you know, whatever all the tiers of your service, so it can stitch together the information.

B

So if I go ahead and run this there shouldn't be anything interesting right now, but I just want to you know show this. So this is sort of information that we've annotated on the activity. There's a couple of other things like you know the start time and the duration that it creates for us. I remember talking about events like I added a you know, an event of my own and it gave me like a timestamp, so I happened to see there like a couple questions pouring in about the exporters in zipkin and bro.

A

B

Promise you folks, I will get to it.

A

Okay, so now that suspense.

B

So so I was talking about what library authors have to do right. They have to do something like this now. Obviously we would be amiss if we didn't do it in our own libraries. So what we've done is in asp.net core uh and in http client. We've already done this, for you.

A

All right, okay, so.

B

Now, if even if you say, if you're speaking from like a python application to a.net core application, you know the python application is instrumented with open telemetry and then it uses the w3c trace context. So that's a recently ratified w3c standard that tells you how to propagate this information using http, headers right.

A

Which is really cool because then, like like you're saying it's w3c standard and it's this includes front-end back-end microservices, all that stuff, so instead of poking around and saying like is the problem here. It's a problem here. You can get this kind of complete view.

B

Yeah and so that's you know, that's our goal. We've been working very closely on this effort with the uh the team that owns azure application, insights.

A

B

Know I think the the the desire for a vendor neutral specification is great and especially one that works across all languages. So it's something uh you know the thing that I'm showing you is something where application insights will get to it. You know we'll. You know we get to the point where it can be open, telemetry based, but.

A

B

Not just for dotnet it's going to achieve that for all languages.

A

Now one place that I've seen zipkin recently is when I've been going through, the project tie things and, and that has integration with zipkin as or there's a um an extension. I forget what it's called. It's.

B

A

Plug-In or extension for that, so so that's nice to see that kind of is there any kind of overlap between what they're doing with project tai and this work with open, telemetry.

B

Right, uh it's funny. You ask that question because I was doing something unrelated, but I happened to have this window open right here. uh So uh I think you know earlier talk when I was here. I came and spoke about. uh You know, event pipe and the diagnostic server.

A

So one of the things.

B

Is in addition to you know, I was talking about the raw bcla activity listener that you could use to get this information. We also have a way of egressing this information from out of the process, so you don't have to be present within the process to get to it. So I have this.net trace commander. I can try and zoom in and the provider of interest is the microsoft diagnostics diagnostic source and you can say hey. I want to listen to you know certain events.

B

uh You know I apologize because this is extremely confusing, but we had to define a dsl and.

A

B

Stick it into a string to turn on what provider so in here I've only enabled the uh start of this, uh the start provider and I get the span id. So you know ty, uh while tai inject some stuff in your application. The way it actually gets distributed. Tracing information is effectively using. This they'll subscribe to a couple, more events and a couple more providers, but they are using the activity, information and that's how and then tai populates it in its own process and then sends it to zipkin, which you can use to visualize.

A

B

So what I'm going to do is I'm actually going to show you a demo, I'm running it with dye, but that just happens to be incidental, but I'm gonna be using uh zipkin directly from my application. So I'm not really relying on the you know the the thai uh I'm not relying on type or say so.

B

Oh sorry, just one thing before I jump over, I want to say: yes, while you can use activity listener to directly consume this information, you're more likely going to be using something like open, telemetry and then open telemetry has the notion of exporters. So it's sort of the same core library very similar to like our ilogger syncs right. You can plug in different logger providers and you get the same fidelity of information going under different things. So I was showing you in this example.

B

I had a console exporter that I just ran and I could have added zipkin, but I'm going to hop over now to this asp.net application I have, and what I've done here is I've set up distributed tracing. So I had to do a couple of things: I'm in my startup reconfigure services and, I said: hey, let's add open telemetry and I told you right: uh asp.net code had already has already been instrumented with activity and so has http client. But these sort of extension methods tell you hey.

B

Let's start listening to the information created by this stuff, all right.

A

So I don't know if now's a good time, but this is an interesting question from alfred he's asking for the difference between tracing in this context uh versus that that's in microsoft, extensions logging, when using trace log level.

B

Okay, so uh so the trace log level just happens. You know the word trace happens to be overloaded and the trace log level when we're talking about logging is just one of the different levels of verbosity right. This isn't that. So, when I'm talking about distributed tracing here, I'm really talking about how do I see the causal links between relationships?

B

I I promise you will make more sense. If you bear with me.

A

B

Just another minute or so.

A

It's I mean the whole thing comes down to it's distributed right, so, in a distributed application you're trying to understand I've got a back end. I've got a data. Tier I've got an api. I've got this. I've got front end, you know api, maybe multiple front ends, and how do I understand what failed? What why did this order? Come in and not get trans or why is it slow or whatever right.

B

But uh there's also the the interesting bit, which is you might want to see like this gantt chart or waterfall of these causally related operations, but most likely once you see that you're gonna probably want to go back to your logs and I'm going to get to that in a bit too as part of the distributed tracing work, we also made it. um We made logging distributor tracing aware so I'll get to that.

