๐ ๐ ๐๐ถ๐ฟ๐๐ ๐ง๐ถ๐บ๐ฒ ๐ช๐ถ๐๐ต ๐๐๐๐ฟ๐ฒ ๐๐ฒ๐๐ข๐ฝ๐
Last night, close to midnight, I finally saw those green ticks in Azure DevOps.
After several failed deployments, SSH troubleshooting, and moments of โI am tired oooo,โ my pipeline finally succeeded!
This was my first hands-on experience with Azure DevOps, and it helped me connect the dots between infrastructure, server configuration, source control, and automated application delivery.
What is Azure DevOps?
Azure DevOps is Microsoftโs platform for planning work, managing source code, and automating software delivery. Its services include Azure Boards, Azure Repos, Azure Pipelines, Azure Test Plans, and Azure Artifacts. For this project, I focused on Azure Repos for version control and Azure Pipelines for deployment automation.[learn.microsoft]
One interesting lesson: using Azure DevOps does not mean your application must run on Azure. Azure Pipelines supports deployment to different environments, including AWS, which was my deployment target.
What I built
I connected an Azure DevOps self-hosted agent to a YAML deployment pipeline that delivered a personalized static website to an AWS EC2 instance running Nginx.
The workflow:
Azure Repos โ Azure Pipelines โ Self-hosted agent โ SSH โ AWS EC2 โ Nginx โ Website in the browser.
Rather than treating these as separate tools, I began to understand how each one contributes to the delivery process.
๐น Self-hosted agent foundation
I used a self-hosted Azure Pipelines agent running on an Ubuntu VM, organized under my SelfHostedPool agent pool.
That agent provided the execution environment for my pipeline jobs. This helped me understand that writing YAML is only one part of the process, you also need an available machine to execute the work.
๐น Source control with Azure Repos
I imported the provided static website repository into Azure Repos, personalized the webpage, and stored my deployment configuration in azure-pipelines.yml.
My application code and deployment workflow were now version-controlled together.
๐น Infrastructure with Terraform
I used Terraform to provision the AWS EC2 deployment environment, including the Ubuntu instance and supporting network access.
This separated infrastructure provisioning from application deployment: Terraform prepared the destination rather than transferring website files on every pipeline run.
Get Engineeragelessโs stories inย yourย inbox
Join Medium for free to get updates fromย thisย writer.
๐น Server configuration with Ansible
I used Ansible to prepare the target server by configuring Nginx, the website directory, SSH access, and the permissions needed for deployment.
An important detail was ensuring that the pipelineโs SSH user could update /var/www/html while Nginx could read and serve the files.
๐น Deployment automation with Azure Pipelines
The deployment checks included confirming that index.html existed and checking the local Nginx HTTP response.
The biggest blocker was SSH authentication.
The pipeline repeatedly reported:
โFailed to connect to remote machine. Verify the SSH service connection details.โ
That message alone did not tell me enough.
With temporary SSH debug logging enabled on the target server, I could see that the connection reached the machine and completed key exchange. The failure happened afterward: the client attempted password authentication while the server allowed public-key authentication.
That changed the troubleshooting direction.
Instead of opening more ports or enabling password authentication, I corrected the SSH service connection by uploading the intended private-key file into its Private Key field and saving the connection.
The next pipeline run succeeded. โ Logs turn guessing into diagnosis.
๐ ๐ ๐ฏ๐ถ๐ด๐ด๐ฒ๐๐ ๐๐ฎ๐ธ๐ฒ๐ฎ๐๐ฎ๐
Terraform, Ansible, and Azure Pipelines have different responsibilities, and work better when those responsibilities are clear.
A connection error is not automatically a network problem. Identifying the exact failure stage matters.
Credential configuration deserves careful attention. A masked field does not prove the correct credential is stored.
Deployment is not finished just because files were copied. Verification needs to be part of the workflow.
Security belongs in a learning project too. I used SSH private-key authentication and kept secrets out of the repository and pipeline YAML.
Cleaning up matters. After the lab, I destroyed the provisioned infrastructure rather than leaving the environment running unnecessarily.
This was not a smooth, straight-line exercise. It involved failed runs, troubleshooting, correcting assumptions, and trying again.
I finished with a clearer understanding of how code moves from a repository to a running website, and how to investigate when that journey breaks.
My first time with Azure DevOps was challenging, tiring, and absolutely worth it.
๐๐๐ฏ๐ ๐๐จ๐ฎ ๐ญ๐ซ๐ข๐๐ ๐๐ณ๐ฎ๐ซ๐ ๐๐๐ฏ๐๐ฉ๐ฌ?
P.S. This post is part of the DevOps Micro Internship (DMI) with Agentic AI โ Cohort 3 โ by Pravin Mishra. My graded progress is public: https://dmi.pravinmishra.com/s/EAgeless.html ยท Start your DevOps journey: https://dmi.pravinmishra.com/?utm_source=student&utm_medium=ps-blog&utm_campaign=cohort3
#AzureDevOps #AzurePipelines #AWS #EC2 #Terraform #Ansible #CICD #InfrastructureAsCode #DevOps #LearningInPublic








