SOCRA : Cours du 4 juin

Lien de la note Hackmd

IT in company

Software needs

• Software
• Apache
• Monitoring
• Linux
• Network

Classic segregation it’s not what you think

| IT-Development | IT-Production | IT-System | |—– | ——- | ——— | | Develop software | Manage software in production | Manage hardware and OS+ | | | Assist users | Handles DPR | | Paid for new features | Paid for uptime | Paid for stability, security |

Assume failures

No software is guaranteed without bugs.

User error investigation

• Call support (Niveau 1)
• Use standard procedures (N1)
• Check logs (N2)
• Check configuration (N2)
• Call development : error in software (N3)

Help them using

• Error messages
• Cristal clear logs

Examples : Logs

Create new wish
Error returned

With that kind of log, impossible to know what happened.

2020-03-31 09:48:26.1539|INFO|0HLULB1T307F0|WishController|Create : New wish for #433 by #ce4ed122-3cfd-47cd-a8ea-9d0b9d4c55f4 : My Mobility
2020-03-31 09:48:26.1575|WARN|0HLULB1T307F0|WishController|HasOneNotNull - Property not found : LinkedID
2020-03-31 09:48:26.1577|WARN|0HLULB1T307F0|WishController|Error returned form HasOneNotNull : Property not found : LinkedId

With this kind of log, we see the error right away.

Logs

• Use different levels : Trace, Debug, Info, Warning, Error, Fatal
• Use and existing framework
• There is never too much logs

From dev to PROD

CI/CD Quesaco

• Continous : forming an unbroken whole; without interruption
  • integration : the coodination of processes
  • delivery : delivering letters, packages…. babies…
  • deployement : bringing resources into effective action

CI process goal

Validate code at each steps, for each developers Confirm software stability (a minimum) Create a delivery workflow Avoid human interaction

How to go live

1. Take a (development) ticket
2. Develop using BDD, TDD (and pair programming)
3. Push on VCS
4. Deploy on a test environment
5. Test your software
6. Everything is OK, then deploy on production

Eventually calculate metrics like coverage, technical debt…

A software

Delivery process

1. Get sources
2. Add version
3. Build
4. Run tests
5. Publish components

Environments

DEV, PROD, PRE-PROD, UAT

Create packages

1. Get sources
2. Add version
3. Build
4. Run tests
5. Create one package for each components

CI/DC Tools

• Shell script!
• Jenkins
• Travis CI
• Bamboo
• Teamcity

Monolyth

• AKA Single-tiered
• Self-contained
• Independant

Multi-tier

• Layer separation
• Flexible

3-tier

| Presentation | |–| | Logic | | Data |

Example

| Console : Get report| |–| | Business Logic : list of all sales and aggregate them | | Data : MySQL : access sales’ store |

SOA : Service Oriented Architecture

A service :

• represents a business activity
• is self contained
• is a black box for its consumers
• may consist of other underlying services

Communication

Local : IPC

|File|Shared Memory| |—-|—-| |Local - Bidirectional, one process at time for writing|Local - Bidirectional, one process at time for writing|

Signal	Socket
Local - Unidirectional, not used for data	Local or netowrk - bidirectional, synchrone

Pipe	Message queue
Local - Unidirectional	Local or network - bidirectional, asynchrone

Problem..

What if a I send a message and the service is down ? How to send a message which can be handled by multiples service, but only processed by one ? How to handle a client deconnection ? A master ?

Message queueing (aka MQ)

• Like a mailbox
• Increase TCP/IP features:
  • Durability - purging : is the message saved ? What is TTL ?
  • Delivery policy : should the message be delivered once more ?
  • Security : which applications should have the message ?
  • Notifications : the publisher can be notified when message is received

Implementations : Apache ActiveMQ, OMQ, RabbitMQ, JMS, …

API

CRUD

REST

• CRUD operations
• Stateless
• On HTTP/S

VERB	Request has body	Response has body	Safe	Indempotent	Cacheable
	GET	Optional	YES	YES	YES	YES
	POST	YES	YES	NO	NO	YES
	PUT	YES	YES	NO	YES	NO
	DELETE	NO	NO	NO	YES	NO

/api/users

|GET|POST|PUT|DELETE| |—|—-|—|——| |Retrieve all users|Create a new user|N/A|Delete all users|

/api/users/{id}

|GET|POST|PUT|DELETE| |—|—-|—|——| |Retrieve one user|Create a new user|Update|Delete one user|

OpenData - OpenAPI

• Free to use, reuse and redistribute
• Lots of companies : GAFAM, RATP, gouvernements…
• Lots of domains : genetic, chemical, geographic, justice…
• Swagger oh no it’s 2010 again

An SOA Software

To manage more users :

But there is still a problem..

The current state problem

What did my user? How can I guaranty it? Can I prove it?

So let’s think events!

Events afterall

|Item added| |——| |Item removed| |Payment received| |Order shipped|

1. Item added : book
2. Item added : DVD
3. Item removed : DVD
4. Payment received
5. Order shipped

Events sourcing

• Command : user intention => AddItemToCart
• Event : happened in the past => ItemAddedToCart
  • Immutable
• Event store

CQRS

Command and Query Responsibility Segregation

Cloud

How to run ?

• Software
• Libraries - dependencies
• Operating system
• Hardware

I am a user

I USE SERVICES!!!!

I’m a developer

• Why should I care about hardware failure ?
  • Infrastructure as a Service
• Why must I update the operating system ?
  • Platform as a service
• Is the (right) JVM installed ?
  • Platform as a service
• I know nothing about SQL Server, I just want a database !
  • Platform as a service / Software as a Service
• ITO are against me !

IaaS to SaaS

Me or nothing

|Entreprise|Cloud provider| |———-|————–| |On premise|Cloud|

Entreprise & cloud provider : hybrid

Lambda functions

• One endpoint, one function
• Focus on objective
• Access defined resources
• Call other functions if