Modernization of applications
Experiences in the modernization of solutions
Creted by Alejandro Escanero Blanco / Twitter: @aescanero Documentatión and demo in https://github.com/aescanero/dockerevents/opensouthcode2018 Presentation in Disasterproject
Introduction
What is a Legacy system?
It's a brown dispacher. That has been there for a long time and nobody knows anything about
No documentation
No source code
There is source code but it does not compile
And worst, there is source and compile, but nobody understands why
Make documentation
Cross fingers
Pray
Mad Max
First problem: Time is money
Fuente: Legacy SystemNew needs over the service
Comercial: Our product in the customer's infrastructure is really good
Manager: Let's put it in another client!
Old model:
New model:
Systems vs. Development
Systems and Development against the world
Problem (In case you haven't noticed this): Time is money
First step: Project evaluation
Analyze infrastructure:
How the application works:
The Database and its model:
Physical Servers
Two layers or multilayer applications
(Oracle) Database with business logic in it
Reality (red line) is a long list of intermediate layers: 
How far do we really want to go?
The business model
On Premise or Cloud
Client infrastructure
Prices and fixed costs
Client infrastructure
Control of data
Worst application support
Simpler but slower implementation
Elasticity
Prices by service
Third party infrastructure
Untrust in security
Better application support
Complex but faster implementation
Other requirements
Time
Cost
Licenses
For yesterday
Free
What?
Second step: Life cycle to modernize a application
Analyze the current situation and look for improvements. It is practically a pre-sale task.
Design improvements that meet the requirements. It is an evangelistic work.
Implement. It includes the tests, it is a communication task.
Operate and collecting opinions.
A continuous improvement job is also continuous communication.
All clients support virtualization
You can't waste time on new physical servers
The first step is: Can we migrate the environment to virtual?
Even very old environments can be virtualized (RHEL or Debian + 10 years)
The second is to migrate and create templates (OVF)
Source: Open Virtualization FormatClient's networks are a mystery
And if you need to mount a cluster but you do not have a balancer or Fence?
A widespread OpenSource solution is Pfsense, which has firewall capabilities, balancer with HaProxy and multiples tools
Configuring a cluster with PfSense is easy and fast and does not require special dependencies.
Automatic deployments
We can schedule deployments or updates of the application
But in an environment with a lot of machines we are interested in move to a configuration manager
Ansible is perfectly valid for launching updates, tests and end
This make simple the rollback job
Source: AnsibleSeparate the application from virtual machines
With the MVs does not sufficient to reach the customer, we must package the applications
The containers isolate us the application of the SO
To improve the deployment of the previous point
Advantages of orchestrating containers
Reboot failed containers
Horizontal scaling
DNS Autodiscover and load balancing
Secrets and configuration
Persistent storage
Multiple network solutions
Supported by all the big Cloud platforms
Can be federated
Monitoring and alerts
Zabbix is an easy-to-use monitoring system
It is trending and actually is more used than Nagios
Integrating it into a client is a complex process
Easily integrates with Graphana (better reports)
Source: Zabbix