Jakarta EE JakartaOne Livestream 2021, 14 Dec 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Helidon – The coolest new features to skyrocket your microservices! | JakartaOne Livestream 2021

Description

Helidon is a brave small yet powerful set of java libraries to create great microservices with the full support of MicroProfile. In this talk you will learn about the newest features we have crafted to bring your microservices to another level liке MicroProfile LRA and more!

—
Jakarta EE gives developers a comprehensive set of vendor-neutral, open specifications that are used for developing modern, cloud native Java applications from the ground up.

With Jakarta EE, technology developers and consumers can be confident they have the best technologies for developing cloud native, mission-critical applications. And they can build on decades of Java developer expertise to move existing workloads to the cloud.

Learn more: https://jakarta.ee/

Follow us on Twitter: https://twitter.com/home
Follow us on LinkedIn: https://www.linkedin.com/showcase/jakartaee/
Like us on Facebook: https://www.facebook.com/JakartaEE/

A

Okay, it looks like uh I am live right now, so uh hello, everybody. I know that uh there are more than 800 of you now there uh everywhere around the world, so this is really cool. I think this is one of the biggest online events. I've ever been so far, so I'm so much excited and tanya and eva. So you've got an amazing like studio right now, so so it's so cool to see you there, but it will be really cool to see you offline again and meet you in person.

A

Finally, uh because this pandemic is really getting like, ah I don't like it, but for now we are online and I'm really really happy to be here and I'm really really excited to tell you about the newest and coolest features describe rockets and microservices.

A

So um well. Let us actually begin. uh Okay, so stay hard. So whenever you make business decisions like check, if everything is the same again um just because uh we are evolving very fast and things may change okay, so my name is dimitry alexandrov, I'm a software developer at oracle. uh So uh this is my contact. This is my tweet handle. So whenever you have questions just tweet me or write me, I will do my best to uh to answer you as soon as possible.

A

uh He says just don't hesitate so about coolest news features, we're gonna talk and the actually the tool uh uh to write this microservices and the tool to demonstrate you uh all these cool new features, I'll be using, is actually uh the two I'm working on and this callisto is called halidon.

A

So um I'm quite happy about being this part of this awesome team, uh so just a few words about the tools and then we'll actually jump into some some some interesting code.

A

So um what is helidom a set of java libraries for developing microservices, uh not a servlet container, but it's actually a tool to create a mock services uh based on neti. Actually, that's why it is so fast and asynchronous and so cool um for those of you who don't know how they don't mean swallowing greek- and this is one of the best names I believe for for a microservice framework because, like you know all microservices, they are so small maneuverably fast and as they fly and clouds and blocks.

A

So uh it really reminds us of this bird. Our mascot is frank. You can see here so and, as we say, cloud native ready. That means that we are not just serving microservices to you via wrestling points, we're doing a lot more than that. uh We also do have metrics tracing everything uh to make you know, life uh of microservices, really traceable, predictable and uh and monitorable. I'm not sure if I have just invented the verb, and um actually this will be some of the focuses of today's talk.

A

So I can say that we are like uh almost hundred percent problem native image, uh supporting that means that everything you have here uh like like ninety nine percent of it will compile to gravity and native image, so it become even more fast whenever you write your microservices, so we are really tiny. So that means that uh you have really low memory, footprint and disk usage and by the way here you see three graphs. So you know it's like helidom jvm, heavily, dom j link, jvm and native image.

A

That means that, as we are uh java 11 uh from the very beginning, we are living in a modular world, and that means that uh you can compile your own custom jvm with uh with uh with helidone, and it will be like smaller, faster and uh work better. Actually, so that's why we always have three graphs, and you see this custom. Jvm is really fast because you don't run the code, you don't need.

A

Okay, uh uh as you say so uh I mean uh helen architecture is like in this picture. We have netting on the on the on the foundation, and then we have some dropping reactive services to actually implement all this.

A

uh This health, traces checks, metrics and all the uh everything you need with the clouds, and then we have a micro profile, and actually I will show you all my examples here, since what actually held it on my profile is probably all of you know what my profile is at least the the people here in this call should probably know what my profile is a set of open specifications for developing microservices.

