Demystifying Service Mesh: How it Works and Why You Need It

 Introduction:

In the ever-evolving landscape of modern application development and deployment, the concept of a "Service Mesh" has gained significant traction. As a tech blogger with over 12 years of experience, I'm here to provide a comprehensive update on this crucial topic. In this article, we'll delve into what a Service Mesh is, how it works, and why it has become an indispensable tool for managing complex microservices architectures.

What is a Service Mesh?

A Service Mesh is a dedicated infrastructure layer designed to facilitate communication between the microservices that make up an application. It acts as a transparent, language-agnostic network of interconnected components, providing essential functionalities such as service discovery, load balancing, security, and observability. The primary goal of a Service Mesh is to enhance the reliability, security, and manageability of microservices-based applications.

How Does it Work?

Now, let's dive deeper into how a Service Mesh actually works:

1. Sidecar Proxy: At the heart of a Service Mesh, you'll find a sidecar proxy. Every microservice in the application is paired with its own proxy, effectively forming a "sidecar." These sidecar proxies are responsible for intercepting all inbound and outbound network traffic to and from the microservice they are attached to.
2. Service Discovery: When a microservice needs to communicate with another service, it queries the Service Mesh for the location of the target service. The Service Mesh provides dynamic service discovery, ensuring that services can locate each other regardless of their changing IP addresses or locations.
3. Load Balancing: Service Meshes implement sophisticated load balancing algorithms, distributing incoming requests evenly across instances of a service. This helps in optimizing resource utilization and ensuring high availability.
4. Security: Security is a top priority in microservices architectures. Service Meshes offer robust security features like mutual TLS (mTLS) encryption, authentication, and authorization. With mTLS, all communication between microservices is encrypted and authenticated, significantly enhancing the overall security posture.
5. Traffic Management: Service Meshes allow for fine-grained traffic control and routing. This means you can implement A/B testing, canary releases, and gradual rollouts with ease, all while monitoring the impact on your application's performance and stability.

Observability and Monitoring: Service Meshes provide rich observability features, including metrics, logging, and tracing. This enables DevOps teams to gain deep insights into the behavior of their microservices and diagnose issues quickly.

Why You Need a Service Mesh:

Now, you might wonder why Service Meshes have gained such popularity. Here are a few key reasons:

1. Microservices Complexity: As applications become more microservices-oriented, managing the complexity of service-to-service communication becomes increasingly challenging. Service Meshes provide a centralized solution for handling this complexity.
2. Resilience and Reliability: With features like load balancing, circuit breaking, and automatic retries, Service Meshes improve the overall resilience of your application. They can handle failures gracefully, reducing downtime and improving user experience.
3. Security: Service Meshes enhance the security of your microservices by implementing encryption and authentication. This is crucial, especially in multi-cloud or hybrid cloud environments.
4. Observability: The ability to monitor and troubleshoot your microservices is essential for maintaining high availability and performance. Service Meshes offer a wealth of observability tools that simplify this process.

Conclusion:

In the world of modern application development, a Service Mesh has become more of a necessity than a luxury. It offers a unified solution for managing the complexities of microservices architectures, ensuring reliability, security, and observability. 

Streaming Spring boot logs to ELK stack

In my previous blog, we have done ELK installation on windows 10 and we have even tried to push messages from input console to Elastic Search and finally viewed on Kibana Server.

I will write a separate blog on why do we need ELK?

In this blog, I’ll show you how can we push spring boot application log directly to Elastic search using Logstash which we can analyze on Kibana and If you don’t know how to install ELK on windows 10 then you can refer my previous blog and start Elastic Search and Kibana server.

Prerequisite

• Elastic Search and Kibana running on your machine
• Basic knowledge of Spring boot application

If you don’t want to start your application from scratch then you can download one spring boot application from my GitHub repository as well.

I am assuming that the Elastic Search and Kibana server are running on your machine and you have a fair idea of how to start the Logstash server and what is Logstash conf file.

So, to push spring boot logs continuously to Elastic Server, We have to open one TCP port in Logstash server and for that we have to create one Logstash config file (say elklogstash.conf) under ${LOGSTASH_HOME}/conf directory mentioning on which port TCP port should be listening under input filter and where to push the data once we received under Output filter.

