Building Microservices - Sam Newman

Question 1

What are Microservices?

Answer 1

Fine grained services, with their own lifecycle, built around a business capability, that collaborate together.

Question 2

What is the programming principle that Microservices better enforce?

Answer 2

Decoupling (a change to one service should not require a change in another)

Question 3

Why are Microservices good to test new technologies?

Answer 3

The biggest barrier in adopting new technologies on monolithic systems is the risk associated in changing the whole system.
With Microservices one can choose a service the is low risk to test new technologies.

Question 4

How can Microservices improve system resilience?

Answer 4

Because when a well designed Microservice fails, it can degrade just part of the system, not everything.

Question 5

How can Microservices improve scaling?

Answer 5

Because with Microservices one can scale business capabilities independently.

Question 6

What is the difference between SOA and Microservices?

Answer 6

Deep down they want the same thing: to break monoliths and tackle the problems associated with then.
The difference is that SOA was pretty much driven by Middleware and Tools vendors, that wanted to push their solutions and make money. Microservices arose from real world architectural discussions and solutions for the monolith problem.

Question 7

How can Microservices improve software deployment?

Answer 7

Changes can be smaller and deploys can be smaller and safer.

Question 8

What are the 3 main challenges of Microservices over Monoliths?

Answer 8

Deployment coordination is harder
Distributed transactions may appear
Eventual consistency

(revise this after finishing book)

Question 9

What can help us frame our architectural decision making?

Answer 9

Defining the strategic goals and principles that help us achieve those goals.
Ex:
strategic goal: expand to 30 new countries
principle: every service must be born internationalized

Question 10

What are two good strategies for encouraging best architectural practices?

Answer 10

Using one real-world service as examplar.

Creating project templates for new services.

Question 11

What is Hystrix?

Answer 11

It is a circuit breaker library by Netflix that allow to wrap every request to an external service in order to prevent failures and latency to impact the whole system.

Question 12

In a simple way, what is technical governance?

Answer 12

Is about ensuring that the teams are following the technical vision.

Question 13

What is a bounded context?

Answer 13

A specific responsibility enforced by explicit boundaries (interface)

Question 14

Getting service boundaries wrong can be very costly. What is a strategy to mitigate that risk?

Answer 14

Starting with a monolith and define the bounded contexts inside the Monolith. Divide them with modules and once things stabilize, extract the module to a Microservice.

Question 15

Why you should think about capabilities and not just data to define a service/bounded context?

Answer 15

Because capabilites defines what a service do to other part of the business. Thinking only about data may lead to anemic crud based services.

Question 16

What is the problem of splitting service boundaries based on technical seams instead of business seams? (ex: split into front end service and backend service)

Answer 16

The service may not be cohesive. A change in one business capability may lead to change in multiple services.

Question 17

Why integration through database is bad?

Answer 17

Because we are exposing the implementation details of the service that owns the data. Also, it only shares data, not behavior.

Question 18

What is the difference between orchestrated and choreographed architectural styles for service integration

Answer 18

Orchestrated: a central entity/class controls the process of calling the multiple services that integrate the use case. Generally this is implemented via synchronous request/response.

Choreographed: each services is smarter to understand what it needs to do. Generally a central entity/class would send an event all the collaborating services would listen to the event and act accordingly. This can lead to more loosely coupled services, but the big picture of the process is implicit among the system.

Question 19

What is RPC?

Answer 19

Remote Procedure Call, is a technique of making a local call and having it execute on a remote service somewhere.
Ex: Java RMI, SOAP

Question 20

What is the problem of RPC technologies that hide too much of the remote details?

Answer 20

It can trick the developer that he is not making a remote call. The network is always unreliable, and we should deal with local and remote calls very differently

Question 21

What is HATEOS?

Answer 21

Hypermedia as the engine of application state.

It is a REST principle to reduce coupling between client and server, by using hypermedia links to abstract away the URIs the client need to do some action on a resource

Question 22

What is the downside of creating client libraries for your api?

Answer 22

It creates a stronger coupling between clients and your api than using, for example, hypermedia for api discovery.

Question 23

When communication over HTTP not recommended?

Answer 23

When you need very low latency and message size

Question 24

What is the drawback of asynchronous communication between services using messages?

Answer 24

The need of a broker Middleware increases complexity (more moving parts).
More complex error scenarios

Question 25

What are Reactive Extensions? (Rx)

Answer 25

A mechanism to compose the result of multiple calls together and run operations on them. You compose observables of remote calls.

Question 26

What is a "tolerant reader"?

Answer 26

It is a client that is tolerant to change in the data that is return to it by our services. Example is a client that uses a deserialization technology for our service response that is tolerant to the addition of new data or removal of unused data.

Question 27

What are the difficulties of maintaining two versions of a service running (an older endpoint and a newer endpoint)?

Answer 27

- two branches of the same codebase to be maintained - complexity of routing clients to the correct service - database model must support both versions of the codebase

Question 28

What is the problem of developing an UI that calls multiple backend services?

Answer 28

This communication can be chatty and consume too mutch internet bandwidth from the client.

Question 29

What is a pattern to solve the problem of chatty communication between UI and multiple backend services?

Answer 29

The backends for front-end pattern, whereby we create backend services that serve a specific front-end and aggregate calls for other backends

Question 30

What is the strangler pattern?

Answer 30

It is a pattern used to break an existing monolith into smaller Microservices. You begin by identifying domain concepts and creating proxy Microservices for those concepts, delegating to the monolith. When every domain concept is covered you can begin implementing the proxies with real implementation.

Question 31

What is a seam?

Answer 31

It is a portion of the code of a system that can be treated in isolation and worked on without impacting the rest of the codebase.

Question 32

What are the two phases in a two-phase commit algorithm?

Answer 32

1. Voting: participants vote if they can commit their local transaction 2. Commit: participants commit

Question 33

What is the name of the software that coordinated a distributed transaction?

Answer 33

Transaction Manager

Question 34

In terms of Microservices and boundary definition, what is the best thing to do when you encounter a business process or state that must be kept consistent at all times?

Answer 34

Avoid splitting the state or process in multiple services to maintain the possibility of using transactions

Question 35

If we have a reporting system for our service what is the downside of having it access the service database?

Answer 35

The database becomes a contract (Api) between them.

Question 36

What are the three ways that a reporting tool can consume data from multiple services?

Answer 36

1. Acessing each services reporting specific APIs 2. Create a data pump process on service side that sends new data to reporting tool/db 3. Use service events as data pump to the reporting tool

Question 37

What is a solution for mitigating the long time it takes for provisioning tools (ansible, chef), to set up all the software in the machine?

Answer 37

We can create virtual machine images with the necessary software already installed.

Question 38

What is the concept of immutable servers?

Answer 38

The ideia is to store all environment and service configuration of a host in source control. Service host is always created based on that configuration and manual change in the host is not allowed.