Disclaimer: No Intent to offend anyone. I used to be an Engineering school teacher and I got my students understand tough principles with metaphors and similies. Learned this unique method from my chemistry teacher, whom everyone adored :)
Let us put an example: Let us say Tesla set out to build the fastest car. At it they built a self driving car. Would a faster car be more faster than a self driving car, given the owner could meander and use their time more wisely, that way making journey use less wasted time ??? At this juncture would it have been smart to just build a slow car so fast that we deliver it in three development cycles ? and not be too fast, neither complicate the concept and hire cheap man power ??? Would perceived performance of the manufacturing team be more important than the performance of the manufactured product ??? Tesla is the answer - multitude of innovations implemented at once is needed to survive.
Now persuading via history - eg: PRINTING PRESS. Taking years to study a simple process of writing by hand, but trying to automate it by printing - was the first machine. We used centuries of years to accomplish this feat, had several different intermediary machines for printing, while 100,0000 of our population was jobless and had hands available free to write. Engineering has a goal and effect, it is for that one person who cannot hire 100,000 people to print a book, but has the grand wisdom to write one worth read by 100,0000. If you really think, free market is not possible without adeqaute engineering. Why ? Core free market idea relies on that man printing that book about that neuvo idea.
Link between Printing and Microservices ? Cloud and Microservices, enables small companies to have cheap shared cloud hardware from amazon or own company cloud and scale as they grow - rather than buying expensive unix servers upfront spending 1 million dollar. They will spend 1 million dollar on cloud later and not tank economy by not doing so :), but not just yet!
Now persuading by pure finesse. Any number of hours used to skim milk to create the gentlest delicatessen cream, will never fail to gather consumers, neither will it fail to fetch top dollars.
Now persuading via Time Vs Money axis: All innovations that can be done by fast hard work alone, has been already done - is a safe assumption. Universe had been around for millions of years and work is cumulative. Hence what is left to do is smart work. What is the proof, I would have hired people with money and forced competition using predatory pricing to produce virtual immense fast hard work. I could have worked long period of time to produce all that could be produced via fast hard work eventually and posthumously - and if not yet I will. Elongated periods of slow work simulates fast work - by principle of being cheap. The concept is bending the time axis - to simple past and future using money spent on work as a medium. What is left - smart work ? Any work without an egineering advantage is already done or can be reproduced easy is the proof.
Microservice is a very abused term and Microservices are not supposed to be Micro in size. However they have the capability to be Micro, in case of a complete CICD requirement. Be warned this comes with a very high hardware price and automated testing overhead and fulfillment of complete Microservice independence requirement. Complete independence means, a database per Microservice.
A better design is split Microservices per large homogeneous entity relationship. Eg: Order Entry System in one Microservice, Product Catalog and Inventory Management in another Microservice. Making Product Description iteself an independent microservice in Netflix model does provides ease of independent deployment and speed to market, but it involves large infrastructure. Each tiny Microservice will have to run in it's own docker container.
When everything is running, hit the url as follows -
- Do the following command in docker (you got to understand docker is != localhost or 0.0.0.0)
docker-machine ip- which usually yields 192.168.99.100. You will have to open this port on unix, to browse
- In Browser -
http://192.168.99.100:8091/api-gateway/mls/getDecision1 - In Docker -
curl http://192.168.99.100:8091/api-gateway/mls/getDecision1
A conservative Microservices architecture allows us to scale our deployments properly and efficiently and structurally sound from aspects of failover and availability. That is all that Microservices focuses as a solution. CICD is a different animal than Microservices and do not belong in the species at all. Splitting Microservices properly allows us to scale the busier servcies to have more instances and scale up and down a group of services as per PRESENT TIME load and achieve cost effectiveness on hardware, as well as high availability and fault tolerance(failover). This is the real use of Microservices architecture. Microservices framework achieves fault tolerance using framework capabilities and server farms are integrated using provided mechanisms.
For the sake of demonstrations that would run on same machine, to simulate an East, West region, we use different port numbers for those servers. However in real life you can run on same port as in East on West as the IP Address is going ot be different. The only paramteter that needs to be input in that case is the host port of the main server.
The docker image has the default values for Eureka sever, Zipkin server, MSApiGateway server. If you configure your docker startup scripts to feed an IP and port to the main server, rest of the values can be hardcoded. What do I mean ? Run Eureka server always on port 8761,8762. Run ZipkinServer always on Port 9411. Run MSApiGateway always on port 8091.
The Micro Service Server, SCALE UP OR DOWN up as many on port 8080, 8083,8085,8087,8089 ... etc
Please click the link above for details. This document details our scaling strategies.