For simplicity, I am skipping the filter tag as it is optional.

elklogstash.conf

Now start the Logstash server bypassing newly created conf file.

   bin\logstash -f .\config\elklogstash.conf

Cool! Now Logstash server is also up and running and if you observe the log, you will realize that it is also listening on port 4560 as mentioned in the conf file. Configure the newly created index (elkbootlogs) on Kibana as we have done during the ELK setup.

Now let's do some changes to spring boot application so that it can push all the logs to 4056 TCP port.

For this tutorial, I am using spring-logger project from my Github repository.

Add below dependency to the pom.xml file. We need Logstash encoder to encode messages.

<!-- Added for logstash Encoder-->

<dependency>

<groupId>net.logstash.logback</groupId>

<artifactId>logstash-logback-encoder</artifactId>

<version>6.2</version>

</dependency>

Open logback-spring.xml file which is under the resource folder and create new appender (say elk). The task of this appender is to push logs to the destination TCP socket and under this appender, compulsory use LogstashEncoder.

<appender name="elk" class="net.logstash.logback.appender.LogstashTcpSocketAppender">

    <destination>localhost:4560</destination>

    <!-- encoder is required -->

    <encoder class="net.logstash.logback.encoder.LogstashEncoder" />

</appender>

Add new appender to root level

<!-- LOGGING everything at INFO level -->

<root level="info">

<appender-ref ref="RollingFile" />

<appender-ref ref="Console" />

<appender-ref ref="elk" />

</root>

Save all files and start your application. So, we are done with all the setup. Its time to check whether all the changes are done properly or not.

Open Kibana on your browser (http://localhost:5601) and select your index under the Discover tab. You will see all logs are populating on Kibana as well.

Congratulations! Our configuration is working absolutely fine and it is pushing logs to Elastic Search. 

You can download the source code from here, ELK code chnages are under elkstack branch.

MicroServices: Spring cloud ribbon with Discovery Server

In this article, I am going to share my knowledge on Spring Cloud Ribbon and how can we use Ribbon with RestTemplate as well as with Feign Client. We will also see how Enabling discovery Sever will improve the scalability of Microservice.

Before jumping into Spring Cloud, I am assuming you must be having knowledge of Eureka Server, Feign Client, and Client-Side load balancer. If not then read my below blog before jumping to Spring Cloud ribbon. Also, I am going to use my existing code to implement Ribbon.

URLs:

• Declarative REST Client: Feign
• Microservices: Service Registry and Discovery
• Netflix Hystrix Circuit Breaker

In my previous blog, I have already talked about the Eureka Server and how other applications are taking advantages of Eureka Server to fetch the host/port of client application.

We have seen that three microservices application are up and running i.e.

• Discovery Server
• Product Service
• Price Service

Where Product and Price service will register themselves to Discovery Server and Product-service will always communicate to Discovery Server to get the exact location of Price-service and then only it will talk to the price-service application.

Imagine there is a high load on the price-service application and to handle it, we have started two more instances of price-service.

• How will you make sure your product-service should talk to all three instances of price-service and divides the load equally to each server?
• How will you manage the heartbeat of the application so that product-service should not hit INACTIVE instance of price-service which just got shut down because of some internal reason?
• How will you get to know how many instances are up and running of price-service?

For all the question, there is only answer which is Netflix Cloud Ribbon. It's a Spring cloud library which primarily provides client-side load balancing algorithms.

let's implement it and see how it can solve our problem.

Add below dependency to product-server pom.xml file as it is the one which is going to consume price-service API.

<dependency>

<groupId>org.springframework.cloud</groupId>

<artifactId>spring-cloud-starter-ribbon</artifactId>

</dependency>

Case 1: Ribbon + Eureka Server + RestTemplate

As our application is already acting as Eureka Client and using RestTemplate to fetch price record. Let's just add @LoadBalanced annotation to RestTemplate to enables Ribbon functionality.

It will allow the product-service application to use price-service as the address of price-service application and will discover the host/port of all instances of price-service from discovery-server.