A

uh I think this is the current state, but uh you know 5-0 is like approaching and there will be separate talks about them uh in this in this conference and this um this online conference, so just join them to know more about it. I will only say that halidon is actually supporting microprofile and actually implements this micro profile, which actually means that um we are standard and we care for standard stuff.

A

As you see everything, what jakarta is it's all about standards and we really care about it, and we really do a lot of efforts to uh make it really standard and portable as as a result, and as you see here, so um we are not only supporting microprofile specification but out of the box. Actually, we also support uh jakarta some of these jakarta specifications so whatever, whatever you actually work with helidom um helidon mp out of the box.

A

So if you got this bundle, um so you only have micro profile stuff and you have some some jakarta stuff so as well as persistence transactions, annotations persistence, so the previous guest was saying lucas who actually works on persistence. So I'm uh a colleague I'm working on that. So we are really close and we are really having um well very very so.

A

We will work like very close in this thing, uh so you have not only my profile uh standard specification, as I would call it like this, uh but also you have side projects which are naughty naughty advantage: micro profile, crop ql and some stock, like course, rpc, which are still like uh under under under developer. As far as I know, uh please correct me, and uh I can say that why would you ever like to choose hello because uh um we are like?

A

Well, I'm not sure if there are other uh like non-servlet container um frameworks which provide full cdi support, especially full support. That means that we have a weld underneath and uh that's why our extension mechanism is based on cgi, so we are portable in everything, even even in extensions. We are portable and I'm really proud of this. Okay, let us actually jump to the cool features that I'm gonna try to demonstrate, but today so a lot of them.

A

The demos worked like before that, and uh I really hope that it will work right now so uh uh fine-tuning. I would say, uh first of all so, as I say, we want to make cool and powerful and we will focus on on some of the cool features, we'll really focus on on performance and that's why so uh metrics is an essential part of everything about unique performance.

A

You really need to know how your services perform and that's why we implement this um microprofile metrics, so uh but more than that, so we're not only just getting the standard metrics. As you probably know, metrics can be three kinds like in a vendor application and uh and base metrics, but uh uh in our vendor metrics. Now we uh actually now.

A

As I say, it's like in a release which will come like uh in the new miner version, we'll provide all the information you need for all the executors and threat fools you have, so you will basically know about executors how they perform. What are the task counts? What are the pool sizes and what are the queues and depending on which executor you actually use? um It will give you relevant information. So that means that uh out of the box, you will be able to meter all the insights of how those characteristics work.

A

So uh this is a little bit going to the to the future, but in real life to you, so you to monitor your performance, uh so we provide so-called um so-called key performance indicators.

A

So that means so whenever you, if you take a look at this picture, so um whenever you have this uh quest coming to to helidom, so uh this is the initial part of quest received.

A

We will call everything until we see a request is common. The response is sent as a request handling, so there may be a little bit with some delay a synchronous request processing. So whenever a request is received at some point, the work starts and is work complete. uh The response is sent. So that means that, for every request you have a counter and a meter meter means that it measures everything that one five and fifteen minutes for all the requests how they perform.

A

But uh this request can be both synchronous and high synchronous, which is which is uh which is kind of essential one here. So that means that helium does support asynchronous requests.

A

That means that, for this standard, kpi metrics, you will have all the information that you need for how they perform, but to make it even better for you to find in your microservices, we have created a so-called extended set of kpi metrics, uh which include in flight requests. So there's a concord gauge, a gauge uh which cares about how many in-flight requests are, how many requests are now being processed.

A

So uh we know that requests can be like long running. uh That means that one request can take more than like seconds. uh There will be a separate meeting for you uh for all the requests which are longer like 10 seconds by default, but we'll get it uh current load meter. So how many requests are completed at this point uh and maybe some deferred requests?

A

That means that how many requests are will be before will be put like ahead in time so and what is good about it, that in metrics you can actually uh totally control what you need. So, first of all, so you can enable this key performance metric extended. So that means that you have like you will get them uh in your response.

