Dev Ops North East March 2023
Self-Service Cloud Infrastructure
New Development, Development Teams to New Private Managed Cloud with an Infra Team. Client has Existing Services on their existing Infrastructure and needs to be moved into a new Private Managed Cloud and this needs to be done quickly and dev team needs access to the cloud infrastructure.
Security Sizes, Enable, Efficiency, Policies and Control.
Need to have VMs to have certain level of security. Each of the teams may be working on larger projects and need to have the resources available only to people on that project and not accessible to people outside of this project. Want to have the credentials and permissions that configure the hardware to be kept away for the developers and the rights to the machines should only be kept with the infra team but the developers should be able to create the VMs as needed. The efficiency of this process needs to be high with 90 applications needing to be migrated quickly with a couple of services taking a few months. Need to have standardised virtual machines within a certain range. Organisation has many policies such as naming policies and how resources are tagged and need to have agents to monitor services and to allow monitoring on things like performance. Infrastructure team should be able to give development teams access to resources, provisioning a VM can be a blocker where this would have to have been a manual process to get something provisioned which would take days so the infra team should be able to allow the dev team to easily create VMs. Infra team retains control and should not need to give developers master passwords and also be able to control the estate itself.
SSI (Self-Service Infrastructure) which has YAML, DevOps Pipeline and IaaS and is owned by the infra team and allow the development team to create whatever resources they need on the infrastructure. The developers create a YAML file to describe their requirements and then after this it is submitted and pushed to a repo to dev ops which goes through several pipelines to go through terraform files with the infrastructure using VMWare and vSphere-based environment. Once these artefacts are processed will have the infrastructure that is needed created. YAML files contain details about the requirements such as networking, CPUs and cores needed, memory required along with the database such as MongoDB with details needed to access this.
Portfolio repository with config files and scripts which is owned by the developers which will trigger the SSI template process which will configure what is needed using Terraform and other templates which then create the artifacts to the create the infrastructure needed. Applications that are provisioned will have secrets that need to be saved these will be placed on a vault namespace. Services can be run such as MongoDB, MariaDB, RabbitMQ and others.
Portfolio pipeline for the developer YAML files will trigger orchestrator and create artifacts to create the services with the services pipeline will use the artifacts to create the infrastructure along with creating results artifacts which will be placed back on the Portfolio pipeline for the developers and can be used to access their resources. Developers don't need to destroy and start again as can modify anything such as configuration as needed. It is also possible to take advantage of what other developers have created so don't have to create everything from scratch.
Impact to the User
Simplicity, maintain, Meaningful Alerts, Speed, Vault, Collaboration, Standards, Sight, Choices and Integration
With the SSI setup it is easy to customise environment as needed, the hardware itself is taken away from the developers but also the complexity is too so it is possible to quickly iterate through changes if needed. It is possible to migrate around five services at a time so makes the migration process even quicker. If there is any monitoring software this can be placed onto the server automatically and request that infra can gather this information and this will be configured with naming conventions so the information can be obtained more easily. Maintaining system is a lot easier as there could be various changes in the implementation and these can be versioned and can run changes with more up-to-date changes to keep up with updates or changes to the infrastructure. If a particular error message is returned so need to filter error messages that are being thrown out by the system and have anything that needs to be actionable by the developers but have a more reasonable message to allow developers to debug their own processes. Standards, choices and collaboration requires projects to follow certain standards and protocols to get integrations working and projects need to confirm with the layout needed by SSI.
Remote Access, Load Balancing, Continuous Integration. Connection is done via jump boxes at the moment and is not the most convenient way so are working on group policy and active directory way of accessing resources. Will be beneficial to have load balancing for many services such as message queues. Various projects are being worked on and developers need to be able to access those services easily for continuous integration of projects and need to work to a certain standard so things can be integrated.
What is GitOps?
GitOps is a continue delivery approach that emphasises using Git as single source of truth for infrastructure and application configurations.
GitOps is a process for automating infrastructure using best practices of software infrastructure using workflow based on Git principals. Ensures infrastructure as code, easy to roll back changes, consistency and standardisation, enhances security and compliance, improves stability and increased productivity. Starts with creating a Branch then that goes through Pull Requests and Merge Requests and can have changes from multiple collaborators and then can be approved by a designated code owner before being merged. This then goes through continuous integration pipelines with an approval process and then go through pipeline process to deploy to cloud such as AWS, Kubernetes, Microsoft Azure or Google Cloud. Before GitOps things may not have been fuly validated and potentially not being checked correctly with things happening in a more haphazard way.
GitOps can help improve things as infrastructure is related to what has been checked into source control and the desired state will be in sync with the environments and can also ensure things have gone through proper scrutiny, have been properly reviewed by dedicated code owners. Once things have been merged and things have been deployed to lower environments integration tests can be performed before being pushed to production. For containerised applications the applications will be put into a container and this can be deployed with Kubernetes and can have manifest files that can also be stored in source control such as being able to change version numbers where needed. Deployments are done with semantic versioning and can easily roll back to previous versions if there are any issues and the same code is always being deployed to all environments to ensure that what is expected to be deployed is deployed, along with being checked for any security issues before being promoted and any issues and errors can be identified early before being put into production.
Infrastructure as Code (IaC) along with Configuration as Code, Network as Code and Security as Code then this can be used with CI/CD pipelines with workflow from Code, Build, Test, Release, Deploy, Operate, Monitor, Elaborate then back to Code along with having Merge Requests and Approvals.
Push and Pull Workflows
Push-based approach where CI/CD tools push changes to environment, where applying GitOps is more consistent with the approach used for application deployment and these targets are not limited to Kubernetes.
Pull-based approach where agent in cluster pulls changes wherever there is a deviation from the desired configuration. In the pull-based approach deployment targets are limited to Kubernetes and an agent needs to be installed in each Kubernetes cluster.
Pillars and Principles
Pillars - Single source of truth from Git, treat everything as code, operations performed though Git workflows.
Principles - Declarative model, versioned and immutable, pulled automatically where agents automatically pull the desired state and continuously reconciled where agent keeps applying the desired stage.
Access control for Git, GPG signed commit, block sensitive data, enforce branching policy, dedicated code approvers, linear commit history, static code analysis for security checks, keeping Git and CI/CD tools up-to-date, rotate SSH keys and access tokens and backup git repositories.