How to read SMART data on the hard drive? My cousin was so scared by the 2.1 billion RAW value that he wanted to switch tracks, but that number was definitely not bad at all

last month, my cousin posted a screenshot in the family group chat. His second-hand Synology had a Western Digital Red Drive 8TB, and the CrystalDiskInfo had a yellow 'warning' ring on top, followed by a string of numbers after the Seek_Error_Rate line: 2,181,507,313. He asked me: 2.1 billion pathfinding errors—do you have to change this game on the same day? I opened the large image and glanced at the adjacent sections, then replied: This plate is very healthy, don't place an order yet. This plate is still running around in his group chat, and he hasn't spent a single cent.

this isn't his fault. SMART outputs dozens of rows or ten columns of data, and most online tutorials just say 'check RAW value,' leaving anyone confused by the first time they see it. In this article, I use the output of that disk as a textbook, breaking it down line by line: which columns are key, which numbers look scary but are actually fine, and which are truly deadly. Which indicators are inherently dangerous? I previously wrote about five red and yellow card tables in the article the signs of a bad market not repeating them. Today, I will focus on "how to read them."

to clarify first: SMART is not a physical exam score, but a bunch of independent counters

many people understand SMART as "hard drive health report, 80 points healthy, 60 points passing." No, that's not it. The original English term for SMART is Self-monitoring, analysis, and reporting technology. Essentially, it's a bunch of counters running in the hard drive firmware: how many hours they've been powered, how many times the head has been routed, how many sectors have been remapped, each recording its own account. It has no total score.

so after running SmartCTL, the output is split into two parts. The top line is the overall evaluation, labeled SMART overall-health self-assessment test result: PASSED. This PESSOLD only means the disk's self-assessment is not failing, and is the thickest line. Below is the proper attribute list: ID#, ATTRIBUTE_NAME, VALUE, WORST, THRESH, TYPE, UPDATED, WHEN_FAILED, RAW_VALUE ten columns with dozens of rows. It looks impressive, but what really needs attention are the four middle columns.

one more premise to be sure first: these numbers are customized by the manufacturer. For the same 'normalized current value,' Western numbers often use 100 as full power; some disks use 200, others use 253, and the range is different. So comparing your Western Digital's VALUE with your friend's Seagate is pointless. There is no comparison between plates; a plate only makes sense when compared to itself.

four columns of numbers each matter, one table explains everything clearly

cousin's red plate Seek_Error_Rate business, I'll copy it down for you to see—it's a textbook example:

the iron rule for interpreting it is simple: first compare VALUE and THRESH, then look at RAW. That disc had a VALUE of 100, passing 67, still 33 points short of the listing, and WORST was also 100, indicating it had never come close to the passing line in history. All three sports are green, so this line of work is absolutely no problem. My cousin's mistake was skipping the first two columns and reading the last column directly, interpreting it literally—2.1 billion divided by 21,000 hours of power usage equals 28 errors per second. This game should have been smoking long ago, clearly something was wrong.

By the way, what are the other columns for? TYPE labeled Pre-fail indicates failure precursors, Old_age indicates wear; UPDATED is for continuous updates for Always, while for Offline, updates are only during offline testing; WHEN_FAILED normally appears as a horizontal bar. Once the word 'FAILING_NOW' appears, don't hesitate—back up your data and schedule the replacement.

Of the three pitfalls

RAW listed, my cousin fell into the first

RAW this column is the most prone to trouble, with three pits, each more hidden than the last.

pitfall 1: The RAW for rate-type metrics is packaged coding, not simple counting. Seek_Error_Rate. Read_Error_Rate these metrics, vendors don't just add 'error count' to RAW, but to multi-field composite values that combine the error count and total path seeks by byte segment, sometimes even mixing in timestamps. So you'll see the astronomical figure of 2.1 billion—it's a coded string, not a quantity you can directly divide with a calculator. Don't look at RAW for these indicators—just look at VALUE and trend. Seagate does this too, and on forums, people who were scared by these numbers posted everywhere could be found everywhere.

Pit 2: hexadecimal. Some discs have some RAW indicators stored in hexadecimal, with numbers mixed with letters a to f. I've mentioned this before: don't panic when you see 1000; converting to decimal might be 16. The recognition method is simple: when letters appear, they are basically hexadecimal; pure numbers can also be used. If you're not sure, just compare them with screenshots from other disks of the same model. If the magnitude matches, you read correctly.