B

So I have this, uh you know so the application I was showing you is this trivia game, that's been built, so there's a front end on the back end. It's a blazer front, end talking to grpc, backend and sort of. I had to do this in both my applications. I said hey, let me add, distributed tracing and in this case I say, uh let me add, you know asp.net core instrumentation, http client. I have an always-on sampler here, but there's sort of you know different sampling strategies.

B

Like I mentioned, you can only sample some of the stuff going in and then I sort of just got my connection string and I'm sending my data here so now for the fun bit, which is, uh let me pull over a tab? Okay, so this is the application I've written. So let's go to this bit where I play trivia, uh you know I'm gonna play as john.

A

B

Score doesn't count now what this did is. Actually it started a streaming, uh grpc call, so the service actually streaming me questions. If I go ahead and like say answer this question, the server streams me another question right away, nice and let me go and answer another question. It was a simple game, two questions. You did really well, john. You scored a hundred percent of the quiz and.

A

B

I can go ahead and play again, but I'm going to go ahead and close this for now, and so you know, I I told you all these things there's a front-end back-end which, if effectively there are a bunch of operations that are causally related right and I would love to see that relationship. So how like? How would I do that so because I've been sending this information to zipkin and now, hopefully a lot of the folks have frustrated by not getting the zip code up until now will be happy.

B

I can go in here, so it's up and running locally and you see.

A

It is really cool like when I first that was my first um look at. It was going through the blazer tutorial, or I mean the thai.

B

A

And seeing this visualization is beautiful, like it's it's I mean it's useful too, but it looks cool.

B

Absolutely, and so you know, I was telling you all these things like this was the beginning of the blazer circuit, and you know I did this start trivia. It was a grpc call. You can see this. This information came from jpc. uh I even have the open, telemetry status code reported by the grpc library, because it's aware, and then this is my you know, I said there was a streaming call, so you can see. There's this 17.

B

Second long call excuse me, and then you see uh at the end, at the completion of that I sent another call to get my score and you can see uh you know this is I it's sort of these were small, so you might have missed it first, but.

A

B

Is like a time series view of this right and this stuff is super cool, but I'm now going to jump back over to my cord and show you something else so now bear with me for a couple of seconds: go I'm going to take a small detour and go back and talk about logging right. So let's do this.

B

A

So the tr, the tracing at that level, wasn't showing specific log stuff. It's showing the timing. It's it's a time right.

B

So it has the yes, so there are like certain, like you know, properties associated with it. One is like start stop. You can add like tags to it, which is basically like uh key value pairs of information. You can also have activity links like you know. I showed you a simple thing where one thing called another, that's called another. Sometimes you can have like batched operations.

B

You can you know fire 10 off together, so you can have links that say hey. This is my peer rather than this is my parent. So all of that could be you know, visualized using that sort of think of you know the um you know in chrome. You have that thing. Where you see the order of http operations, uh it was where's that gone.

A

Like the network activity.

B

Exactly uh this one: okay, can you see this waterfall? It's something like you know, it's a very similar visualization too, but so I'm taking a small detour to talk about logging and it will make sense in a second right.

A

B

Want to remind folks so this is simple. You know I'm using logging in a console application and I've done this thing called you know. uh Logger guard begin scope and you know: we've had we use uh an async local to uh persist, scope information. So even if you know you go everything to perform other operation, this remains and what this effectively does is says: hey all these log messages that happen inside this logging scope. Please annotate this information right so.

A

B

And run this sample, you know that took longer than I thought. Okay, so you know these are the logs that I expect, like hey log information, so you see info, you see the stuff, but then you see this information annotated for each of these logs because we're running inside the context of logging scope right and so this because this mechanism exists in logging, we said hey, why don't we make the log smarter and we make it activity aware?

B

So I'm going to hop into this example now and you might have deduced what I'm getting at, which is, if we have this way to just you know, stick arbitrary baggage on logs.

B

Why don't we stick what activity you're running under right now, so we added a couple of options on the logger, so you say now when I'm creating, you know configuring logging, I say activity tracking options and I said: hey, you know annotate each of my logs with the span id trace, id and parent id okay, uh so trace id is you know for the current operation parent is? Who invoked me and span? Is that logical identifier? That goes? You know across the entire trace.

A

B

Right and once I did this, uh uh the other thing I have to do is I have to actually be running in the context of an activity for this to you know, yield anything useful, but in this case I am so I created an activity so now I'm running a sentence inside the context of an activity. So now, if I go ahead and run this, uh you can see so I still have the same. You know scope information from last time right this one, but also you see this new stuff added.

B

I have my span id trace id and parent id now you know we would be amiss if we didn't put two and two together and realize we could offer you something that works right out of the box.

