Jakarta EE Tech Talks, 9 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Jakarta Tech Talk - Building cloud native microservices with MicroProfile and Jakarta EE

Description

Show how to build a microservices with Microprofile by adding cloud native features from Microprofile. I believe a hands-on live coding session with incremently adding features will show how easy it is to use Microprofile to build production grade microservices.

A

A

Hello, everyone and welcome to another Jakarta Tech talk. My name is Serena and joining us today is Ron Veen, who will be presenting on the topic of building Cloud native microservices with micro profile and Jakarta ee. If you have any questions for Ron as we move through, today's presentation feel free to ask it in the chat or use the ask a questions tab without any further delay. Ron over to you.

B

Thank you. Yes welcome everyone and great that you actually want to join me on this introduction to Jakarta, EE and micro profile, um I'm, just going to start with a short presentation to give you an idea what we're going to be talking about, and then we'll do some Hands-On coding. So um that would be the agenda. A sure thing about me: I'm doing something like 20, plus Java development on and off I'm a developer and a special agent at a company called team rockstars.

B

It I do like speak at conferences and currently I'm writing a book on how to migrate, Jakarta ee to a newer version right. So what is Jakarta ee I guess? We should first look at the two technologies. Well, it focuses on developing large-scale Enterprise applications with the Java language. Of course, it's a set of apis really, together with a reference implementation and a technical compatibility kit to prove that they actually meet the demands of the apis. It was originally started as j2ee and later j-e-e by Sun Microsystems.

B

um It was then renamed again to Java ee by Oracle when they bought Sun, um but in 2017 the eclipse Foundation became the steward and they rebranded it to Jakarta EE.

B

um There have been three releases since Java ee was the transition release from Oracle to um Eclipse jata Jakarta ee9 had everything moved from the Java X namespace in packages to Jakarta and finally, just recently released Jakarta ee10, with new functionality um being added to the project and already talks are being held now for what to include into Jakarta EE 11..

B

Looking at the platform, it consists basically of three profiles: there's a new core profile, which is new in Jakarta ee10, which is really targeted at um micro, surface style applications, Edge Computing things like that, where you just need a really a subset of the functionality, which is really small and concise.

B

There's the web profile, which includes more functionality and finally, there's the complete Jakarta ee-10 platform, which includes all of these specifications.

B

There's a number of vendors who are actually working on my on on on on Jakarta ee or implementing it most, notably, of course IBM Oracle, pajara tummy tribe, red hat, which have their different implementations.

B

Then, moving on to micro profile, so microprofile basically is another set of applications that bring to Jakarta ee or formerly Java IE. What wasn't there yet, which is the cloud nativeness um Java ee, had its releases, but they were very infrequent and with long intervals, which was basically a no-go in the time when we had all these developments going on in with web related Technologies and things moved so fast, then, and then you had to wait like two three four years before you could get new things into a Java ee release.

B

So what basically happened is that all the vendors from the different application service they started to implementing their own set of apis, but if you think about it, that goes against everything. What the whole Jee ID stands for. That is really to to have a standard set of apis, and now everything everyone was building their own set.

B

So a group of Hendersons of individuals started micro profile in 2016 and they said well we're going to make some specifications and we, you know we're going to use some default Jakarta ee specifications and on top of that, we're gonna build our own set of specifications so no longer every vendor having their own specifications, but we're going to make another standard on top of this one, um and so they came up with specifications for resilience, configuration metrics, distributed tracing Etc. We will go through a number of them Hands-On.

B

B

um It was very interesting to see that the first thing they address with microprofile was having a really short release. Cycles in the beginning, they were extremely short where they just said: okay, we we hammering out this API, we kind of put it there in the market. We're gonna see if it gets adduction. If it gets adapted great, we'll continue it. If it doesn't we'll see if we need to tweak it so um that went really really quick um to indicate started, then in 2016 and now we're already at release. 6.0.

B

So what are the specifications and the apis in micro profile? Oh, as you can see here, um there's a default set that everyone who have wants to be micro profile compatible, has to implement and there's a set of Standalone apis like the long running actions API or the graphql API, no necessarily needed to be implemented, but they're still there and and there's still this demand of having some standards there as well.

B

Again, there's a lot of people, but also a lot of the same people and companies working both on Jakarta, EE and microprofile. So again we see IBM Red, Hat, Tommy tribe. We see Microsoft having a great part there, pajara, Oracle and so on.

