START HERE

What you will learn

  • Connect domain, host and web service responsibilities without confusing them.
  • Separate staging and production access and configuration.
  • Define the evidence needed to operate and recover a small website.

Before you start: A cloud account and domain controlled by the business, plus an authorised administrator. Provisioning changes require budget and account approval.

Reference edition: Fictional website walkthrough; providers are examples, not required services. Sources checked 2026-11-13.

Decide whether you need a virtual server

An information-led site may fit a managed static host or application platform, which can remove operating-system maintenance. A virtual server is useful when the architecture needs that level of control, but it also creates responsibility for patching, access, storage, monitoring and recovery.

Use AWS Lightsail as a concrete virtual-server example here. The workflow applies to comparable hosting. Choose a region, capacity and recovery approach against the business’s data and operational needs. Avoid selecting a bundle by an introductory price without understanding the resources and maintenance it requires.

Prepare the environment record

Name the account owner, administrators, domain owner, region, host, expected public services, deployment identity and recovery contact. Record where configuration and backups live, and which person can approve a production change. Keep credentials in a controlled secret store.

Separate staging from production by identity, configuration, endpoints and data. Separate hosts reduce some shared-resource risks; separate paths on one host can be a small-project compromise but still share an operating system and failure domain. Document the compromise rather than treating two hostnames as strong isolation.

Connect a stable address and the right domain records

In the Lightsail example, attach a static IPv4 address so a stop/start does not require changing the domain’s A record. Point the intended hostname to the serving address, verify DNS from outside the account and configure HTTPS at the web server or managed edge. Publish an IPv6 AAAA record only if that route and its firewall are actually configured.

Check both the canonical hostname and redirects. A DNS record reaching a host does not prove the right site or certificate is being served. Confirm the expected environment and release marker over public HTTPS. Keep domain and DNS control available to the business if the host changes later.

Constrain administration and serving separately

Allow the public web traffic the site requires. Restrict administrative access to an approved route such as a VPN, bastion or limited-source SSH path. Configure the operating-system boundary as well as the cloud firewall. Lightsail’s public firewall rules are inbound rules; they do not by themselves restrict outbound access or replace private-network controls.

Use a dedicated deployment identity with access to release locations, not a general administrator login for every pipeline run. Verify SSH host identity using a fingerprint obtained through a trusted route. If hosted CI cannot reach a restricted host, design an approved runner or network path instead of solving the problem by exposing unrestricted administration.

Serve versioned releases and prove recovery

Place each release in an immutable directory or equivalent versioned deployment unit. Switch the serving reference only after checks pass, retain the previous version and record which revision is current. A Node development server is not the production serving plan for a static build.

Back up what cannot be reconstructed from source: configuration, approved persistent data and the recovery information required to restore service. Snapshots are useful but should be restore-tested, and should not become the only copy of the application source. Exercise recovery on a safe target, confirm certificates and DNS assumptions, and assign an owner to monitoring and patching.

What stakeholders should be able to ask for

BoundaryTechnical actionEvidence
DomainSet the intended DNS records and canonical redirectsCorrect public hostname and HTTPS response
HostConfigure capacity, patching and service startupDocumented environment and restart check
AccessLimit administrative and deployment authorityNamed identities and verified connection route
ReleaseServe a versioned artifactPublic revision marker and rollback record
RecoveryRestore configuration and required dataRecorded restore test and accountable owner

Step-by-step process

  1. Choose the hosting responsibility model

    Compare a managed host with a VM and record who will maintain whichever option you choose.

  2. Prepare a separate staging target

    Use non-production data, distinct configuration and an explicit access/indexing policy.

  3. Create and record the serving address

    In Lightsail, attach a static IP; then configure only the domain records you can support and test.

  4. Configure HTTPS and restricted administration

    Verify the certificate and the SSH or managed connection identity. Keep the deployment identity narrower than the administrator identity.

  5. Install the versioned release path

    Test startup, public health checks and rollback before calling the environment ready.

  6. Run a recovery rehearsal

    Restore to a safe target and document the missing steps before a real outage exposes them.

FOLLOW ONE EXAMPLE

Worked scenario

A server restart should not change the website’s address

A fictional business runs a small information site on a VM. The administrator plans an operating-system maintenance restart.

  1. They confirm the stable address, DNS, service startup configuration and current release before maintenance.
  2. The administrator tests the same setup on staging and retains the previous working release and recovery information.
  3. After the production restart, they check HTTPS from outside the host and verify the release marker and enquiry path.
  4. If a service fails to start, the runbook identifies its owner and recovery procedure instead of relying on the person who originally installed it.

What this shows: The address, serving process and recovery route are managed independently, making maintenance less dependent on undocumented setup.

The process at a glance

Visitors reach a serving address through DNS and HTTPS. Deployment, administration and recovery use separate controlled paths with different responsibilities.
ProcessDecisionHuman checkpoint
Read Mermaid diagram source
flowchart TD
 V[Visitor hostname] --> D[DNS record]
 D --> I[Stable serving address]
 I --> H[HTTPS web service]
 H --> R[Versioned release]
 P[Checked deployment identity] --> R
 A[Restricted administrator route] --> M[Host maintenance]
 B[Recovery records and backups] --> T[Restore rehearsal]
 T --> M

What you should have at the end

  • An environment record covering access, service ownership and public verification.
  • A release and recovery path tested before depending on the server.

Common mistakes and how to avoid them

Treating a DNS change as a completed deployment

Verify the actual site, certificate, environment and revision over public HTTPS.

Giving the pipeline a full administrator key

Use the narrow authority required for release and retain a separate maintenance route.

Backing up without restoring

A useful recovery claim needs a rehearsal and the information needed to reproduce it.

APPLY THE IDEA

Try it yourself

Your task: Draw the domain-to-release path for a site you own and identify every separate identity involved.

What to produce: An environment record and a recovery rehearsal checklist with named owners.

Check your understanding: Can the business recover access and move the site if the original implementer is unavailable?

Key terms in plain English

DNS
The system that maps names such as a website hostname to records used to find services.
Static IP
An address retained across supported instance lifecycle changes.
Failure domain
A set of resources that can be affected by one underlying failure.

Sources and further reading