You can run the app quickly by doing the following
- git clone https://github.com/cheerans/Microservices-Architecture.git
- cd Microservices-Architecture
- chmod 755 run-docker-compose.sh
- ./run-docker-compose.sh
A cool place to run it would be, docker collaborated - https://labs.play-with-docker.com/
Please click the link above for details. This server manages our service registry and discovery. Our Microservices registers with Eureka servers. The consumer applciations and ZUUL proxy queries Eureka server for discovery.
Please click the link above for details. This is the Microservice implementation project. For demonstartion sake, one Microservice here calls another microservice. However the call is made through service discovery. Hence it demonstrates load balancing concept clearly. If you run the application, the output traces port number and it will show you that the server is calling through load balancer.
Please click the link above for details. Zipkin server logs all the service calsl and maintains traceid for one set of service calls, that are part of same call. This could come handy debugging performance or other issues.
Please click the link above for details. The idea is to implement an API Gateway. Why Gateway, several Microservices can be tailored to one API gateway and load balancing can never go wrong when a service is consumed via a gateway, plus security can be implemented easily hiding microservices behind a DMZ zone.
AutoScaler will scale the services running on the local docker. The details can be seen in this section
Please click the link above for details. This area shows how we can avoid cascading failures and avoid certain circuit paths by resolving to return failure.
Refer to this article - https://reflectoring.io/spring-boot-health-check/
http://localhost:8761/eureka/apps/mlservice-east - The above shows the metrics about a server and all its instances
http://localhost:8761/eureka/apps/mlservice-east/localhost:mlservice-east:8091 - The above shows the metrics about a server instance
http://localhost:8091/actuator/health - The above shows the status of the server
http://localhost:8091/actuator/prometheus - shows CPU usage etc if you add followinf dependency
<dependency>
<groupId>io.micrometer</groupId>
<artifactId>micrometer-registry-prometheus</artifactId>
</dependency>
add the following entries in spring properties file
management.endpoint.health.show-details=always
management.endpoints.web.exposure.include=health,info,prometheus
The ports of a docker container is not automatically forwarded to the localhost:port. Hence we need to do the following instruction. The magic is done by microsoft\iis which spawns up IIS server on the windows machine and makes the docker container available. Here port 8080 on docker container is available. To build an image do the following
docker build -f Dockerfile.iissite -t santhoshcheeran/iissite .
docker push santhoshcheeran/iissite
docker run -d -p 8080:80 --name iissite_instance santhoshcheeran/iissite
However again the intricacy is that this port is only available on the docker container IP Address and not on local host. The following will obtain docker containers IP Address.
docker inspect -f {{ .NetworkSettings.IPAddress }} <container>
Hence forth typing dockerip:8080 in internet explorer provides the access to local docker container via browser.
On Linux the ports are forwarded to localhost or 127.0.0.1.
- git status
- git log - shows all the changes
- git reset - remove all changes
- git stash - stage and save the changes
- git reset
- git pull
-
docker container ls
-
docker container stop - Gracefully stop the specified container
-
docker container kill - Force shutdown of the specified container
-
docker container kill $(docker ps -q)
-
docker container rm - Remove specified container from this machine
-
docker container rm $(docker container ls -a -q) - Remove all containers
-
docker image ls -a - List all images on this machine
-
docker image rm
- Remove specified image from this machine
-
docker image rm $(docker image ls -a -q) - Remove all images from this machine
-
docker login
-
docker ps --filter "name=mlserver-east"
-
docker exec -it mycontainer sh -i means interactively -t using tty opens a shell named mycontainer(just a dummy name
-
docker run -P -d --name employee --link mycontainer microservicedemo/employee --link links to an existing container -d means run as a daemon in background -P tells Docker to expose any container-declared port in the ephemeral range
-
docker run -P -d –name mycontainer mongo Runs mongodb in mycontainer. There are ‘official’ Docker images of many popular software packages. MongoDB’s is simply ‘mongo’ – you know it’s an official image by the lack of an owner name (e.g. owner/image)
-
docker ps - shows process
-
docker port - all docker ports
-
docker inspect - all config dump
-
docker logs -f "container name" - shows logs of the container
-
docker stats [container name/ID] -
-
docker exec container_name curl http://localhost:8761 - does curl inside a container
Good Luck and spend time and code slow, but smart.
- Fork the project
- Create a feature branch based on the 'master' branch
- Create a PR and feel proud.
MicroServicesArchitecture is released under the MIT license, see LICENSE.
# Microservices Architecture
# Microservices