B

um That was the part where we actually talked about all the slides and now we're gonna switch to the code. Editor and we're just gonna have first have a look at Jakarta, EE and microservices, specifically within Jakarta ee um I. Have this tiny program started up here and the most interesting part is even though we are getting in all these specifications, we basically only need two dependencies.

B

I need a Jakarta ee, API dependency and I need the org Eclipse micro profile dependency, both state that they are provided, because the underlying application server were actually implement it. So I just need the apis here to get my code compiled.

B

um So that is our application, starting at the basics of the application and also in our jobs, arrest, Works, Jox RS is the restful API, which is part of both Jakarta IE and also leveraged in micro profile. You can see that um we're basically telling here. Okay, you are an application, and this is a class defined within the Jakarta framework, and we tell it with the application path. This is where your endpoint is and where clients can reach you. So this is the application part.

B

Next up would be that you would have a project a resource with a lot of annotations on there, so it tells here with the path annotation, okay, after the slash API, if you use the slash projects, this is where you can reach me. Then it tells that every input it takes like from the body content of calls, it should be application, Json and everything it produces itself- will also be Json format.

B

um Those are the jobs um are X implementations. Then, if we move on to, for instance, this method, this is something where it tells us. Okay, I want to retrieve via HTTP, get response, some data from this path, and this path basically gets appended to the part we saw above and having it defined like this. It is just a parameter.

B

So what is really says here, if you would send something to API projects, slash projects, slash one, it would try to retrieve the resource named projects with the identifier one which we're going to see in a second one. More thing I want to show you uh first is there's no database behind this application, so this application here allows us to create new projects, retrieve projects by idea or retrieve them all. There's just no database.

B

Every time when I'm making a change to the project here, the server gets quickly compiling the code and restarts, which then leaves my in-memory database emptied. So you'll see me performing an action where I'll add some data to the database multiple times.

B

Right so starting with the rest of this microservice, because we first said: okay there's a few endpoints in there, but we need to add one. We don't have a get all end point yet where we can retrieve all projects. Now, let's see how simple it is to implement something like that, we would have a public method where we would call it get all. There's no parameters as we want to retrieve everything and basically repay and just say what we're going to return.

B

The response and a response is a response object where we can either say: okay and it will return HTTP 200, but in our situation we actually want to retrieve some data. So we have a service we'll talk.

B

How we head to this surface within a second and surface, find all method, and then we just need to say to the Builder Okay. That was it, and that is basically how you can make a get end point.

B

What we still need to do is annotate this endpoint with the get at get annotation so that this endpoint actually gets exposed.

B

Now, if I would save this, then you can actually see that I have the terminal here, and it said: okay, great. This application has now been updated in 0.2 seconds and it is available.

B

So that should give us at least the possibility to run the option here, and you see we get a response code 200 now, there's just no data in there yet. So that's why the application doesn't return anything. Yet if I would enter some data and I would run our endpoint again, then you can actually see that we're getting the data returned here in a list, and that is how simple Jakarta ee makes restful web services going back to one point that we talked about before I said: okay, we're calling this service method here.

B

So how do we get to the surface method? It is defined up here and it has an at inject annotation and inject comes from another API.

B

That micro profile is kind of loaning from the Jakarta project, which is CDI or context and dependency injection.

B

um Basically I have a project surface here which I gave a scope. In this case it was an application scope, there's different Scopes available. There's, request scope, this session scope, um there's conversations code and, depending on the profile you're choosing there can be different. Scopes application scope basically says kind of a Singleton I have one object which will be used for every part of the application.

B

um Because it is application scoped, it gets registered by CDI. So as soon as I say in my resource class that I want to inject project, surface CDI will look if something there has been registered with that name, and is it if it is, then it will get automatically injected into the variable we're talking about here service, so that is part of the dependency injection part of CDI. Then there's also the context part, and this annotation here is all about that.

B

What it tells here is basically, it says: okay, this resource, this class has request scope, meaning for every new request coming in I will create a new instance and and hand it. So that's completely opposite of the one we saw with application scope.

B

Having gone through all of this then I think we have covered the main parts um from Jakarta AE used in this simple application.

B

um Perhaps there's one more feature that we should just highlight a bit: um I was I, am returning a list of java project objects. Here, still my output gets to be in Json.

