Loading...
Publication
AutomatizaĆ§Ć£o do processo de deployment
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 10.27 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Num contexto organizacional cada vez mais exigente, a automatizaĆ§Ć£o do processo de
deployment tornou-se uma componente crĆtica na melhoria da eficiĆŖncia, fiabilidade e
previsibilidade dos ciclos de entrega de software. Esta dissertaĆ§Ć£o teve como objetivo conceber,
desenvolver e validar uma soluĆ§Ć£o de automatizaĆ§Ć£o do processo de deployment, adaptada Ć
realidade de uma organizaĆ§Ć£o tecnolĆ³gica de grande escala, com mĆŗltiplas equipas, ambientes
e restriƧƵes operacionais.
A investigaĆ§Ć£o seguiu a metodologia Design Science Research (DSR), articulando uma fase de
diagnĆ³stico com iteraƧƵes de conceĆ§Ć£o, implementaĆ§Ć£o e validaĆ§Ć£o. O trabalho inicia-se com
uma revisĆ£o do estado da arte, abordando prĆ”ticas consolidadas em DevOps e CI/CD, como
testes de sanidade automatizados, rollbacks condicionais, monitorizaĆ§Ć£o contĆnua e reporting
automĆ”tico. SĆ£o tambĆ©m analisadas ferramentas como Jenkins, GoCD, Splunk, Sensu e Grafana,
avaliando a sua aplicabilidade no contexto da pipeline da organizaĆ§Ć£o.
A soluĆ§Ć£o proposta materializou-se no desenvolvimento de uma aplicaĆ§Ć£o autĆ³noma ā o
Release Assistant ā que centraliza e automatiza etapas crĆticas do deployment, como a gestĆ£o
de trĆ”fego, validaƧƵes pĆ³s-deployment (sanity checks, mĆ©tricas, alarmes) e produĆ§Ć£o de
relatĆ³rios. A arquitetura modular permite integraĆ§Ć£o com sistemas existentes e reutilizaĆ§Ć£o em
mĆŗltiplos serviƧos e ambientes.
Os resultados demonstram uma reduĆ§Ć£o objetiva no esforƧo manual e no tempo de execuĆ§Ć£o.
Com base em mediƧƵes, registou-se uma reduĆ§Ć£o mĆ©dia de 15 minutos por execuĆ§Ć£o,
equivalente a cerca de 41% da duraĆ§Ć£o dos deployments, refletindo-se diretamente na
eficiĆŖncia operacional. A soluĆ§Ć£o aumentou ainda a fiabilidade das validaƧƵes, assegurando
maior consistĆŖncia nos testes e deteĆ§Ć£o precoce de falhas. Em termos de contributo, o trabalho
apresenta uma abordagem replicĆ”vel de automatizaĆ§Ć£o de validaƧƵes integrada em pipelines
CI/CD, validada num ambiente de produĆ§Ć£o real e sustentada por evidĆŖncia quantitativa e
feedback qualitativo.
Apesar dos resultados positivos, reconhece-se que a soluĆ§Ć£o depende da estrutura atual das
pipelines e da existĆŖncia de sistemas de monitorizaĆ§Ć£o acessĆveis por API. A adoĆ§Ć£o noutros
contextos poderƔ exigir adaptaƧƵes. Ainda assim, o trabalho demonstra o potencial da
automatizaĆ§Ć£o como vetor de transformaĆ§Ć£o digital orientado Ć melhoria contĆnua dos
sistemas e processos operacionais.
In an increasingly demanding organizational context, deployment automation has become a critical component in improving the efficiency, reliability, and predictability of software delivery cycles. This dissertation aimed to design, develop, and validate a deployment automation solution tailored to the reality of a large-scale technology organization with multiple teams, environments, and operational constraints. The research followed the Design Science Research (DSR) methodology, combining a diagnostic phase with iterative cycles of design, implementation, and validation. The work begins with a state-of-the-art review, covering established DevOps and CI/CD practices such as automated sanity checks, conditional rollbacks, continuous monitoring, and automated reporting. Tools such as Jenkins, GoCD, Splunk, Sensu, and Grafana are analyzed, assessing their applicability within the organizationās pipeline. The proposed solution took the form of a standalone application ā the Release Assistant ā which centralizes and automates critical deployment stages, including traffic management, post-deployment validations (sanity checks, metrics, alarms), and report generation. Its modular architecture enables seamless integration with existing orchestration systems and reuse across multiple services and environments. The results demonstrate a measurable reduction in manual effort and execution time. Based on recorded measurements, an average reduction of 15 minutes per deployment was achieved ā approximately 41% of the total execution time ā directly enhancing operational efficiency. Additionally, the solution improved validation reliability by ensuring greater consistency in postdeployment testing and enabling early fault detection. As a contribution, this work presents a replicable approach to automating validation processes within CI/CD pipelines, validated in a real production environment and supported by quantitative evidence and qualitative feedback from technical users. Despite the positive results, the solution has limitations. Its effectiveness depends on the current pipeline structure and the availability of monitoring systems with accessible APIs. Adoption in other contexts may require architectural or integration model adaptations. Nonetheless, the project demonstrates the potential of automation as a driver of digita
In an increasingly demanding organizational context, deployment automation has become a critical component in improving the efficiency, reliability, and predictability of software delivery cycles. This dissertation aimed to design, develop, and validate a deployment automation solution tailored to the reality of a large-scale technology organization with multiple teams, environments, and operational constraints. The research followed the Design Science Research (DSR) methodology, combining a diagnostic phase with iterative cycles of design, implementation, and validation. The work begins with a state-of-the-art review, covering established DevOps and CI/CD practices such as automated sanity checks, conditional rollbacks, continuous monitoring, and automated reporting. Tools such as Jenkins, GoCD, Splunk, Sensu, and Grafana are analyzed, assessing their applicability within the organizationās pipeline. The proposed solution took the form of a standalone application ā the Release Assistant ā which centralizes and automates critical deployment stages, including traffic management, post-deployment validations (sanity checks, metrics, alarms), and report generation. Its modular architecture enables seamless integration with existing orchestration systems and reuse across multiple services and environments. The results demonstrate a measurable reduction in manual effort and execution time. Based on recorded measurements, an average reduction of 15 minutes per deployment was achieved ā approximately 41% of the total execution time ā directly enhancing operational efficiency. Additionally, the solution improved validation reliability by ensuring greater consistency in postdeployment testing and enabling early fault detection. As a contribution, this work presents a replicable approach to automating validation processes within CI/CD pipelines, validated in a real production environment and supported by quantitative evidence and qualitative feedback from technical users. Despite the positive results, the solution has limitations. Its effectiveness depends on the current pipeline structure and the availability of monitoring systems with accessible APIs. Adoption in other contexts may require architectural or integration model adaptations. Nonetheless, the project demonstrates the potential of automation as a driver of digita
Description
Keywords
DEvOps CI/CD AutomatizaĆ§Ć£o Deployment Sanity checks MonitorizaĆ§Ć£o OrquestraĆ§Ć£o Observabilidade Automation Monitoring Orchestration Observability