Server Will Not Power On: A Diagnostic Order That Saves Time
A server that will not power on has a small number of likely causes, and working through them in order is faster than replacing parts. This covers the sequence.
Establish what is actually happening
Three different symptoms that look the same from across the room.
No lights, no fans, nothing. The machine is getting no power, or the supplies have failed.
Fans spin briefly then stop. The machine is attempting to start and aborting, usually on a power or hardware fault it detected.
Fans run at full speed but nothing appears on screen. The machine is powered but not posting. This is a different problem from not powering on, and the section below covers it.
Knowing which you have narrows the causes considerably.
No lights, no fans
Work outward from the machine.
Both power cords seated, at the machine and the PDU. Cords work loose, and on a redundant machine one loose cord is invisible until the other fails.
The PDU outlet is live. Test with something else, or move to a known-good outlet.
The supplies show a status light. A supply with no light at all is getting no power; a supply with a fault light has failed. Our power supply guide covers reading them.
The management controller is alive. On most enterprise servers, iLO, iDRAC or IMM has standby power and responds even when the machine is off. If the management controller is reachable, the machine has power and the fault is elsewhere. If it is not reachable and the supplies show no light, the problem is upstream. Our out-of-band guide covers this.
Fans spin then stop
The machine detected something and aborted. Four common causes.
Insufficient power for the configuration. A machine that lost one supply and cannot run on the other alone will attempt and abort. Check whether both supplies are healthy.
A missing fan module. Many platforms refuse to start with a fan absent or failed, because they cannot guarantee cooling. Our fans guide covers this.
A processor or heatsink problem. A processor fitted without its heatsink, or a heatsink not seated, aborts immediately.
A memory configuration the platform refuses. Some platforms refuse an invalid population rather than running degraded. Our memory install guide covers valid populations.
Powers on but does not post
The most common cause after a hardware change is memory not fully seated. Reseat every module you touched, pressing at both ends until the clips close themselves.
Then, in order.
Read the diagnostic indicators. Enterprise servers have LEDs on the board or a display panel that identify the failing component. This is much faster than guessing, and it is the thing people skip.
Check the management controller log. It frequently records what stopped the boot, and it is readable remotely.
Remove everything non-essential. One processor, minimum valid memory, no expansion cards, no drives. If it posts, add components back one at a time.
Check processor firmware support, if a processor was recently changed. A newer generation than the firmware supports fits and does not post. Our processor guide covers this.
If a hardware change preceded it
Whatever changed is the first suspect, regardless of how unrelated it seems.
Remove the new part and confirm the machine returns to working. That single test separates a faulty part from a coincidence, and it takes minutes.
Common culprits: memory not seated, an expansion card in a slot that cannot supply it, a processor without firmware support, a drive causing the controller to hang during initialisation.
When to stop diagnosing
If the machine will not post with minimum configuration, known-good memory, and a single processor, the fault is the board, the processor or a supply — and on an older machine those are frequently worth more as a decision point than as a repair.
Weigh the cost of a system board and the time against a replacement machine. Our repair or replace guide covers the call, and our spares guide covers why a cold spare machine is worth holding for platforms that matter.
Frequently asked questions
My server will not turn on at all. What do I check first?
Both power cords seated at machine and PDU, the outlet is live, the supplies show status lights, and whether the management controller responds. A reachable management controller means the machine has power and the fault is elsewhere.
Why do the fans spin briefly then stop?
The machine attempted to start and aborted on a detected fault. Common causes are insufficient power for the configuration, a missing or failed fan module, a processor or heatsink problem, or a memory population the platform refuses.
Server powers on but nothing on screen. What now?
Reseat memory first, then read the diagnostic indicators or display panel, then the management controller log, then strip to minimum configuration and add components back one at a time.
It stopped working right after I added a part. Coincidence?
Assume not. Remove the new part and confirm the machine returns to working. That single test separates a faulty part from a coincidence and takes minutes.
Can a missing fan stop a server booting?
Yes. Many platforms refuse to start with a fan module absent or failed, because they cannot guarantee cooling. The same applies to missing air baffles on some designs.
When should I stop diagnosing and replace the machine?
If it will not post with minimum configuration, known-good memory and a single processor, the fault is the board, processor or a supply. On an older machine, weigh a system board and the time against a replacement.
Tell us the model and what the diagnostic indicators show, and we will help you identify the part before you buy anything.