B

All that comes because I'm saying here up here: it produces media type, application Json, so the the jobs Eric's runtime actually knows. Oh wait. I got this object, but I need to translate it to Json. So it does all these things automatically. You don't have to work with Jackson or Jesus. It's all been done for us. So that's that's pretty pretty neat.

B

Okay! Moving on um these are the things we really want to talk about with in relationship to Jakarta ee itself. You've got this great microservices, it's doing really great, but then you find out when you deploy it on the big bad internet, the things actually might break and things probably will break you have this monolithic application and suddenly you have all these smaller microservices moving Parts, much more Network interaction between them. Something definitely at some point. Will go wrong so we need to add a bit more um Cloud nativeness to it.

B

So, let's see how we can do something like that. The first thing out of the box with me without me doing anything is that you get an open, API specification by default. If I would go to the server here and refresh this part here, you can actually see that out of the box I'm getting a lot of open API specifications so for each of my endpoints it is it notes down. Okay, um was it a post? Was it again was it the put? Was it the delete?

B

What are the parameters, uh what what what what needs to be in there? So that is actually quite good.

B

It's actually a complete set of all the values in there. It is out of the box like this. It is very easy to change. If you decide that you really want to add some extra information to it. That is very easily done.

B

For instance, if you say I want to rename this parameter and it says retrieve a project gave it a good description. You just edit here with these standard annotations and if we would refresh it here, it would here show that information. So it's very easy to add your own flavor to your apis and make it very easy for your clients to use your API.

B

Next thing on, because um one of the next steps in making an application like Cloud native, would be configuration according to the 12 Factor app configuration sources should always be outside of the application, so that it's easily to change.

B

um Think containers where you probably would use environment variables to configure data source, URLs user, IDs, passwords and stuff like that they get injected. In the last moment.

B

For this there is the config every API and micro profile, where you basically can define a project, a a configuration value, and it will look in certain hierarchical places for the right configuration value. The most default value where it will look, will be inside the micro profile, minus configure properties in the meter in folder.

B

Next, it will look at property files. It will look at overrides on the jvm, with the minus d uh parameter and finally, it will look at environment variables and environment variables will always overwrite the ones underneath and the ones underneath. So so, there's like this hierarchy, and that makes it very flexible to work with these kinds of things. Now, how do we access these configuration failures if we really want to access them and and have them inject it into a program? Well, that's exactly again how it works. We inject them in our program.

B

Take a look at the output here. You can see that a project contains an active, a name and a uid, but if we would actually look at a project, it also contains a company ID in this case we're not using that company idea at all because it doesn't get a value and if it doesn't get a value to no property and it won't show up in the Json.

B

So how can we change that? Well, I. Have this optional value, Company ID here and when I create a post. I. Actually check is compiled me id if it's present then set it to that value. So. Currently it is nil and it won't be set, but it's possible to do here is that you basically say Okay add inject, and then you say this is a configuration property.

B

And it has a name in this case it was company. That is the name that it also has in this configuration property watch the value 24.

B

And then we'll just make it automatically injected here so now, if I would run this application again and if I would post a new value, you can actually see here that suddenly Company ID gets filled with this value. So it's really injected, and this is the way you can inject your URLs, your passwords, whatever you need, can be injected this way in to your code now. What is interesting about configuration, for instance, is that in version 2.0, suddenly you have the possibility to Define profiles and profiles.

B

Basically saying okay, I got a different set of parameters for development, for test, for staging and for production and I can have all these values in my environment and based on a setting mp.confactor profile, where I can specify, which is the current active profile, it will take the right configuration properties um that really makes it very handy to use and and and very easy to switch environments.

B

So um there was a great addition in version 2.0, another edition of ones, that we could add property expressions before they were really static, and now we can have in our properties itself have Expressions, which is really nice and finally, another option that is quite usable is something called a config value, because I just told you there's a there's all these sources, where I probably tell you can come from and sometimes you're just not getting the property you would really want to until foreign.

B

Where does it come from? Well, if we would do something like this, we say: okay, this is still a conflict property I want you to inject it and I want to know the config values. Where does it come from.

B

Now, if I would take one of these endpoints and I would just tell it. Okay, just give me a system, I'll crank line.

B

For the config value, where from and then I want to see, okay, what is this actual Source name? I'm? Saving this and I will again call the endpoint.

