Why maintenance must be scheduled
Maintenance that only happens when there is time does not happen. A fixed maintenance window per month, such as one evening or an early morning, announced to employees, ensures that updates and restarts do not take place in the middle of the working day and that someone checks afterwards whether everything is running again.
Monthly
- Install operating system updates (Linux: package updates and a restart for kernel updates; Windows: patch day updates), then check services
- Application updates: database, web server, PHP, line-of-business software, antivirus definitions
- Check backups: did all jobs succeed, is the size right, are there error messages?
- Disk space on all volumes, growth over recent months, clean up old log files and temporary data
- Check logs for anomalies: failed logins, hardware errors, recurring warnings
- SMART values of the disks and RAID status, fix degraded arrays immediately
- Review monitoring alerts of the past weeks: what triggered, was it fixed?
Quarterly
- Restore test: recover a file, a directory and once a year an entire server from backup, note the duration
- Review users and access: employees who left, unused accounts, administrator rights, SSH keys
- Certificates and domains: expiry dates, does automatic renewal work?
- Compare firewall rules and open ports against the documentation
- Performance: utilisation of CPU, RAM and disks, is the hardware enough for the next twelve months?
Yearly
- Update firmware and BIOS of server, RAID controller and network devices
- Physically check hardware: fans, dust, temperature in the server room, UPS battery test
- Operating system versions: is anything still running that loses support in the next twelve months?
- Update the emergency plan: contacts, access, order of recovery
- Go through the documentation completely and compare it with the actual state
What monitoring should cover
Monthly maintenance is only enough if monitoring keeps watch in between: reachability of services, disk space, RAID and SMART status, backup results, certificate expiry, utilisation and security events. Alerts must reach someone who can react, by email, messenger or SMS, and not just light up on a dashboard.
Documentation
Every maintenance is briefly logged: date, what was done, what was noticed, what is due next time. It sounds like bureaucracy, but it is the difference between "someone did something at some point" and a traceable operating history that saves hours with every problem and every change of provider.
Frequently asked questions
Should updates be installed automatically?
Security updates yes, on Linux servers with automatic updates and a restart in the maintenance window, on Windows with a controlled patch day. Major version jumps and application updates belong in scheduled maintenance with a prior backup.
How long does monthly maintenance take?
For a single server with monitoring usually one to two hours including checks and log. Without monitoring and automation considerably longer, because everything has to be checked manually.
Do you also take over maintenance for a flat rate?
Yes, as a maintenance run at a fixed price of €149 per server including updates, backup checks and report, bookable monthly or quarterly, without subscription, for Linux and Windows, remotely and in Vienna on site if needed.
Need help implementing this? I implement it for you, remotely or on site in Vienna, at a fixed price or based on effort. Send request →