VolttiDiscuss your pilot

What happens to your AI environment if hardware fails?

If the hardware behind your AI environment fails, we restore the environment from the latest encrypted backup onto replacement hardware. That backup sits at a separate location in your own network and never on the node itself. Your projects, settings, knowledge base, and the agent's memory do not disappear with the computer.

Hardware breaks at some point. The real question is whether you are ready for it. Below you can read what happens step by step and how AI backup and recovery works at Voltti. You will also see why you should test a backup and what you can prepare yourself.

Blue loop with arrows around a shield above a row of servers
Backup and recovery: prepared for hardware failure.

What happens when the hardware fails?

A failed drive, a broken power supply, or a computer that no longer boots: the cause differs, but recovery follows the same five steps every time.

IT teams describe recovery with two terms. The recovery point objective (RPO) is how much recent work you can afford to lose. You set it by choosing how often a backup is made. The recovery time objective (RTO) is how long you can do without the environment. It depends mainly on how quickly replacement hardware is available and how large the backup is. That is why we do not quote a fixed recovery time here. What you can expect from us is part of your support agreement.

  1. 01

    The failure is identified

    Hermes stops responding, or the computer does not boot. You notice it yourself, or we see it through the monitoring that is part of support. First we check whether the hardware is the cause. A stuck update or a full drive can usually be fixed without restoring anything. In that case your current environment stays in place.

  2. 02

    Replacement hardware is set up

    If the computer is broken, it needs a replacement. That can be hardware you already have or choose yourself, or a computer Voltti supplies as an option. Not sure what you need? Read our overview of what computer you really need for local AI and our article on when you need a Linux computer for an AI agent. If the language model runs at an external provider instead of locally, simpler hardware is often enough.

  3. 03

    The latest backup is restored

    We restore the latest encrypted backup from the separate location in your network. It contains your full Hermes environment: projects, settings, knowledge base, the agent's memory, skills, and plugins. Anything that changed after that is not included. How much work you lose therefore depends on how often a backup is made. You agree on that during installation. Code and documents in your own systems, such as a git repository or a shared drive, are not affected by a failed AI computer.

  4. 04

    We check that everything is complete

    After the restore, we go through everything. Are the projects and settings complete? Is the knowledge base still there? Does Hermes work as before? We also check the connections to your other systems, because they need access again on the new hardware.

  5. 05

    You get back to work

    Only when that check passes do we put the environment back into use. We tell you what went wrong and whether anything in the setup should change. Think of backing up more often, or keeping an extra copy somewhere else.

How do backups work at Voltti?

Backups run automatically and go, encrypted, to a separate location in your network, never to the node itself. That means:

  • The backup is not on the same system as the environment

  • The backup is encrypted, so even if someone gets into the storage, they cannot read the data

  • The backup includes the full Hermes environment: projects, settings, knowledge base, the agent's memory, skills, and plugins

  • You choose how many versions are kept during installation, just like how often a backup is made

The Dutch National Cyber Security Centre (NCSC) recommends that businesses follow the 3-2-1 backup rule (page in Dutch). Keep at least three versions of your important data (the original and two backups), on two types of storage, with one copy at another location. The backup of your AI environment is one of those versions. If a fire or theft could hit the computer and the NAS at the same time, an extra copy outside your office makes sense. We then discuss how that fits your setup. For more on that trade-off, read local AI vs cloud AI.

Why test a backup?

An untested backup is not a backup you can rely on. The Centre for Cybersecurity Belgium (CCB) says the same in its guide to managing backups. Check regularly whether your copies can actually be restored and read. Storage media can get damaged, and sometimes a sync does not run. That way you know in advance that the five steps above work, not on the day the hardware gives out.

On our own Hermes installation, after setting up the backup, we checked that we could open it and that all the data we wanted to keep was in it. It was. We have also moved our own configuration from a desktop computer to a server. That shows a Hermes environment is not tied to one computer.

What can you prepare yourself?

Make sure the backup location is reliable: a NAS, an external drive, or another server in your network. Agree with Voltti on how often a backup is made. That depends on how much changes per day. And look beyond backups alone: our practical checklist for AI and business data also covers access, permissions, and human oversight.

AI backup and recovery is one part of a local AI setup. Hardware is not the only thing that can fail: a cloud model can also stop in the middle of a task. Our article on a local model as a fallback explains what to do then. Read more about local AI for businesses or about security and privacy.

Portrait of Seppe Gadeyne

Seppe Gadeyne

    • Seppe Gadeyne is a developer and the owner of Voltti. He sets up AI agents for companies in Belgium and the Netherlands.

    Updated on

    Frequently asked questions

    01How often are backups made?

    That depends on how you want your backups set up. You agree on it with Voltti, or you handle the backup setup yourself with your own IT team. The backup goes, encrypted, to a separate location in your own network. If a lot changes per day, for example because the agent works on projects all day, it is better to back up more often. The more often you back up, the less work you lose if the hardware fails.

    02Where are the backups stored?

    At a separate location in your own network, never on the node itself. That could be a NAS, an external drive, or another server. This way a failed computer does not take the backup with it. If you also want a copy somewhere else, as the 3-2-1 rule recommends, we discuss how that fits your setup.

    03What does the backup contain?

    Everything needed to restore your AI environment: projects, settings, the knowledge base, the agent's memory, skills, and plugins. You choose how many versions are kept during installation. Code and documents that already live in your own systems, such as a git repository or a shared drive, stay there. A failed AI computer does not affect them.

    04How long does recovery take after a failure?

    It depends mainly on how quickly replacement hardware is available and how large the backup is. If a computer is already on hand, it goes faster than when new hardware has to be ordered first. We do not quote a fixed recovery time here. What you can expect is part of your support agreement.

    Questions about backup and recovery?

    Discuss your pilot