B

And I would look at the output terminal. It tells me, oh, this comes from a property, Source config and it comes from the classes, meta in micro profile, configure properties. So that's a very nice neat easy trick to debug your application. If you're in doubt where the value comes from.

B

Right well, moving on to the next one. We have our API. We got it exposed via open API. We got the application configured, so the next Cloud native point that we would address would be health, and that would be the health of your application. Your service can actually tell outside sources. How he's doing the whole idea is that, with the health surface, you check the validity, the availability and the status of your microservices.

B

For this there's actually three annotations and three endpoints that can be exposed.

B

um There's the ready endpoint, which basically tells you is my business logic, ready to process events. There is the life and point that Basics is my Surface still responding at all, so it's still acting as I would expect it to act so more on a technical level as opposed to the Readiness and there's. Finally, a startup and point telling to the outside world I'm, actually still in the process of starting up.

B

As you can imagine, all these endpoints they're really machine to machine endpoints they're, not really meant for us use this to to to to check them all the time. They're really meant for container managers like humanitas to check the health endpoints and see. Do I need to restart this part, because it's liveliness and point really doesn't give us any good information anymore. So that's really uh what it's meant for still. We need to implement them ourselves.

B

It is again a CDI beam, so let's see how we can Implement such a thing. If we Define a normal Clause, then we need to implement the health check interface in order to make it.

B

Health endpoint and it has one method, and that is the health Dutch to the method, call that shoot return and health check response right. So what to do well, the health check response.

B

We can give it a name in our case, we'll call this the project thing and then we can decide. Okay, is it going to be up or is it going to be down depending on? If the server is running or isn't running, you can also give it return it status.

B

You could say it is up and then you say build, and this is basically your end point saying I'm doing very well now, in normal implementations, you would typically.

B

Add some additional data maybe check your database connections check if you can reach external system that you need to reach check if you can actually write to the Kafka queue stream stuff like that, so so that's normally a bit more than to just returning up and saying you're ready.

B

If you would would want to supply extra data, maybe they're just saying I'm having a problem with my database, so um I'm trying to reach the reddit's cash and it gives uh gives a problem or something you can actually make it available to the outside world. As additional information like this, so you can add an unlimited set of information to this now, even though we implemented the health check, it is still not a healthcare. We need to do a few things.

B

First of all, we would make this applications code because for a complete application, we're defining just one health check here. Next, we need to say in which situation should this health check be called? Well, we just said we have the redness.

B

We had the liveliness and we had the startup.

B

Well, with all these added, our health check application part here is completed, it got newly deployed, so if I would now go to the endpoint, which is always the health endpoint in a project which is at the surfer name and then health and I would run it, you can actually hopefully see here that it returned the information that we gave it. So it still says my status is up. This is my project and without the title it would have said, I am not available.

B

So this gives you really really nice quick information and on how your system is doing and actually it gives likely an container Management Systems the needed application. So what did we learn from this? So Health Service means again they're CDI beans, in this case application scoped the need to return a health check response.

B

So we have our API, we have a configuration we have. The health next thing we actually might want to know is: how is my server actually doing?

B

So how are its? um How is the? um How often is a method called and what is the average response time for a method or what are the longest delays?

B

You know all these kinds of metrics and for that there's the metrics project Within micro profile, and it provides us with a number of metrics that we can apply there's the times annotation that has all kinds of information about: okay: how how how long did it take and how often and which one which buckets of uh response percentages did it land in there's the counted just basically saying: okay, how often have I been invoked and there's different gauges and meters and histograms that we can use?

B

We are in our situation, we're going to just look at a few of these and going back to our resource that we had here now. We might want to know. How often does this get all method? How often does it get cold um now? Maybe you would like to know it on this level, where you say this is really I just want to know how often it gets invoked, and it is really as simple as saying I'm doing accounted here.

B

There's additional parametric parameters that you can use to specify what name it should have and I'm I'm in which bucket it should be so there's all lines of conf or options for configuration out of the box on the ad counted is enough.

B

Now, let's see how this works out, if we would go to the HTTP request and fire multiple times on this, get old endpoint.

B

It's returning us a number of two and now, if I would actually say. Okay, give me the metrics for this specific application.

B

It actually comes down with quite a large payload.

B

So I'll just use this in the browser and I didn't work at all. Did it.

B

Here actually I'll just zoom in a bit.

