Prevention

Reading SMART values before it starts clicking

Almost every storage device keeps records about itself. These values can be read out, and in many cases a failure announces itself there weeks in advance — you just have to know what to look at.

What SMART is and what it is not

SMART is a self-monitoring system that drives have included for decades. It counts events: how often a sector had to be reallocated, how often a read operation failed, how many operating hours have passed, how hot it got.

What SMART cannot do: make a prediction. A drive with impeccable values can fail tomorrow, and one with conspicuous values can run for years. The overall assessment “status: OK” shown by many programs is therefore of little value — it often only trips when it is long past too late. What is meaningful are the individual values.

The values that matter on hard drives

  • Reallocated sectors

    The drive has identified a location as unusable and replaced it with a spare. The single most important value there is. A handful over the years is inconspicuous — but if the number rises within weeks, the device is in decline.

  • Pending sectors

    Locations that proved problematic when read but have not yet been replaced. They are the precursor of the first category and a very serious sign, because data can be sitting exactly there.

  • Uncorrectable sectors

    Locations that could definitively no longer be read. Data has already been lost here.

  • Read error rate and seek error rate

    Be careful with the interpretation: some manufacturers write raw figures into these fields that look enormous by default and mean nothing. Only the change over time is meaningful.

  • Operating hours and power cycles

    Not a warning sign in itself, but context. A drive with a five-digit hour count in a RAID is a candidate for the next failure — and for the question of whether the backup is current.

  • Temperature

    Permanently high temperatures shorten the service life. Less interesting than the current reading is the maximum the drive has ever seen.

With SSDs something else counts

  • Wear level

    Indicates how much of the intended write endurance has been used up. In normal use this value stays inconspicuous for years; in servers and with continuously writing applications it can become an issue.

  • Amount of data written

    Compare the amount actually written with what the manufacturer specifies for the service life. That is the most honest measure of wear.

  • Spare blocks

    SSDs keep replacement cells in reserve. When that reserve declines it is time to act — with flash memory things then often move quickly.

  • Unexpected power losses

    A high counter points to an abruptly disconnected power supply. That is the most common cause of damaged translation tables and thus of SSDs that suddenly appear with the wrong capacity.

When you should act

As soon as reallocated or pending sectors appear and the number is growing, the rule is: back up, now, and before anything else is undertaken. No cloning with on-board tools, no test program, no defragmentation — all of that creates load on a drive that is in the process of showing it no longer tolerates load well.

And the most important sentence on the subject: if the drive is already making noises, SMART is no longer the question. Then it is time to pull the plug.

A practical note

SMART values can be read out with freely available programs — pure reading, without intervention. Anyone running a NAS will usually find the values directly in its interface and can set up notifications. Configuring that takes five minutes and is the cheapest data recovery there is.
Common questions

Answered briefly.

My program says “status: good”. Can I rely on that?

Only to a limited extent. The overall assessment often only trips once thresholds have been exceeded — and those are set generously by the manufacturer. Look at the individual values, above all at reallocated and pending sectors.

Can bad sectors be repaired?

No. The drive replaces them internally with spare locations — nothing is repaired, and the data that was there is generally lost. Programs that promise otherwise are merely postponing the problem.

My drive is clicking but shows good SMART values. How can that be?

Because self-monitoring has to be read out — and a drive with mechanical damage often no longer supplies any values, or only outdated ones. In that case the noise is the more reliable information.

A specific case?

We will take a look at your storage device.

The Economy diagnosis is free of charge. You find out what is possible before anything costs money — and what it would cost is put in writing beforehand.

Wasserweg 8–10 · 60594 Frankfurt am Main