Note:
@EnableCircuitBreaker <-> Used to enable Netflix Hystrix
@EnableHystrixDashboard <-> Used to check circuit state on Dashboard.
@EnableFeignClients <-> To enable feign client (Needed for case 2 scenario)

Start discovery-server, product-service and start two instances of price-service. You can do it easily by just overriding port number under Eclipse Run As Configuration.



Let's confirm it, whether all the instances are up and running or not by hitting discovery-server URL (http://localhost:8761/)



Now hit product-service URL multiple times from the browser and then go and check both the price-service logs. you will observe that few requests are coming to one instance and others on second one.



Great! Netflix Cloud Ribbon is successfully implemented and working absolutely fine.

If you start another price-service instance and hit the product-service URL again then you will find the request logs in the third instance too without doing any modification/configuration to any files.

Case 2: Ribbon + Eureka Server + Feign Client

If you are not aware of the feign client then you can read my blog here.

I'll just talk about Ribbon Integration with existing FeignClient application assuming you are already aware of Feign Client and implemented it.

In product-service application, I have already exposed another Endpoint (http://localhost:7001/products/feign/1) which consume price-service API using Feign Client.

To enable Netflix Ribbon, Add @RibbonClient annotation to the feighClient interface and pass your consuming service name (price-service).



Now restart your application and hit new endpoint. you will observe that the requests are distributed to all the instances of price-service.

Case 3: Ribbon + (RestTemaple/FeignClient) + NO Eureka Server


Can we use Netflix ribbon without integrating Eureka Server, The answer is YES but it would not be a good design. So, I would not recommend it.

When your application is not integrated with Eureka Server, in that case you have to list down all the address manually to properties file.

Remove @RibbonClients annotation.
Add below entry to your application.properties file under product-service.

#Enable this property if you are not using Eureka Server
price-service.ribbon.listOfServers=http://localhost:8002,http://localhost:7002

Imagine there are 1000 instances of price-services are up and running so we have to add all the instances manually. It could be a nightmare if we have to do it. :)

That's all for Netflix Ribbon now and do let me know if you have any confusion/query or you think I am not right somewhere. Please feel free to comment. Thank you!

As usual, you can download the spring-boot-microservices from GITHUB.

Declarative REST Client: Feign

We already know that how microservices communicate with each other using RestTemplate. In this blog, we will see that how this can happen using Feign Client as well. 

What is a Feign Client?

Feign is an abstraction over REST based call. It makes writing web service clients easier. It Is as easy as creating interface and then annotate it. It has pluggable annotation support including Feign annotations and JAX-RS annotations. It also supports pluggable encoders and decoders and has supports for Spring MVC annotation.

Using Feign, microservices can easily communicate with each other and developers don't have to bother about REST internal details and can only concentrate on business logic.
Implementation.

I am going to use my previous applications to demonstrate the working of Feign Client. So before starting feign-client application. Make sure all the below three applications are already up and running. 

• Discovery-server
• Product-server
• Price-server

Feign Client Implementation

• Create a spring boot application and add Feign starter dependency to our pom.xml file.

• Add @EnableFeignClients annotation. With this annotation, we enable component scanning for interfaces that declare they are Feign clients. 

• Declare a Feign client using the @FeignClient annotation.

• Either name or url must be passed under @FeignClient annotation. If name attribute present then it will fetch the service location from service registry and then hit product-service and in case of url, it will directly hit the application to get the data from product-service. In case of both present then it will always first check for the name attribute. 
• Use Spring Web annotations to declare the APIs that we want to reach out to.

Note: If your application is not registered with service registry then directly use url attribute to get the data but I would recommend to use name attribute and to use name attribute, make sure your application is connecting to Eureka server to get the service location. 

Read my blog to understand on how to add support for Eureka Client to your application.

Now start your application and you should be able to fetch product-service from feign-client. 

You can download the source code from here. 

Custom Configuration

Feign Client also support for Custom configuration changes like telling Feign to use OkHttpClient instead of the default one in order to support HTTP/2. 

Let’s go deep down to check how can we customize Feign Client.

There are two ways to configure it i.e. using properties file or override them using a @Configuration class, which we then need to add to the FeignClient annotation. 