A

um You can control these thresholds uh about how many long running actions that you have there and more than that, you can control everything, even the scope, so what metrics will be monitored and in which scope and which will be reported, you can totally disable enable them, and you can even include and exclude specific metrics, uh which means that uh not everything has to be like monitor, because when you monitor every small microservices there are a bunch of data that probably you may not need, and it will just be like uh losing so just delivering.

A

This data will be like uh wasting your time and, of course, uh you can even disable enable, as I said, but you can exclude and include- and here the idea is that uh everything which excludes it excluding overrides, so they include stuff.

A

So I would before jumping there. I really wanted to show you this thing, so we have our idea and we have a project which is like a typical, quick start project, so um I will just open it and make it actually run.

A

For that. I will use our special special tool, which is helidom, and I will put it because everything everything initial initializing application like building the application and actually working with the application with uh developing your application. It really helps you very much as, for example, for this project. As we start uh that loop, it will actually build everything for us and start it and monitor for every changes that we have.

A

So uh what does that mean? So um so there's a very simple greeting application. This is our our our main application and for demo purposes I would say uh that means that it just sends creating, but we want to monitor how it happens. So we want to leave all the metrics that we have so uh it usually starts much faster because now in the zoom um in zoom uh it consumes a lot of resources, but it's usually fractions of a second, and here you see mp, we're starting mp project.

A

That means you have full cdi and full profile support here, okey dokey. So as we go to the call- and uh we do, for example greeting it will send us hello, world, hello, jakarta.

A

It will say hello jakarta, but if you wanted to know more about how they perform the services, uh we will just uh go to the unsafe metrics. This is automatically uh uh and pull metrics. Yes, endpoint that will come automatically to you and being prometheus, for that would be, are actually not much um reading it. So that's why you will just go and say header.

A

uh Second, I just say header which accepts application json, and um then you have a better option getting more better with and what we have here. So uh that's what does that actually mean? So, as you see, we have some thread information which is very specific. We have stretch request, so we have made five requests and there was a meter for that.

A

So this is really good, but, for example, if we want to make more uh more like uh focus more on how the services perform, I have my helper files or that uh so I will just go and do I guess, additional additional properties in the profile configs as we're standard, we will just add them and just if we switch the purpose, we see that uh the helidon death actually rebuilds everything for us and we want it.

A

And now, if we go to this vendor specific information, we will have more more information about how many in-flight requests, how many deferred requests. We have, uh how many long running requests are we having right now. So that means that um this information is extended to us and we can control it. So we set the default to 2000 milliseconds, so everything which runs more than two seconds will be will be like uh interpreted as something and then um so.

A

For example, if we want to create control some other metrics, so we will go to and say count it so to our default message. We will create something to name.

A

Yes, yes, I would say absolute true. uh The same will be done for personalized personalized metrics personalized codes. Yes, so get message for personalized subject, and here I will also give times and we will get like name gets personalized. So whenever a personalized metric, it will monitor off all the information person.

A

Sorry, I wanted all the information that will come to a standardized message. So once again we switch focus and this is happening live so it's rebuilding. It's informatic, it's starting everything for us. So next time we get more metrics.

A

So, uh yes, it will also.

A

uh Yes, as you see, there's gets a point, get personalized and it will get more information about how many calls are being made and whenever we make some more calls, it will uh here, but.

A

So it will monitor everything for us, but once again, if we want to uh to control what we have, so let's go back to this helper file. I would say- and we will say that everything as a type as a return type will be, uh will be monitored here. So we said only for application and we will exclude everything is with get personalized uh as pattern.

A

So let's rebuild it and then you see the development process is really really fast here and whenever so, we will do some requests. uh We'll do some some typical requests like hello world. So what does that mean? Is that whenever we get the metrics, uh so it has not been reloaded. So let's give a second.

A

uh Yes, whenever we do this, and we do this only for applications code, we can just go.

A

ah Well, it should reload and uh whenever it gets reloaded uh it should actually demonstrate everything with gets. So let me fix that personal line, personalized I've said it. Actually. Frankly, I get personalized. Now, though, now I see why it's not worth because it had to be gets personalized.

