Staggered Spin-Up: Drive Power Surge and Missing Bays at Boot
A server that runs perfectly but sometimes comes up with three drives missing, or a fully populated enclosure that will not power on at all while the same chassis half-filled works fine, is usually not a drive problem, a backplane problem or a controller problem. It is a power problem that exists only for a few seconds at start-up.
This guide explains why spinning a hard drive up costs far more power than running it, what staggered spin-up does about it, how the symptoms present, and why a rack full of equipment sometimes fails to come back after a power cut.
Spin-up is the expensive moment
A mechanical hard drive has to accelerate a stack of platters from stationary to full rotational speed. That takes considerably more current on the 12V rail than keeping them spinning does, and drive datasheets quote a peak spin-up figure well above the steady-state running figure for exactly this reason.
For one drive it is trivial. For twenty-four starting simultaneously it is a large instantaneous load arriving within a second or two of power being applied, at the same moment fans are also spinning up from rest.
Solid state drives do not have this problem in the same way, because there is no motor to accelerate. An all-flash chassis avoids the spin-up surge entirely, which is one of the less-discussed operational advantages of flash.
What staggered spin-up does
Rather than letting every drive start at once, the controller brings them up in groups or one at a time, with a short delay between each. The peak load is spread over tens of seconds instead of arriving all at once, and the power supply never sees the combined surge.
On SAS drives this is handled through the controller, which holds drives in a not-yet-spun state and releases them in sequence.
On SATA drives the equivalent is Power-Up In Standby, where the drive does not spin until told to. Support varies by model, and on some drives it is enabled by a jumper rather than in firmware, which is easy to miss when the drive arrives.
The cost is boot time. A large enclosure with aggressive staggering can take a noticeable while before all drives are present, and a controller that starts scanning before the last drives are ready may report them missing. If you enable staggering, check that any discovery timeout is long enough to cover it.
How the symptoms present
The confusing part is that these faults are intermittent and look like something else.
Drives missing at boot but present after a rescan. The classic signature. The drive was not ready when the controller looked, then spun up normally a moment later.
A different drive missing each time. A genuinely failing drive fails consistently. A rotating cast of missing bays points at timing or power, not at the drives. Our guide on a server hard drive not being detected works through the order of checks.
The chassis works half-populated and fails when full. A strong indicator that you are at the edge of the power budget rather than facing a component fault.
The power supply shuts down immediately on start. Protection circuitry reacting to an over-current condition that lasts a second. The server appears completely dead, which sends people towards the board or the PSU rather than the load. Our guide on a server that will not power on or POST covers the diagnostic order.
Size the supply for the surge
Power supply sizing calculations usually use steady-state figures, which is the right basis for thermal planning and the wrong one for start-up.
A chassis whose running load sits comfortably inside its supply can still exceed it briefly at spin-up. On a redundant pair this frequently goes unnoticed, because two supplies share the surge — until one fails and the machine will no longer restart on the survivor. The fault then looks like a bad power supply when it is a sizing assumption. Our guide on server power supplies, redundancy and sizing covers this properly.
External enclosures usually have their own supplies sized for a full population, but a JBOD filled with drives it was not specified for can hit the same limit. See JBOD enclosures explained, and SAS expanders and cascading for the bandwidth side of populating one.
The rack-level version
The same physics applies one level up, and this is why a room sometimes fails to come back cleanly after an outage.
When power is restored, every device in the rack that was set to power on automatically tries to start at the same instant. Every drive in every chassis spins up together, every fan starts from rest, and the inrush can trip the circuit breaker feeding the rack. The breaker trips, power is lost again, and on reset the whole thing repeats.
Managed PDUs address this with per-outlet power-on delays, bringing equipment up in a defined sequence over a minute or two rather than all at once. It also lets you control order, so storage is ready before the hosts that mount it. Our guide on rack PDUs and power planning covers configuring that, and UPS sizing and power protection covers the supply side.
This is worth testing rather than assuming. A planned shutdown and restart tells you whether your rack comes back on its own. Discovering it does not during an actual outage is an expensive way to find out. Our guide on moving a server room covers shutdown and startup order, which is the same problem.
Drives and the restart itself
One more reason to care: spin-up is mechanically the hardest thing a hard drive does, and drives that have run continuously for years are most likely to fail on the restart rather than during operation.
That is not an argument against restarting, it is an argument for expecting it. Before a planned power-down of an old array, confirm the backup is current and have spares to hand. Our guides on how long enterprise drives last and the 3-2-1 rule cover the risk.
Sourcing
We supply enterprise SAS and SATA drives, carriers, power supplies and controllers for server and JBOD deployments. If a chassis is dropping bays at boot or failing to start when fully populated, tell us the chassis model and the drives fitted and we will help work out whether it is a power budget problem before you replace parts. Request a bulk quote or email sarah.jane@techsellerusa.com.
Common questions
What is staggered spin-up?
Bringing drives up in groups or one at a time with a short delay between each, rather than letting them all start together. A mechanical drive draws considerably more current accelerating its platters than it does running, so twenty-four starting at once creates a large instantaneous load. Staggering spreads that over tens of seconds so the power supply never sees the combined surge.
Why are drives missing at boot but present afterwards?
The drive was not spun up and ready when the controller enumerated the bus, then came up normally a moment later. A rescan finds it because by then it is ready. The giveaway that this is timing rather than hardware is that a different drive is missing each time; a genuinely failing drive fails consistently. Check whether the discovery timeout is long enough to cover the staggering delay.
My chassis works half-full but not fully populated. Why?
That pattern is a strong indicator of a power budget problem at start-up rather than a component fault. The running load may sit comfortably inside the supply's rating while the combined spin-up surge briefly exceeds it, and the protection circuitry shuts the supply down. Enabling staggered spin-up usually resolves it; if not, the supply is undersized for a full population.
Does staggering work the same on SAS and SATA?
Not quite. On SAS the controller holds drives in a not-yet-spun state and releases them in sequence, which is well supported. On SATA the equivalent is Power-Up In Standby, where the drive does not spin until instructed. Support varies by model and on some drives it is enabled by a physical jumper rather than in firmware, which is easy to overlook when the drive arrives.
Why did my whole rack fail to come back after a power cut?
Because everything set to power on automatically tried to start at the same instant. Every drive spun up together, every fan started from rest, and the combined inrush tripped the breaker feeding the rack. Power is lost again, and on reset the cycle repeats. Managed PDUs with per-outlet power-on delays fix this by sequencing equipment over a minute or two, and they also let you bring storage up before the hosts that mount it.
Do SSDs need staggered spin-up?
Not in the same way, because there is no motor to accelerate, so the large mechanical spin-up surge does not exist. An all-flash chassis avoids the problem entirely, which is a genuine if rarely mentioned operational advantage. Flash drives still draw current at power-on, so a very dense enclosure is not completely free of inrush considerations, but it is a different order of magnitude.
If a chassis is dropping bays at boot or will not start when fully populated, send us the chassis model and the drives fitted before replacing parts.
