Knowledge baseCloud & infra

We manage, you keep the repo: managing multiple Linux servers on your own Ansible and Terraform

18 November 20266 min read

Outsource server management and you usually get the provider's tooling with it. The configuration moves into their system, their panel or their repository, and your own Ansible roles or Terraform code fall out of use. The misconception is that this is how it has to be: that you can only have multiple Linux servers managed for you if you hand over the automation. We work the other way round. If you already have a repo describing your infrastructure, we manage inside it, the way a good colleague would. In terms of the layers of managed hosting, this is Managed+.

Merge requests, not silent changes

Every change we make arrives as a merge request in your repo. A patch round, a new firewall rule, a PostgreSQL setting: it sits in a diff, with a description of why. With CODEOWNERS you decide which paths need your review and which we may merge on our own. The repo is then not a copy of what runs, but the place where it gets decided.

There is a side effect that customers ask for more and more: an audit trail you do not have to keep separately. Who changed what and when, and who approved it, is in your repo history and your pipeline logs. A customer that falls under NIS2 and has to account for its suppliers gets more out of that than out of a monthly report; see the Dutch Cyber Security Act.

The drift report: dry run and plan in the pipeline

Drift is the difference between what is in the repo and what runs on the servers. It always happens: someone logs in during an incident and tweaks something, a package update drops its own config file. You cannot prevent it; you can make it visible.

For that, your pipeline periodically runs a check that changes nothing:

  • Ansible — ansible-playbook --check --diff against the whole inventory. Every task that would report "changed" is drift, with the diff attached.
  • Terraform or OpenTofu — plan shows which resources differ from the code. A plan that wants to do nothing is an environment that matches.

Two caveats from practice. Not every Ansible module supports check mode well, and command or shell tasks are skipped in check mode by default. A clean report is only as reliable as your roles; idempotent roles without loose shell commands make it sharper. And drift is not automatically wrong. Sometimes the manual change is the right one and belongs in the code. Root stays with you, so the rule is not "nobody touches the server" but "whatever happens outside the repo gets reported, and either lands in the repo or gets reverted".

Access and secrets: your keys, our own identity

A provider that works with your personal keys cannot be cleanly offboarded. So:

  • Own deploy key or runner — we push and run with an identity that is ours alone, with exactly the rights the work needs.
  • Own accounts on the servers — no shared root password; what we do is traceable to us.
  • Secrets in your vault — whether that is encrypted files in the repo or a separate secrets manager, the key stays with you and the automation fetches what it needs.

The result is that parting ways is not a migration. You revoke our key and accounts, and everything we built is already in your repo.

In practice: 31 servers on the customer's own model

For a SaaS company we manage a fleet of 31 servers in four roles this way: database, application, control and big data. Not on our standard package, but on the automation model the team had built itself. We work within that model instead of replacing it with ours. Our engineers also contribute to the application and debug it when something goes wrong. That is what makes the difference during an incident: whoever runs the servers also knows the code, and does not have to wait for someone to explain the application. How the application side becomes visible is covered in monitoring versus error tracking; how changes go through a pipeline in CI/CD without buzzwords.

Management in your repo

Do you have your own Ansible or Terraform repo and want someone to work in it rather than replace it? See managed servers. If you are working on the application itself, software development fits alongside, and for watching your network there is NOC as a service. Curious how that would work on your repo? Get in touch.

Frequently asked questions

Frequently asked questions

Do we have to adapt our Ansible roles or Terraform code to your way of working?

No. We work on the model you built: your roles, your inventory, your modules. If we see something that could be better, we propose it as a merge request, like a colleague would. Whether it goes in is your call.

How do you get access without us sharing our keys?

With our own deploy key or our own CI runner, and our own accounts on the servers. Secrets stay in your vault; we get read access to what the automation needs. If we stop, you revoke that access and the handover is done.

What if someone with root changes something outside the repo?

Then the drift report shows it. A dry run with diff, or a plan, shows the difference between the repo and reality. After that it is a choice: take the change into the code, or let the next run revert it.

Answer not found?

Ask an engineer directly — we usually respond within one business day.