Select server maintenance methods, procedures and tools
Key takeaways
- Restore returns lost or corrupted data to a working state from a backup; recover is the broader return of a server or service to normal operation, which may also mean rebuilding configuration, reinstalling roles or failing over to another server.
- Preventive maintenance is planned work before any problem, predictive acts on an early warning sign in monitoring data, corrective fixes a fault that has already occurred, and periodic runs at fixed intervals regardless of the server's condition.
- The test between periodic and preventive is the trigger: purely the calendar is periodic, a deliberate plan to reduce risk is preventive.
- Maintenance covers every resource a server depends on: network, disks, time, memory and devices, not only its disks.
- A programme built on corrective maintenance alone is simply reacting to disasters; preventive and periodic work form the baseline, with predictive maintenance added where good monitoring data exists.
A server that is installed, configured and left alone degrades: disks fill, logs grow, certificates expire, and small faults accumulate until one of them becomes an outage. This closing paper covers how that decline is prevented and managed. This opening chapter sets out the vocabulary of disaster, loss, restore and recovery, the four distinct types of maintenance, and the specific server resources maintenance routinely covers.
Key terms
Types of server maintenance
Maintenance is classified by when and why it happens, and confusing the four types is the most commonly tested error in this chapter.
| Type | Triggered by | Example |
|---|---|---|
| Preventive | A fixed plan, acting before any sign of trouble | Cleaning dust from server fans quarterly |
| Predictive | An early warning sign in monitoring data | Replacing a disk whose SMART data shows rising errors |
| Corrective | A fault that has already occurred | Restoring a file after accidental deletion |
| Periodic | A fixed calendar interval regardless of condition | Monthly security patching on the second Tuesday |
Server resources maintenance covers
Maintenance is applied across every resource a server depends on, not only its disks.
Server maintenance techniques
Server maintenance techniques combine these maintenance types and target resources into a coherent programme: most servers benefit from preventive and periodic maintenance as a baseline, supplemented by predictive maintenance where good monitoring data is available (see the chapter on server monitoring), and corrective maintenance whenever something does go wrong despite the rest. A programme built on corrective maintenance alone, fixing things only once they break, is not really a maintenance programme at all; it is simply reacting to disasters as they happen, exactly the state the rest of this chapter's preventive, predictive and periodic techniques exist to avoid.
Classifying five maintenance actions by type
A data centre logs five actions from the past month: replacing a UPS battery on its scheduled three-year replacement date regardless of its tested condition; replacing a disk after its SMART data showed a rising reallocated-sector count over several weeks; restarting a web server process after it crashed under an unusual load spike; cleaning dust from server intake fans during the routine quarterly maintenance window; and patching all servers on the second Tuesday of the month as always.
UPS battery replaced on schedule regardless of condition: periodic maintenance, since it is triggered purely by the calendar interval, not by any observed sign of failure.
Disk replaced after SMART data showed a developing trend: predictive maintenance, since the action was triggered by an early warning sign in monitoring data rather than a fixed date or an actual failure.
Web server process restarted after it crashed: corrective maintenance, since the fault had already occurred before the action was taken.
Fan cleaning during the routine quarterly window: preventive maintenance, planned work carried out before any sign of a problem, to reduce the chance of overheating occurring at all.
Monthly patching on a fixed date: periodic maintenance, the same classification as the UPS battery, since both are triggered by a fixed schedule rather than by condition or by an already-occurred fault.
Self-check
(a) Distinguish restore from recover. (2 marks)
(b) Define the four types of server maintenance. (4 marks)
(c) Classify each of the following by maintenance type: replacing a fan on its fixed three-year replacement schedule; replacing a battery after monitoring showed its charge capacity declining; fixing a corrupted configuration file after a server crash. (3 marks)
(d) State four server resources maintenance routinely covers. (4 marks)
(e) Explain why system clock accuracy matters for a domain server specifically. (2 marks)
(f) Explain why a maintenance programme built on corrective maintenance alone is not really a maintenance programme. (2 marks)
Show worked answer
(a) Restore returns lost or corrupted data to a working state from a backup. Recover is the broader process of returning a whole server or service to normal operation, which may involve restoring data but also rebuilding configuration, reinstalling roles, or failing over to another server.
(b) Preventive: planned, routine work carried out before a problem occurs. Predictive: uses monitoring data and trends to anticipate a likely failure before it happens. Corrective: carried out after a fault has already occurred, to fix it. Periodic: carried out at fixed, regular intervals regardless of condition.
(c) Fan replaced on a fixed schedule: periodic. Battery replaced after monitoring showed declining capacity: predictive. Configuration file fixed after a crash: corrective.
(d) Network, disks, time, memory, devices (any four).
(e) Domain authentication (Kerberos) depends on server clocks being closely synchronised; if a domain server's clock drifts too far from the others, authentication attempts can fail even though the credentials themselves are correct.
(f) Corrective maintenance only acts after a fault has already occurred, meaning every issue becomes a live disruption before it is addressed; relying on it alone means the organisation is simply reacting to disasters as they happen rather than reducing how often they occur in the first place, which is what preventive, predictive and periodic maintenance are for.
Exam tip
Preventive versus predictive versus corrective is a frequently tested distinction: preventive is scheduled regardless of condition, predictive responds to an early warning sign in the data, corrective responds only after a failure has already happened. Periodic is often confused with preventive; the test is whether the trigger is purely the calendar (periodic) or a deliberate plan to reduce risk that may or may not be calendar-based (preventive).