A

So we just reloaded, you see, it happens really fast, so I just switched the focus and we got it working. So that's why you see we control it and we can totally disable everything uh which means that we're doing that.

A

We will switch to focus.

A

uh Yes, it's been reloaded and now, if we go to like metrics.

A

It should still need some time to reload.

A

ah It should sometimes reload, but we will get a message here like 404, that it's okay, uh it's enabled true, it's false man. Oh sh! Yes, sorry about that. I really wanted to deliver a lot of information to you and I have only like, like 20 more minutes to do it. Yes, and uh that means that whenever we do this, we not only get like initial. Let's say it's four and four, and we will just say: metrics is disabled. That means that metrics will not be gathered.

A

You may say why would you do this and uh and properly so just not include a specific specific.

A

Specific uh dependency just how to build it, but, for example, you have different configurations for different services. So that's why you can just enable it for one service and disable it for any other service. So it's really really fast. I would call it yes. So, let's move to another thing, another cool feature is actually axemployer support, so what is what is axon par and how we do support actually exemplar is something that brings uh two things from two worlds. Like you know, it's bringing metrics and tracing together.

A

So uh what does that mean? So whenever we have tracing output, uh it always had like.

A

Sampling and if you uh go metrics output- yes, it gets something. So it gets samples of of of the data which is being collected and if you get trace, you can trace every request and what the x employer do. Actually, let's go first to the definition of exemplar which, by the way, merriam-webster dictionary is like as an idea or a typical or standard specimen.

A

It actually uh corresponds to special cell well sample to a most close tracing. So what does that mean? So whenever you gather some metrics, it will deliver you information about how do they correlate to metrics, because sometimes it's like they're living in two or different worlds, and now uh with exemplar support?

A

It's very easy for you to gather how to actually current metric samples correlate to specific tracing. So you can really find what's what's the problem, for example, in your performance tests and uh let's actually jump to this uh to another example.

A

So, um first of all, I will just need to run the zipkin uh server, so I will just go to uh and I will have my zipkin so I'll just run the circuit on my local machine, wherever you can do it on on your other uh stuff or like other docker images or whatever it is um so, and let's jump to the next example where we actually do this, and this example is like examples for example. So what is the main difference between these two files?

A

It is actually that you add one dependency for tracing and, for example, support, and this is it. This is everything you need the examples to be running so and now, if we we actually run it.

A

So I will use I already pre-compiled. I guess version of it so uh targets to save your time.

A

So um usually starts much faster once again because of the generality of the um now. If we do curl with with matrix, you will start saying that you will not only get like type health information, but you will also get where is it uh get some additional information about? What is the yes? What is the trace id here? It is so that means that whenever we do a probe here, it finds the most suitable trace id and actually shows us. What is the trace id?

A

How did how did this metric correlate to this trace id and now, if we go to uh to to zipkin, so I will just open it.

A

A

Just copy it yep here it is and when the query, that means that we can actually see how this current response a request response, actually uh get our information. So that means that whenever you have a metrics, it will find the corresponding trace id and correlate it to the metrics that we do have. So I believe this is. This is really cool for fine-tuning your services, and this helps a lot for you to improving performance with services that you you actually have.

A

So, let's just stop everything. Once again, it's very easy. You just get exit power, support and that's it so you just everything else is the same as the code, so this is the same application. You just add this uh this um this year, this this two dependencies and that's technically it.

A

um So let us jump to the to the slideshow so uh once again, zipkin or jager by the way is supported as well, and that means that whenever you get a standard, prometheus output, you will also get a trace id which actually corresponds to uh to the uh trends that you have. So I believe this is really cool for fine-tuning your services.

A

Another cool feature is the micrometer support, so micrometer is very popular, and actually I know that uh for as far as I know, please correct me if I'm wrong in the chat is that uh sorry, I don't see the chat right now, so um maybe you've got questions I'll, be happy to answer them in the in the end.

A

uh So micrometer is like an instrumentation for sate, so it's it's quite widely used around around many many vendors and I learned that there should be a something like um like joining this two different worlds of monitoring in uh metrics in microprofile and micrometer together. So basically, what do you have? So this is also just an annotation um just a typical annotation.