B

You see that we have a counter here, it's the application and then the name of the project, the method being invoked on what service and the type of counter- and it tells you okay well discount. It has a total of six times that it has been in a vote.

B

All these metrics they're in a Prometheus format, so they're easily exported into that format. Basically, there's endpoints! You can just get the data out there get it loaded in, for instance, Prometheus and have um have it displayed in graphical views.

B

So this was um for this very specific one and also already a whole lot of stuff being exported by default by your running application.

B

So it tells you, for instance, okay, how many threats are running into in the system? How many classes did the class loader unload? How long has the jvm been up? So, there's a lot of things from the application from java itself that is actually being insupported exported and where you can either use them or you can't.

B

But it's very nice if you want to see okay, what CPU usage right now on this service, um how much this space is left, although you probably have to add that one in yourself how much memory is being used so very, very useful uh stuff which is in there and then we're gonna, add one more because we just added on the health surface. I'm, sorry, sorry, on a project service, the counted, but maybe for a normal call to get all I would like to know.

B

Also let you know the turnaround time so how much time does it actually take for the complete trip to complete so I'm, getting a request in I'm calling the back-end service, the back-end service, retrieves data from the database reports it back. This data is going to get converted into Json and that is being returned to the client. So how long would that take?

B

Oh there's, another indication timed again with this annotation. You can use a lot of um a lot of parameters to set your stuff for, for specifically for your metrics, but this at times as it is suffices to have it working so I'm saving this.

B

Going to the end point again.

B

Calling it a number of times and then see what ended up in the metrics now nothing here, um I should go to that. Endpoint.

B

Metrics application.

B

And though it almost looks, if, if we have a complete set again, it didn't, it still got to a counter that got increased again, but you can also see that it also has the information for the get all methods and how long it took and that's the average one minute minimum rate per second, so that there's like all these percentiles as well.

B

So there's a lot of information being gathered there about how the performance of your system is, and you can imagine that, if you look at these figures and you'll have some graphical system to represent them, you can get very, very good insight into the ongoing performance of your system and your the average throughput of your system, so metrics very important um to make your application um more visible and more reliable, see. What's going on on the inside.

B

Good and then moving on we'll get to the final API that we're going to be discussing today, and that is for tolerance.

B

Okay, one more time we got the API, we got the configurations, we got the health, we got.

A

B

But now something might sometimes go wrong in our application. Now the Health Service is already helping us part there right, because, basically it says: okay communities, if, if you think I'm not doing well anymore, I'll tell you over my likeliness service and, if I'm doing really bad, please restart me, and that is what kubernetes will do, but that's kind of a very course current solution. Isn't it I mean you, don't really want to restart your complete application or your part of your container or whatever, just because there might be some hiccups somewhere?

B

So maybe you want to have more fine-grained controls over what um what is.

A

B

The system and try to make the system like you know, self-correct itself, self heal itself. Well. For this there is default tolerance, module within micro profile, and we will now look at a few options that it actually offers to us if I would go to the project service and then there here there is this very basic, find all method that we've just had before I'm just taking out the content.

B

um Well, the most basic and the the most easy option would be that you gave it a retry. We say at retry and you state the number of times it should retry the operation. Should it fail?

B

Okay, you can also specify a Max duration in this case, I put in a thousand and thousand, and this would be milliseconds.

B

But it is possible to actually specify the duration unit yourself. So if you would like one minute, then you could just specify 3600, milliseconds or just one and then a Time unit of a minute, as you can see, there's also some.

B

Other parameters that can be specified like Jitter, which is some kind of a random interval where it will wait until retrying again, but basically uh we'll get quite a long way by just specifying what we have here now. Should these 1000 milliseconds.

B

Expire before we had the maximum of three retries, then the system will just say: okay I'm down here, where we're not going to try any more retries. So this is really the cutoff time and talking about cutoff time. There is something similar which would be a timeout.

B

Where you again can specify in the time unit you want, how long does this call maximum is allowed to take so how quickly shoot it, and so, in this case we're saying it can actually take 1500 milliseconds.

B

So it would be one and a half seconds and if it doesn't complete within that time it will give a timeout exception.

A

B

We'll later see this one election, when we do the last piece of coding first, we will move on to different.

B

Options we can use, which is the circuit breaker and the circuit breaker is really quite an interesting one. You can really see it like a a power circuit in your house like a power line. Basically, the circuit is always close. Nothing is wrong.

