Small teams need infrastructure they can operate
Choose infrastructure with updates, incidents and recovery responsibilities in mind. A small portal illustrates when more orchestration can help.
A small team's infrastructure decision should start with operations: who can restore the application after a failed update, and what happens when a server fails? An additional platform earns its place when it makes these tasks more manageable under your actual conditions.
A portal with a limited structure
A hypothetical team of three developers maintains a customer portal, a database and a background worker that transfers orders. In this example, load is predictable. Automatic distribution across many machines is not yet a requirement. Recovery objectives have been agreed for outages; the team still needs to demonstrate that it can meet them.
The options are documented container operations on one server and a Kubernetes cluster operated by the team itself. This is an assessment of this example, rather than a universal ranking.
Operational question | One server running containers | A self-managed cluster |
|---|---|---|
What needs updating? | Operating system, container runtime and applications | Cluster components and their configuration as well |
What happens if a machine fails? | The portal goes down; replacement and recovery are required | Distribution can help if enough independent nodes and suitable configuration are available |
How is data recovered? | A tested backup and documented restoration | A separate plan for application data and its recovery is still needed |
The simpler option could suit this case if a demonstrated recovery procedure meets the operational objectives. A single server does not provide high availability. If that outage is unacceptable for the business, this version of the option is ruled out.
Automation shifts the work
Kubernetes can restart failed containers and distribute applications across available nodes. A process running again does not, however, prove that orders and attachments are complete. Data recovery and checks of business functions still need planning.
A managed service also needs a clear division of responsibilities: does the provider manage only the control plane, the nodes as well, or additional tasks? Application errors and the recoverability of your data do not automatically disappear.
When the decision changes
If the company already has a maintained platform with an operations team responsible for it, using it for the small portal can make sense. For the three developers, a separate server might then be the option requiring more special handling. Highly variable load or binding availability requirements may also demand a different architecture.
Before choosing, ask to see a failed release and a recovery scenario demonstrated in a test environment. Record who acts and which business function is checked afterwards. That evidence explains an infrastructure choice better than its component count.
Sources
Read more
All articles
Cloud & Operations with ASES
Hosting, deployment, maintenance and support