5.2.1. Hardware Failure Symptoms and Causes
💡 First Principle: Hardware failure symptoms fall into two categories: those you see (LEDs, LCD readouts, error screens) and those the system tells you (POST codes, event log entries, diagnostic alerts). Reading both categories together gives you the full picture.
Memory Failure Patterns
Memory failures are among the most varied in their presentation — RAM errors can mimic almost every other failure type:
| Symptom | Probable Memory Cause |
|---|---|
| Blue Screen of Death (BSOD) with MEMORY_MANAGEMENT | RAM cell failure or timing incompatibility |
| Kernel panic (Linux) | Memory corruption |
| Random lockups (no pattern) | Uncorrectable memory errors (ECC fixes single-bit errors silently and logs them; multi-bit errors lock or crash the system) |
| Memory dump generated | OS captured RAM state before crash — indicative of memory or driver fault |
| High memory utilization with no change in workload | Memory leak in application or OS component |
| POST error codes at boot | RAM not recognized, incompatible, or failed |
Power Failure Patterns
| Symptom | Probable Cause |
|---|---|
| Random shutdowns under load | PSU insufficient for load, failing capacitors |
| Server won't power on | PSU failure, blown fuse, bad outlet |
| Damage to multiple components simultaneously | Power spike, failed PSU providing incorrect voltage |
| Burning smell | PSU overheating, capacitor failure |
Thermal Failure Patterns
| Symptom | Probable Cause |
|---|---|
| System crashes after warm-up period | Thermal throttling → crash when temperature threshold exceeded |
| Fan alerts in event log | Malfunctioning fan (bearing wear, blocked airflow) |
| CPU/GPU overheating alerts | Improperly seated heat sink, dried thermal paste, failed fan |
| Performance degrades gradually under load | Thermal throttling (CPU/GPU reduces speed to manage heat) |
Indicator Systems
Servers include multiple physical indicator systems to communicate status without requiring OS access:
| Indicator | What It Shows | Example |
|---|---|---|
| Drive LEDs | Green = healthy; Amber/Red = failed or rebuilding | Amber LED on drive bay = drive fault |
| System health LEDs | Overall chassis status | Blinking amber = fault condition |
| PSU LEDs | Power supply status | Green = OK; Off = PSU fault |
| NIC link lights | Link established, activity | No link light = physical connection problem |
| LCD/VFD panel | Fault codes, IP addresses, status messages | Shows specific error code for quick diagnosis |
| POST codes | Progress through hardware initialization | Specific code identifies which component failed initialization |
| Beep codes | BIOS-era diagnostic | Specific patterns indicate RAM, video, CPU failures |
CMOS Battery Failure
A dead CMOS battery causes a specific, recognizable symptom cluster:
- System date and time reset to default (often 1/1/2000 or similar) after power loss
- BIOS settings reset to defaults after power cycle
- May cause "CMOS checksum error" at boot
The CMOS battery maintains BIOS settings and the real-time clock when the server is powered off. It does not affect server operation while powered — only the state that's preserved between power cycles.
Misallocated Virtual Resources
In virtual environments, resource misallocation can produce symptoms that look like hardware failure:
- VM showing 100% CPU but host shows low physical CPU — vCPU overcommit creating scheduling delays
- VM with severe disk latency — IOPS overcommit on shared datastore
- These symptoms appear in the guest as hardware problems but are actually hypervisor configuration issues
⚠️ Exam Trap: A purple screen (PSOD — Purple Screen of Death) is specific to VMware ESXi and indicates a hypervisor-level crash, not a guest OS failure. A BSOD (Blue Screen of Death) is Windows; a kernel panic is Linux. Knowing which OS produces which crash indicator is directly tested.
Reflection Question: A server crashes every Tuesday morning approximately 30 minutes after the weekly backup job starts. Temperatures are normal during the first 20 minutes. What hardware component is most likely failing, and what specific indicator would you check first?