A

Which is this way is just timed uh or counted. It has a little bit different uh uh different uh set of.

A

Parameters, uh but it actually delivers you as again a slightly different output, but before actually jumping there. Let me let me go back to the code and demonstrate a little bit, um so I have prepared a an example for you.

A

So here it is uh yes, it's here we will just go here so once again, this is the same greeting application that might be used and I will just go and say target because I pre-compiled it for you.

A

I guess it should be micrometer mp.

A

It's not good. um No, it should be like wait. A second uh okay, but let me just have it down there and we don't death this stuff make it run so.

A

Wait a second integrations example. Oh yes, because, oh sorry,.

A

Yes, now it should work.

A

I just did it the wrong the wrong the wrong the wrong way, so it just puts a little bit different annotations, like you see this timed here, but uh this time there's not the mic profile one, but it's like uh higher micrometer and it's set as experimental feature to do this, and the output will be slightly different from slightly different endpoint. So we will do not matrix but micrometer meter and it will just deliver us close to the same output, but it might be needed for that.

A

So um whenever you have like this two different way to monitor about microservices, uh exactly you can choose between micrometer or or just a standard micro profile way of doing stuff. So uh it is here, it's available just put a different another another uh dependency, and that's probably it so it's so much easier.

A

Okay, okay, come here, feature number two. So now I'm a little bit jumping to something a little bit different, but it's also essentially part for for your microservices uh from very tiny node performance. You will run into something which is called wrong running actions, uh so my profile already is actually one of the implementation of saga pattern for distributed transactions, so um it does relax some of the constraints for assets, so everything for distributed transaction.

A

While it's it's very hard to make an asset, I'm not sure if it's not possible, but it's a computer science question and the exactly second part is one of the one of the patterns is quite old, but uh the good part of it. Now we have a micro profile, every 1.0 spec already available to us, and that means that we are available. We are now ready to work with uh really distributed transactions across different microservices and, let's see how we do it.

A

So uh actually there are two main components involved in hl array, so these are participants. That means that uh usually these are juxtapress microservices that will be involved in one transaction. uh They are annotated in a standard way following the uh standard uh eclipse reallocations uh and there will be a coordinator. So there's like a separate service running out there and it actually coordinates all the activities and decides. Okay. Are you introduction now?

A

Are you out of transaction now or, for example, if something failed, we have to do a compensation action for that and uh that's why huh uh it actually cares for that and that's why so the coordinator to know about it.

A

uh We have this set of annotations, so that means that we annotate something with lra um well like microservices, with already means that it will be compounded. Part of of uh of this long running action and there are stupid annotations with different methods for compensation or leaving completing.

A

So it's really really self-describing very, very much so and let us actually see how it works. So, let's go to our code, it's so much nice that, uh let's see it's very easy to start and work with everything I will close uh the stuff. We don't need- and I will do some lra example, so this by the way, are just uh typical examples for for uh on our health repo, so you will be able just to play with them uh and for coordinator.

A

So in this demo I will use our own implementation of an array coordinator, although we are compatible with nariana, um so you can just run it in a docker image somewhere, but I want to keep it really simple. So that's why we will run our own lra call it. It's still like experimental features like with the disclaimer.

A

It does everything, but uh it's like work in progress, but I'm really happy that it's so strong. So what does that mean? Small uh small means that we only have one method which actually uh runs, and um uh you just start it and it works. uh At least. We need some time to start once again. uh Zoom is consuming a lot of resources and uh we will jump into our lra example. So let me just open it. So let's say uh where is it? uh Yes, it's coordinator uh integrations.

A

Oh my god. Where are you yes, microprofile hillary? By the way, the coordinator started it's freedom with at 70. how it works, and now we'll do some really really small resources which have like two main methods. To start an example, and as we say, we are part of lra, so we just declare it as a lra.

A

We have a complete method. The yellow array, when, when something is likely distributed, transaction is completed. So this method annotated that completed will be called so we'll complete the example and whenever something goes wrong there will be a compensate action.