A

B

What I did uh so you don't have to do this by the way if you're running an asp.net core. This stuff happens for you directly, but if you're running in uh you know in another context or newing up your logo and you want to be distributed tracing and where you might have to do this. So if I go back over to this uh tab right, so you see we were looking at this.

B

You know operation, so you see there's this trace id associated with this operation and I'm going to go ahead and copy this, and now I'm going to go to where my logs live right now, uh and so you know, I just happen to be using diet to run it, but whatever your eventual log store is, and I'm just going gonna do a naive. You know search and page and stick in that race, id and boom, ah and this is sort of the workflow you would be using right. You'd be like.

A

B

I go look at this, uh you know distributed race and something seems amiss. Let me use this as the way to key into my logs and identify what went wrong and all of this you know just works. You got it. You don't have to do like anything for the logs to be activity. Aware.

A

Okay, okay, just by setting those things, then it's it's! It's added into your logs as well right.

B

A

So there's a question here and I I've seen zipkin, I haven't seen jagger so there's a few people asking how this compares to jagger so.

B

I personally have not used jager. Okay, uh I I mean I I wouldn't be the you know best person to comment. However, I do want to point out that uh it's all you need is a different exporter for open telemetry, like the.

A

Code, you write.

B

Should not change and that's the goal of this project.

A

So you just write to the standard, the open, telemetry standard and then any thing in that.

B

Ecosystem can go back in here, you know, so there are a couple of things when you set up open telemetry, you say what all instrumentation types do. I want right.

A

B

Case I said I want asp.net code, I want http client, uh there was the other one in which I think I said hey. I want grpc as well right. Then you say what is the sampling, so how many of these samples do you want to consume and there are different philosophies you can have and then the last bit is you add the exporters you're like? Where do you want to send this information.

A

B

Right- and you know I added zipkin- I believe, there's a jaeger exporter and if that isn't, I trust you know, there'll be one very very soon uh and you know we're working we're looking at adding an application insights uh exporter, I know a couple of the other apm vendors uh have it. I think you know I I don't remember, which ones I think maybe was it splunk and I want to say: datadog have exporters, and this is sort of you know the move across the industry. Everyone's moving to this.

A

Okay, here's a a relatively big question here, I'm not sure. Okay uh do! Do you see this on the screen here.

B

Yeah so they say, themes historically have used the guide.

A

And that the link that star star star is asp.net, that's it starts out links.

B

Oh, I see got it.

A

Yeah, so they want to map incoming guide header as a parent trace id.

B

So I don't know that uh the asp.net layer is pluggable enough. However, you could sort of, I I think I I don't think you can change the asp.net behavior. However, what you can do is you can create a new activity like new up your own activity context and you.

A

Can instantiate.

B

That activity context uh using that incoming header.

A

Okay, so you'd be able to map it to that.

B

uh The the format, though, like the good, is not compatible with the open telemetry specification right like oh, so you know.

A

What would be better is to yeah what would be better is to instead, instead of doing your own custom good and a header to actually just use the w3c like use the standard stuff right, all right.

B

If, if so, I I'm not sure how they're using the good, if the good is just like to correlate their logs, they could very much just add tags like you know, in my example, oh shoot where'd. This go give me one second I'll hop back over here. You see. I added this tag with the server name.

A

uh Right so you can you.

B

A

Add a tag, so it's.

B

Based on how you're using it, but the expectation is, the solution should be complete and you know it should be usable.

A

Okay, so I could put multiple tags based on whatever I mean: okay, cool, all right.

B

And so so I can you know the the one thing I want to talk to is about uh the the project and what the maturity level is.

A

B

Right now the opentelemetry.net implementation is in beta. The idea is as soon as dotnet 5 goes ga or soon after that we want the the open, telemetry project will go. Ga again, you know that's a recommendation. I can't speak for it because it's a independent project with its own governance yeah, uh but I mean.

A

Ga means general availability, which.

B

A

Basically released yes,.

B

It'll be a stable version which is released. The good thing is library: authors can, you know, start working on um instrumenting their applications because they don't need to rely on wait for the open, telemetry stuff- oh my god, so my fire alarm is being tested.

A

Oh good: well, if it's a real fire, then uh I'll just.

A

B

Unfortunately, john, I think we have to.

A

Okay, well, that was a great demo. We'll share the links, be safe, all right, okay, I guess we'll end there then!

A

Well thanks a bunch syrah. This was uh this, was a fun show yeah, I told you suspense. I did promise suspense for the show, so um the community links I shared them earlier. um I added in the link to the opentelemetry.net.

A

So, as he said, it's in um it's in ga or it's in beta now will ga at or shortly after um.net 5 releases and that's expected for dotnetconf.netconf is coming up and that's uh.netconf is november 10th through 12th, and uh so a lot of cool stuff happening then and and including.net5 so very much looking forward to that, and uh I guess we'll wrap up there for those that are watching on twitch, we'll go raid, somebody and uh all right the. I think that that breaks some records for suspense in the show.

A

So um we'll let you know when, uh when we hear from sarah that he's okay, all right thanks.

B

B

B