A drive in my old four-bay Synology threw a warning email at 2am, and I nearly ignored it because the drive was reporting "healthy" in every dashboard I checked. Two weeks later it dropped out of the array mid-rebuild and took a chunk of my photo library with it.
The warning had been there the whole time. I just did not know which number to read.
SMART (Self-Monitoring, Analysis and Reporting Technology) has been baked into hard drives since the late 1990s, and modern drives report dozens of attributes. Most of them are noise for a home user. A small handful genuinely predict failure, and Backblaze has spent years publishing data from hundreds of thousands of drives to prove exactly which ones. Here are the attributes worth your attention, in the order I check them.
1. Reallocated Sector Count (SMART 5)
This is the one I look at first, every single time. When a drive finds a bad sector, it quietly swaps in a spare from a reserved pool and increments this counter. So a non-zero value means the drive has already started dying, just slowly.
A brand new drive should read 0. One or two reallocated sectors on a drive that has been running for years is not an emergency, but it is a reason to start watching closely.
The number I treat as a hard line is anything that climbs. A drive sitting steady at 8 for a year is different from a drive that went from 0 to 40 in a month.
According to Backblaze's own failure analysis, this attribute is one of the five that correlate most strongly with drives that actually died.
2. Current Pending Sector Count (SMART 197)
This is the scariest one, and the most misunderstood. A pending sector is a spot the drive tried to read, could not, and has flagged as unstable but has not reallocated yet. It is waiting to see if a future write fixes it.
The reason it matters so much: a pending sector is unreadable data right now. If that sector held part of a file, that part is gone unless you have a copy elsewhere.
Any value above 0 here gets my full attention. On a NAS with redundancy, a scrub or RAID scan will often force the drive to either recover or reallocate the sector, which then moves the count over to SMART 5.
3. Uncorrectable Sector Count (SMART 198)
Officially called Offline Uncorrectable, this counts sectors the drive gave up on entirely. Where a pending sector is "I am not sure yet," an uncorrectable sector is "I tried and failed."
A non-zero value here usually means you have already lost readable data on that sector. Combined with a rising SMART 197, this is a drive I pull from service without much debate.
4. Reported Uncorrectable Errors (SMART 187)
This attribute counts errors that could not be fixed by the drive's error-correction hardware. Backblaze specifically calls this out as one of their best predictors, and in their fleet, drives with a non-zero value here failed at dramatically higher rates.
My rule is blunt: if SMART 187 is above 0, plan the replacement. It does not mean the drive dies tomorrow, but it has crossed from "probably fine" into "living on borrowed time."
5. Command Timeout (SMART 188)
Command Timeout counts how many operations the drive failed to complete in the expected window. Some drives report inflated raw values here because of how the vendor encodes them, so do not panic at a large number on its own.
What matters is the trend and the context. A drive that suddenly starts racking up timeouts, especially alongside dropouts from your array, is often on its way out or has a cabling or power problem worth ruling out.
6. Spin Retry Count (SMART 10)
This counts how many times the drive had to retry getting the platters up to speed on startup. A healthy drive spins up first try, every time, so this should sit at 0.
Any non-zero value points at a mechanical or motor issue. It is less common than the sector attributes, but when it shows up it tends to mean the drive is genuinely tired.
7. Power-On Hours (SMART 9)
This one does not predict failure by itself, but it is essential context for reading everything else. It simply counts how long the drive has been running.
A typical desktop-class NAS drive is rated for a few years of continuous use, and consumer drives spinning 24/7 often show real wear somewhere past the 40,000 to 50,000 hour mark. That is roughly five to six years of always-on service.
| Attribute | ID | Healthy value | Action threshold |
|---|---|---|---|
| Reallocated Sector Count | 5 | 0 | Any rising value |
| Reported Uncorrectable | 187 | 0 | Above 0 |
| Command Timeout | 188 | Low and stable | Rising sharply |
| Current Pending Sector | 197 | 0 | Above 0 |
| Offline Uncorrectable | 198 | 0 | Above 0 |
How to actually read these on your NAS
On a Synology, open Storage Manager, click the drive, and choose Health Info to see the SMART table. QNAP hides it under Storage and Snapshots. On TrueNAS or any Linux box, the command smartctl -a /dev/sda from the smartmontools package dumps everything.
The column you want is the raw value, not the normalized "value" or "threshold" columns that most dashboards summarize into a green checkmark. That normalized number is exactly why my dying drive read "healthy" for two weeks.
Run the long test, not just the quick one
Reading current attributes tells you the drive's history. Running a SMART self-test forces the drive to check itself right now, and it can surface pending sectors that a passive read never touches.
The short test takes a couple of minutes and mostly checks the electronics. The long, or extended, test reads the entire surface and can take 6 to 12 hours on a large drive, but it is the one that actually finds weak sectors before they bite you.
I schedule an extended test monthly on every drive in my main NAS. It is free, it runs overnight, and it has caught two drives for me before either one caused real trouble.
Reading the drive before it reads you
None of these numbers are magic. A drive can die suddenly with a spotless SMART report, and one with a couple of reallocated sectors can soldier on for years.
But the pattern is consistent enough to bet on: attributes 5, 187, 197 and 198 climbing together is the sound of a drive dying, and it usually gives you days or weeks of warning. Set up email alerts on your NAS, glance at the raw values once a month, and keep a spare drive on the shelf so replacement is a chore and not a crisis.
Frequently asked questions
Does a SMART warning mean my drive will fail immediately?
Not necessarily, but it means the drive has crossed from healthy into at-risk. Attributes like Reported Uncorrectable and Current Pending Sector usually give you days to weeks of warning, not minutes. Use that window to copy data off and order a replacement rather than gambling on how long it lasts.
Why does my NAS say the drive is healthy when SMART shows bad sectors?
The "healthy" or "PASSED" status reflects the drive's normalized values against a conservative manufacturer threshold, not the raw counts. A drive can accumulate reallocated or pending sectors and still pass that check for a long time. Always read the raw value column yourself instead of trusting the summary.
How many reallocated sectors are too many?
There is no universal number, but the trend matters more than the count. A drive steady at a handful for a year is less worrying than one that jumped from 0 to 40 in a month. Any consistent upward climb in reallocated sectors is a strong reason to replace the drive.
Should I run the short or the extended SMART test?
Run both, but rely on the extended test. The short test checks the drive electronics in a couple of minutes, while the extended test reads the entire disk surface and can take 6 to 12 hours. Only the extended test reliably surfaces weak or pending sectors before they cause data loss.
Can a failing drive damage the rest of my NAS array?
A single failing drive in a redundant array will not corrupt healthy drives, but it can stall rebuilds and cause timeouts. If a drive shows rising Command Timeout values and keeps dropping out, replace it before you attempt a rebuild. Running a degraded array on a shaky drive is how single failures turn into total losses.
Which tool should I use to check SMART values?
On Synology use Storage Manager's Health Info, on QNAP look under Storage and Snapshots, and on TrueNAS or Linux use the smartctl command from smartmontools. All of them expose the same underlying attributes. The key is to view the raw values rather than the summarized health status.