Properties file:

Using @Configuration

Logging:

By default, a logger gets created for each Feign Client and to enable it, we have to declare it in the application.properties file using the package name of the client interface.

logging.level.com.com.waheedtechblog.feignclient=DEBUG

There are four logging levels to choose from:

NONE – no logging, which is the default

BASIC – log only the request method, URL, and response status

HEADERS – log the basic information together with request and response headers

FULL – log the body, headers, and metadata for both request and response

Handling Errors with Feign

By default, Feign’s default error handler, ErrorDecoder.default, always throws a FeignException.

To customize the Exception thrown, we can override it by writing our own Custom Error class and implements ErrorDecoder and declare this bean under @Configuration as we did earlier.

Spring-boot microservices can be downloaded from GITHUB

Happy Coding...!!!

Reference:

https://cloud.spring.io/spring-cloud-netflix/multi/multi_spring-cloud-feign.html

Microservices: Service Registry and Discovery

In my previous blog, I have already talked about Netflix Hystrix Circuit Breaker Microservices and explained it by creating two applications i.e. Product Service and Price Service.
In this post, I’ll use both applications to explain what is the Service Registry and Discovery Server.

• Circuit Breaker and Microservices Architecture

• Netflix Hystrix Circuit Breaker 

What is Service Registry and Discovery and why do we need it in the first place?

In our monolithic application, mostly service invoke one another through language methods or procedure calls and even in traditional distributed system deployment, services run at fixed, well-known locations (hosts and ports) and so can easily call one another using HTTP/REST. Over here the network locations of service instances are relatively static.

With the microservice architecture system, this is a much more difficult problem as service instances have dynamically assigned network locations. Service instances change dynamically because of auto-scaling, failure or upgrade. So, you can see that it’s not possible to hard code the locations (hosts and ports) of service instances in our application.

Apart from that, most of the time we have to deploy the same instance on multiple hosts (E.g. During Big billions day sale where 1000’s of instances is running to support high demand) which will be again a tedious task for the load balancer to register and deregister itself for all these services. 

What is Service Registry and Discovery?

Service Registry is the server instance where all the service provider register itself when it starts up and deregisters itself when it leaves the system. The service instance’s registration is typically refreshed periodically using a heartbeat mechanism.

Whenever a client has to communicate with other service instances, it has to first query to Service registry to get the network location and then uses a load-balancing algorithm to select one of the available service instances and invoke a request. 

The above instance discovery pattern is called a Client-Side discovery pattern and this can be easily implemented using Netflix Eureka as a Service Registry. It provides a REST API for managing service‑instance registration and for querying available instances.

Netflix also provides Netflix Ribbon which is an IPC client that works with Eureka to load balance requests across the available service instances.

Implementation

As we already have two microservices product-service and price-service and we know product-service is internally invoking price-service to get the price detail of the products. We will do minor changes with these two applications to support service registry.

Discovery Server Implementation

• Create a spring boot application and Eureka Server starter dependency to your pom.xml file 

• Add @EnableEurekaServer annotation to Enable Eureka Server.

• By default, Each Eureka Server is also a Eureka Client which needs at least one service URL to locate service registry. We have to disable this feature as we are creating single node Eureka Server. 
• Add below properties to application.properties file.

• Now start the discovery-server application and access http://localhost:8761 which will display the UI similar to below screenshot. 

• As of now, no application is registered with the Eureka Server. 

2. Registering product-service and price-service to Eureka Clients

• Let’s make out product-service and price-service as Eureka Client which should register to Eureka Server itself on starts up.
• Add Eureka Client dependency on both pom.xml files.

• Configure eureka.client.service-url.defaultZone property in application.properties to automatically register with the Eureka Server. 

• Now start product-service and price service application and hit http://localhost:8761. We will see product-service and price-service is registered with SERVICE ID as PRODUCT-SERVICE and PRICE-SERVICE respectively. We can also notice the status as UP(1) which means the services are up and running and only one instance of product and price-service are running. 

3. Update hard code URL with application name in product-service and restart the application. 

4. Now open browser and hit product URL and you will get the below response.

Congratulation!!! Our application is perfectly working with Discovery Service. 

Spring-boot microservices can be downloaded from GITHUB

Happy Coding...!!!

