Power cuts in Antananarivo: preparing SME servers to recover

A power cut can stop a server, its storage and the network at different times. The useful decisions are which services must return first, how long the business can wait and how much recent work it can afford to lose. JIRAMA publishes planned outage notices, which help with preparation; the same recovery plan should also work for an unexpected cut.

1. Write down what must keep working

Start with business activities rather than a list of machines. If invoicing, shared files or the main application stops, which teams cannot work? For each service, note the longest acceptable interruption and how much work could be re-entered.

On one page, connect each activity to its dependencies: server, storage, local network, router, internet connection and the people who can respond. A server on a UPS is still unreachable when its network switch or storage loses power.

2. Check real UPS runtime

A UPS buys time to ride through a short cut or shut systems down cleanly. Runtime depends on the connected load and battery condition, not just the number on the product box. Measure it with the actual equipment and set a shutdown threshold before the battery runs out.

Automatic shutdown must follow the order required by your environment: applications and virtual machines, host, then storage where applicable. Test both shutdown and restart, preferably outside production hours. A qualified professional should handle electrical work and sizing.

3. Prepare a restore, not just a copy

The UPS protects shutdown; backups protect data when shutdown is not enough. Identify essential data and configurations, choose a copy frequency based on acceptable data loss, and keep a copy that does not depend on the same server or power supply.

A “successful” backup job does not prove that the business can resume. Restore a representative file, database or machine in a test environment. Record the time taken, access needed and the person who can repeat the operation.

4. During the cut: follow the written procedure

Record the start time and tell the affected people which service is unavailable. If expected runtime will not cover the cut, start an orderly shutdown before the battery runs out. Avoid repeated restarts while power is unstable.

Keep instructions, useful phone numbers and emergency access available outside the stopped server. A procedure stored only there will not help when it is needed.

5. When power returns: test the business activity

Wait for stable power, then restart in the documented dependency order. Check storage, virtual machines, network services and applications. A green server light is not enough: ask a user to perform a simple real task, such as opening a shared folder or finding an invoice.

Check that backups and monitoring resumed. Note what took longer than expected and update the procedure. NIST recommends testing contingency plans; an outage gives the team concrete evidence for improving one.

The minimum plan to keep close

• Priority services : Which work stops, and in what order should services return?

• Runtime : How long do the UPS and the actual connected devices last?

• Shutdown and restart : Who starts each step, and where is the procedure?

• Backup : When was the last successful restore test?

• Business check : Which simple task proves that people can work again?

A principle used in our work

Before production VMware to Proxmox migrations, we tested backups with a real restore and prepared a rollback for each phase. The same principle applies after a power cut: plan the recovery before you need it.

Review your servers and backups

The Infrastructure package includes configured and checked backups, monitoring and a recovery plan. A free 30-minute first conversation helps define scope; the starting price and included work are published.