B

The circuit is closed, but then, if something goes wrong, it might end up in a different state, in this case we're telling it if, in a request, volume threshold of 10 cores so 10 cores, if there's a real failure rate of 0.5, meaning from the 10 cores, five failed I actually want you to open the circuit halfway and what it will do is actually wait for a thousand milliseconds until it accepts any further traffic.

B

Then, once these thousand milliseconds have passed slowly, traffic will be flown, will be sent again to this endpoint or to this method, and once there's a success threshold of three meaning that three successive cores are successful and completed successfully, then the circuit will close completely again and all traffic can flow through again.

B

So this is a really powerful, powerful option to use, because you really can give the system some time with the delay- yeah, maybe you're kind of busy right now or you're struggling with something or you have something like um memory that is being reorganized at the moment. So so just just give yourself some time and maybe you'll recover yourself.

B

Based on a pattern.

B

A

B

Limit the the output of of the problems that are occurring, there's another pattern, which is called the bulkhead pattern, which is basically a very simple pattern and there's this analogy to shipping. Hence the name bulkhead. If there's a big container ship and the hole gets breached because it touches an iceberg or whatever, then, um if nothing has been done, then the whole ship will flood with water and it will eventually sink.

B

So what they've done with these big container ships is that they make different compartments and if one compartment gets breached, it will be closed off from the rest and water can't get in anymore and this another so meaning the shift won't sink, and this analogy works perfectly for what is happening here.

B

If we would have all these methods that you see here, these five methods, and only to find all, would really perform very badly that at some point, all the threats will be busy with processing find all methods that the rest of the system is basically starving and doesn't get any response anymore. So, there's one little part in your system that is performing badly, maybe because you're going to an external system, it's suddenly kicking down your complete system now with bulkhead pattern. Basically, you can prevent those things. There's two flavors won't go too much into them.

B

That's the semaphore pattern and there's the threat pool. Basically, here the semaphore pattern says: Okay, um there's only five maximum concurrent invocations of fine find all so that that five calls can come in from a client and I will process them. I won't do a sixth one. Why well, if we have a thread pool for 20, and we would do it unlimited then, basically, calls would come in.

B

They would take long to process threats, get blocked and, and none of the other 15 threats which would be in our 20 thread thread pool, would be able to answer to the other methods anymore.

B

So this is really a good way to protect your system and there's. Finally, a final annotation because sometimes.

B

Even though you try your best and all the other things, they really try to fix the system give it a retry gave it a longer. Delays, try to make the system still responsive with the bulkhead and the circuit breaker. Sometimes things just fail and there's a one final.

B

Method that we can actually use.

B

And that would be the retry I'll show you not to retry. Of course, that would be um the fallback and what the fallback basically says is: okay, if there's nothing, I can do anymore. I really can't reach that Source, where you want me to get the data from I can either be really nasty and give 500 server, error or I can just be a bit more graceful. Now, let's see what's happening, if we don't do it gracefully.

B

So, if we put an annotation on here- and we say.

A

B

There's a timeout.

B

Of 30 milliseconds and then we'll just make the system a bit slower machine. We'll do a threat sleep of two and a half thousand milliseconds I. Guess you can imagine, what's going to happen, the system will say: okay. Well, so sorry he didn't respond in time and something is wrong. So if we would do it like this and I would call the method.

B

I'll just clean this down here.

B

Sorry, it doesn't have the end point anymore.

B

It's funny you should have had that.

B

Seems to be something wrong.

B

B

We can't use this one yet because it.

A

B

There yet I said so: that's why it won't fail Now. It worked so I'm trying to call it again. Within These delays.

B

System will wait a bit and we will end with the infamous 500 internal server error, which isn't very elegant now. If you decide no I want to do it differently.

B

I'll, maybe instead of giving a five returning a 500 I would just like to well I. Don't know, I'll return, an empty list then, or something something or.

A

B

Got your data: maybe you got your data somewhere in a caching system and you could say: oh well, I'll I'll see if that works for me, so I'll just return, maybe in all the few, but still better than giving a 500.. Now we said here we have a backup method, so we just need to Define one here, and the only thing is that is important is that it should have returned the same data type as the method. It is a backup form. So in this case we would say list project backup have no parameters.

B

Now we don't have anything to really to return right now. So in this case I guess I could just say: I'll return, an empty list.

