Server administration

Why “the server just sits there” isn’t enough

An unsupervised server works exactly until it stops - and then it turns out nobody knows what was running on it, which services depended on it, or whether the backup covered all of it.

Three things decide whether a failure is an incident or a crisis: how up to date the system is, monitoring of parameters, and documentation of what’s actually running. All three are boring and all three are cheap compared to rebuilding an environment from memory.

What we most often find

• a system past End of Life that can no longer be safely patched
• no monitoring, so you find out about a problem from users
• a configuration known to one person who’s already left the company - that is, no inventory
• a backup nobody has ever test-restored

What administration covers

• installation and reinstallation of the system on physical and virtual servers
• regular updates plus emergency ones for critical security bugs
• service configuration - files (SMB, FTP, SFTP), terminal access, Active Directory, databases, HTTP/HTTPS, reverse proxy, SSL, SSH, Proxmox
• parameter monitoring for the server and services on an agreed list
• physical server configuration - disks, out-of-band management
• platform selection - physical or cloud, matched to the specific task
• inventory of servers and of service and application licenses

Databases

We configure and maintain MSSQL, MySQL, PostgreSQL, Firebird and SQLite, along with a backup to the local file system. A copy off the server is a separate topic - covered in the backup service.

Updates without stopping the business

The most common reason servers don’t get updated isn’t technical - it’s fear that something won’t come back up after a restart. That fear is justified if there’s no documentation and no way to roll a change back.

That’s why we take a virtual machine snapshot before every major update and have a documented rollback path. Maintenance windows are set outside the company’s working hours, and for critical security patches we don’t wait for the next window - the risk of an unpatched vulnerability is then bigger than the risk of an unplanned restart.

We treat End-of-Life systems separately. They can’t be patched, so the conversation is about migration, not updates. We present a replacement plan with dates and costs in advance, so it doesn’t land in the budget as a surprise expense.

Virtualization: why bother with a single server

Even with a single physical server, virtualization pays off for three reasons that only become obvious during an outage. A snapshot lets you roll back a failed update in minutes instead of restoring the system from a backup. Splitting services across separate machines means a problem with one application doesn’t take the others down with it. Rebuilding the environment on different hardware stops depending on whether the new server has compatible drivers.

We use Proxmox VE because it lets us keep virtual machines and containers on the same host, and its backup server integrates without extra tooling.

Who this works for

Companies where the server is part of the process, not an extra: production with an ERP system, accounting firms with client databases, trading companies with an online warehouse. If a server stopping stops people working, oversight stops being optional.

Frequently asked questions about servers