A

So uh let me let me jump into it and make it uh run, so I will go to lra valerie. um I will just so I pre-compiled it. So I just uh sorry I'll just do my homework as I say, and uh we'll just run it uh it's once again. It's hit the application to start really really quick, and uh I won't like you know you see you can write it as a docker image.

A

For example, if you have like kelly the only real cornette or narayana, but uh and uh and uh to make it simple, we will just um um make it the simplest way. So we have started and then let us actually start an ally transaction. So we did start it and as we see we completed the action, so the complete method was called. So you see, every lra has its own id and most probably we will get it as an output in here.

A

So yes, this is the id, so we have a an id which is completed and now, let's let us do some exception- uh to force uh compensation action to to uh be called, and you know so. This is boom unexpected exception and the action is compensated, and you see the coordinator said that hillary id was compensated. Then we go to the lra itself. It will just say that okay, we compensated this id, so that means it happens like a distributed transaction and you are able to nicely work with it.

A

As the specification is ready now um we are really free to use it. Once again, you see our our own. Implementation is really really tiny, so you can just run it from the first case. Just the idea, uh let us jump a little bit back and uh let us see what we have more so just just to mention a few minutes. So, uh as you say, we have this. um If you remember this architecture, this micro profile is based on reactive apis and instead of hiding, we just published these apis.

A

They provide the same actually set of specifications, so provide you the same functionality, but it's pure java. So um no cdi here involved. So uh we haven't have additional stuff, like a reactive web, defined around 2db client.

A

So uh it's not much focus of jakarta community, but whatever you need really high performance services and you count every cycle. We still provide you this possibility and, as you see, uh the the memory footprint is really tiny and the applications become really small, but then you can of the standards. But then you get all the power of java in the world uh by the way uh before that, rudy did a wonderful talk about some new stuff inside java 17, um just whenever you're interesting.

A

So together with david cabela's uh from from verizon team in oracle, we did a presentation when we also demonstrate this java 17 features uh in helidon, so helidom perfectly works with java 17 and you can enjoy uh all this uh benefits. Like sealed classes like uh like uh switch expressions, uh et cetera, et cetera, you may just follow this link.

A

Once again, we are growling and native image friendly. So what does that mean that everything you have here is gravity and native image?.

A

Compatible, so what does that mean uh whenever I did a same quick start example, but I pre-compiled it to native image. So now you get all the same stuff, uh just an executable which starts like so much fast and if we go to like do some curls oops curls, uh like like great.

A

You will get absolutely the same stuff uh like like here. So no limitations. Oh hell, I don't know it's uh metric metrics.

A

It does have something the same, but now this is a single executable about any device, but still with full cdi working. So we are extremely proud. We have this cdi uh in native image compatibility, and this makes this quite a unique, so uh you may also play with it. uh You probably saw the slide. I showed you so already about project loom uh no time for this at this time, but we already do support project club out of the box.

A

That means that uh whenever you type this this this small item make it small, uh like small property in your properties file. You will run absolutely the same stuff on virtual threads, which means that um which means that you will get that ultimate performance of loom. So only only only one one small property will make it run all your requests run on virtual threats, not the usual one, and that will I believe that will be a separate talk quite soon about it. uh For those who lost confidence, we do support the latest conference.

A

Boxing can play with the same kind of stuff uh with the same apis, but all in content. So there's a bunch of examples that we did there, and so, if you want to play test containers so hello, those are so much suitable. Do this there's a quite a big article describing how you can test your application uh microservices in test containers, including kafka, so you can run as a black box.

A

Your application is a black box inside of test enters or you can test your application against services running in and test containers like, for example, you can test her up in against kafka or other database.

A

So it's quite easy, and with this I just want to say thank you so so much um I'll have two minutes for questions and uh I'll be really grateful. So whatever you just, if you could ask me on this twitter or just mail me, so uh thank you so so much for your intention.

A

I think, by the way, open, open questions.

A

A

A

Questions now: okay, uh no questions. That means that. uh Thank you so much uh thank everybody for your attention and um I really have enjoyed, and I really now you have to find tunes fighting your your micro services. Actually, this cool new features are there to help you so uh thank you so much and have a wonderful, jakarta one and I really pray to be part of it.