B

um Might not be great, but it might be an indication to the user, at least that okay, he didn't find any projects that can't be right, so something must be, uh must be going on so having saved this and trying to invoke this method again.

B

I'll just clear it here invoke the method. Again, then you actually see that we're getting a 200.

B

Saying, okay and we got a response of two bytes, meaning just the empty Json array, so so we can still return some some kind of value that could be of use to uh to our users great.

B

um So this was the end of the presentation, so I guess now is a good time. If there are any questions there.

A

Is thank you so much Ron for that amazing presentation and and demo. uh We do have two questions. We have one from Helmut, um so it's if you have a real database as backend, would you use at uh ejb annotations instead of CDI.

B

Oh, that's a very interesting uh question: um I I guess it really depends upon because agbs in the old type um uh in the old style, there were like two things: there were like surfaces and there would also be persistence. Persistence is now separately in this case. I think I would even if I had a real database. I would still use um our normal things that we're using right now. The way I used it now, because there we also can have transactions in there.

B

So there's no reason to use agbs I can see why you want to want to do that, um but you can use the same I understand because there's also a Singleton scope, Within AGB beans, that kind of compare to the application scope, but um you can use basically both it really depends also if you would go to the core profile that I was showing before I'll just pick it up. Here we have this core profile on the right. Basically, beans are no longer part of that.

B

The Enterprise means they're in the web profile, but they're not in a core profile, and would you working in a core profile with your microservices? You would have no other alternative than to use the annotations I already used.

A

Okay, great and another question um from the same uh same person is: is it possible to react in a Health, Service, System, metrics, uh P, dot e force a reboots by marking a system as down? If the memory footprint surpasses a limit.

B

um Yes, what you you could force a reboot of course, because you have all these metrics available. You can even inject them individual, metrics and or retrieve them from the metrics registry, and what you could do in your Health. Service, of course, is go to this magic registry. Look at okay! What's my current memory footprint? Is it below a threshold?

B

Then I'll just return, not the up like I did before, but we turned it down and the down would then be an indicator uh to The Container management platform like kubernetes to say: oh, wait, a minute. You gave me a down, then I'll automatically reboot the container, so it would still be like the container management platform that would do the reboot, but you can actually give it an indication saying: let's reboot, yes, definitely question.

A

Great and then the last question is what happens if the circuit breaker can't fix the problem? Are there any automatic fix fixtures.

B

Yeah, that is actually also a very good question helmet. Let me show you one example that I didn't even get to, but I would really love to show I'm just taking out everything. I had here, go back to the original one and then because I just showed them um in all individually, but it is actually very very possible to to have all these things run together as one one one big group of set that is um running in in conjunction. So let's look at this code here.

B

Basically, I got a circuit breaker, a retry, the bulkhead and the timeout, and a fallback method so that that they're all in there right so first, what happens is um we get to this message find all and the circuit breaker will first check? Is the circuit open? If the circuit is open, no traffic will flow to it and it will find a different method now if it is not open but close, that means the traffic will go through and the bulkhead the timeout will actually start.

B

A timer will start running, see that we don't surpass the 500 milliseconds. The next thing that will happen is that the bulkhead will look. Is there still a spot for me? Am I allowed to execute this method, or are there already too many executing now?

B

um If it isn't, then it will just execute as normal. If it would be, if the bulkhead would say no, we can't do it right now, because all the fats are used, it will actually throw a bulkhead exception which will then be caught.

B

And then the retry will take over again, and this continues to work in conjunction, so they keep working like this. The circuit breaker might go open it. It will retry again and in the end, if we can't get a decent result, it will go to the fallback method. So um long answer to the question: yes, it can fix it. The circuit breaker will throw an exception, but if you use them in conjunction, then you actually might have automatic fixtures in place.

A

Okay, great, it doesn't look like there's any other questions, um so we're also looking to book some more Jakarta Tech talks through 2023. So if anyone is interested, I did link uh the form in the chat. And finally, if you have any feedback on the tech talk programs, we would love to hear from you. So when we close out this chart you're going to get a little thumbs up or red thumbs down- and just you know, give us some feedback.

A

If you enjoyed this talk and and if you've and if not, um you can also give us some feedback on ways that we can improve. So I just want to say thank you so much Ron again, um and everyone for attending and I hope you all have a wonderful day.

B

A

You very much thank you so much take care bye-bye.