Pitfall 3: Unable to tell which RAW files can be read directly. Here's a simple binary for you. Read directly: Current_Pending_Sector, Reallocated_Sector_Ct, UDMA_CRC_Error RAW is the value, 0 is best; Temperature, Power_On_Hours, Start_Stop_Count are also plain codes and read directly. Don't read it directly: all velocity indicators marked with Error_Rate or Retry_Rate—see pit one for the reason. The three things I really need to check on my cousin's disk: Pending is 0, Reallocated is 0, and UDMA_CRC is 0—everything is clean.

the indicators to really watch: the absolute value group and the trend party

narrow the scope to the few worth watching, dividing usage into two camps.

the correct way to use the trend party is to screenshot and archive. My habit is to glance at SMART once a month, take a screenshot, and toss it into a dedicated folder in my album to compare next time. Start_Stop_Count this is something I deeply regret: my NAS at home started and stopped forty or fifty times in one night. I relied on screenshots from two consecutive nights to detect abnormal speed, and in the end, it was found to be a downloaded container stealing the drive. The complete investigation from that time was written in the of dormant article. For second-hand properties, you also need to keep screenshots of the day's numbers. My cousin's unit was powered for 21,000 hours, which is even older than the was considered refurbished when relocating. Every increase in the future will have to be reconciled.

two paths that don't want to give orders

smartctl running in SSH, some find it troublesome; there are two alternative paths with their own choices.

First, the NAS comes with a built-in health page. Both Synology's storage manager and Feiniu's storage management have built-in SMART viewing, with graphical interfaces available just by clicking and featuring scheduled detection tasks. The downside is that they only provide conclusions and a few items; the full output of the four-column system is not visible, and when encountering a "yellow warning but not sure if it's false," it's just a blank guess—that's how my cousin's screenshot came from. If you want to be serious, you still need to go in SSH and run smartctl -a. Use lsblk to confirm drive letters first, don't just try to fix the SDA operation and the result is flashing SDB. When installing a drive in a USB hard drive enclosure, you need to add the -d SAT parameter; otherwise, it will show empty. I mentioned this pitfall in my article on testing new drives in .

second, you can remove the disk and connect it to the computer, install CrystalDiskInfo on Windows, a well-established tool that has been free for over ten years. It folds states into three colors: blue good, yellow warning, and red anomaly. The color determination logic is based on the VALUE rules for THRESH mentioned above, except the machine compares for you. A reminder: yellow warnings often cause false alarms, most commonly because the normalized value of the rate-type indicator is naturally low (some disks have factory VALUE right above the passing line), so at the slightest disturbance, it turns yellow. So don't panic when you see yellow—open the details and see which item is yellow: 05 remapping, C5 pending mapping, and C6 cannot fix RAW non-zero. Yellow means it's real, back it up quickly; Huang Seek_Error_Rate this kind of speed is most likely a cousin-style false alarm.

here's a fun fact: SMART data is stored in the disk's own firmware area and follows the disk. The drive is removed from the NAS and connected to the computer, and not a single number changes; Switch machines to read the same disk, but still read the same ledger. So there's no such thing as 'changing the environment and getting a fresh check-up.'

proactive testing and a fair word

can actively test beyond passively looking at numbers. Smartctl -t short test lasts two to three minutes; my sense of smell for bad tunnels is sharper than human ears, so if I'm suspicious, I run one. -t Long Long It takes several hours to scan the entire disk, but it's worth running before placing a replacement order or when you get a second-hand drive. Plus, during the long test, the head is fully loaded, so don't squeeze the backup tasks into the same night. I have written about the applicable scenarios for both in the previous two articles, so I won't elaborate here.

to wrap up, to be fair, and this is the phrase I most want you to remember: SMART all green doesn't mean this steady . In 2007, Google published a report analyzing 100,000 disks, and the most striking conclusion was: 36% of damaged disks had zero counts for all SMART variables before they failed—not a single sign was given. For sudden failures like a magnetic head hitting a disc or a motor jamming, the counter simply can't keep up with the accounts. So SMART is a routine check-up of two minutes each month, not an insurance policy. It's about 'catching the more than 60% of risks,' and the remaining 30-plus percent is just backup. My account has three layers: the 3-2-1 backup principle the foundation, the cold backup disk archiving, and the SMART monthly review as a sentinel — Prices have value; you can refer to the about selecting properties if you want to buy them. Data is priceless, so don't reverse-order this order.

finish with a complete order, so you won't miss anything: first look at the overall overall review line — then go through the absolute values of Pending, Reallocated, UDMA_CRC, temperature, and power on time — RAW interpretation — check VALUE versus THRESH — archive screenshots for next month's trend comparison — run short-term tests if suspicious — run long-term tests before market change. In just two minutes a month, my cousin's red plate, which was wrongfully accused of '2.1 billion,' is still saving photos for him.