Netflix Hystrix Circuit Breaker

Netflix Hystrix Circuit Breaker

In one of my previous blogs, I have already discussed the Circuit breaker pattern and its usage, Today, we will see how can we implement it in our application using Spring Cloud Netflix Hystrix.

In this document, I’ll walk you through the process of applying circuit breakers to potentially-failing method calls using the Netflix Hystrix fault tolerance library.

Hystrix is watching methods for failing calls to related services. If there is such a failure, it will open the circuit and forward the call to a fallback method.

To understand it in a better way, I’ll take the same problem statement that I have already discussed in my previous blog i.e. E-commerce Portal. 

I am assuming you must be aware of Spring boot framework as this implementation is completely based on it.

As per problem statement, we need two applications i.e. Product Service & Price Service but in this tutorial, I’ll talk about just Product Service as Price Service is simple web application exposing a single API and you can clone it from GitHub.

I’ll talk about Product Service in which we will integrate Circuit breaker pattern and Cache Service implementation and internally, it will invoke Price Service to get the Price Detail.

Optional: You can download the below project from GitHub or you can try it by creating a new one. I would recommend you to try it from scratch.

Product Service:

Create Spring boot application and add below dependency to your pom.xml file.

<!-- Optional. Application health monitor check -->

<dependency>

<groupId>org.springframework.boot</groupId>

<artifactId>spring-boot-starter-actuator</artifactId>

</dependency>

<!-- Compulsory for circuit breaker implementation -->

<dependency>

<groupId>org.springframework.cloud</groupId>

<artifactId>spring-cloud-starter-netflix-hystrix</artifactId>

</dependency>

<!-- Optional. Dashboard in case you want to see the circuit state UI -->

<dependency>

<groupId>org.springframework.cloud</groupId>

<artifactId>spring-cloud-starter-netflix-hystrix-dashboard</artifactId>

</dependency>

To enable Circuit Breaker in spring boot application, add @EnableCircuitBreaker annotation on product-service entry-point class. 

Now use @HystrixCommand annotation on themethod we want to apply timeout and fallback method.

I have annotated @HystrixCommand(fallbackMethod = “getProductDetailFromCache”) on getProduct(String Id) so that if it doesn’t receive the response within the certain time limit or request get failed while calling price service API then fallback() should get called over here and fetch the price value from Cache Service. 

Make sure, the fallback method should be defined in the same class and should have the same signature.

We can customize the @HystrixCommand default behavior by configuring properties using @HystrixProperty annotations. Will check in depth in another blog. 

Add RestController class and other required class as per source code present in GitHub.

Add server.port to the application.properties file.

Start both applications and try to fetch the product Detail. In our case, our product-service app URL is:

http://localhost:7001/products/1

Now try to fetch product which does not exist

You will observe that over here circuit breaker is still in the closed state even after the exception. 

Why? The reason is that we have added  ignoreExceptions = { ProductNotFoundException.class } to @HystrixCommand so that it should not trip circuit if the product is not present.

Now, Stop the price-service application and hit the product URL again:

Check the log:

Congratulations! Our application is working as expected.

Monitoring Circuit Breakers using Hystrix Dashboard

Hystrix comes with a decent dashboard where we can monitor the status of Hystrix commands

To enable it, Add Hystrix dashboard dependency to the pom.xml file 

<!-- Optional. Application health monitor check -->

<dependency>

<groupId>org.springframework.boot</groupId>

<artifactId>spring-boot-starter-actuator</artifactId>

</dependency>

<!-- Optional. Dashboard in case you want to see the circuit state UI -->

<dependency>

<groupId>org.springframework.cloud</groupId>

<artifactId>spring-cloud-starter-netflix-hystrix-dashboard</artifactId>

</dependency>

Add @EnableHystrixDashboard annotation on the entry point class

Add management.endpoints.web.exposure.include=hystrix.stream to application.properties file.



Start the product-service application and then go to http://localhost:port/hystrix to view the dashboard. 

In Hystrix Dashboard home page enter http://localhost:7001/actuator/hystrix.stream as stream URL and give Product Service as Title and click on Monitor Stream button. 

Now, we can see that the Circuit status along with how many calls succeed and how many failures occurred, etc. 

Spring-boot microservices can be downloaded from GITHUB

Happy Coding...!!!