Should a new NAS hard drive undergo a full inspection? I tested it in 18 hours to write zeros; these three types of people really don't need to learn from me

recently bought a 4T Cool Wolf for my old Feiniu to save trouble. Unboxing SSH, I checked Smartctl, and it powered on for 9 hours, the model matched perfectly. I thought the new drive was stable, so I just formatted and uploaded the data. On the third day, when I checked the backend, a 1 appeared Current_Pending_Sector. It was just a pending sector, and not a single file was broken, but I felt uncomfortable for a whole week. Later, I added a second drive to the small case. This time, I diligently idled for 18 hours with badblocks writing zero, and only felt reassured when I read the line "0/0/0 errors" after running. Is new property inspection really worth it, and whether you should do it? This article explains everything thoroughly.

new hard drive can break too, not just to scare people

many people think that brand-new tape seals have anti-counterfeiting protection, and if they break, it's unlikely to be their turn. There is a term for mechanical disc failure called the bathtub curve: the failure rate is high in the first few months, then bottoms out in the middle few years, and then lifts up again toward the end of its life. The opening section is called the early expiration period—edge components that the factory didn't screen out thoroughly, weak sectors on the platter that weren't exposed are basically exposed within the first few hundred hours. Manufacturers only perform random inspections and short-term tests after delivery. A 4TB drive has over 8 billion sectors, so it's impossible to write and read every single disk. The chance of this loophole is left to you.

there is another layer for transportation. The hard drive is rated to withstand 250G shocks when not running and 70G when running. It sounds solid, but that's the lab's pulse limit—after loading and unloading packages or constant bumps on the road, weak sectors are created by vibration. So the essence of a "new listing inspection" is one thing: to fill in any unfinished inspections by the manufacturer yourself, so that the problematic listings are exposed during the warranty period and before you store your data. You don't even have to pay for the shipping for a new one, just ask if it's worth it to find out what you need.

skeptical, you can check Backblaze's annual hard drive failure statistics. Their data centers run with tens of thousands of yuan in storage, but the yearbook shows that the annualized failure rate in the first few months of new drives is higher than the second year. If the data center environment is stable and humid, the bumps from your parcel locker to the hallway will only be more exciting. Encountering a weak sector isn't just bad luck; it's the probability that it's already there. The difference is whether you catch it before or after you log in.

three types of testing, the severity is far from the same

The

phrase "test it" actually has three levels, and the results are completely different:

my pending disk is a living example: after getting it, I passed in two minutes, I slacked off and just uploaded the data. Later, after reviewing it, I took it apart and wrote zero. When it ran over 60%, I read back to verify a sector — this part reads passable, but if you write it in and read again, it drifts. For these weak sectors with 'poor writing ability,' both short and long are blind, and only writing zero can force them out. That's why I later only recognized bad blocks.

badblocks write the full process for zero, and those who can wait will run

to say the first thing: writing zeros will clear the entire disk's data, so you can only do it on the empty disk; If you break midway, you don't stop and keep running; if you do, start over from the beginning. Pick a weekend to AFK and don't rush. Three specific steps:

Step one: Recognize the disc symbol. When entering SSH, first check which part is the new drive. If you misread the drive letter, writing zero is the data drive. There's no cure for this kind of accident. To be safe, you can unplug the data drive after powering down, leaving only the target drive in the machine, making it hard to write errors.

Step two, run the command:

badblocks -wvs -t 0x00 /dev/sdb

-w writes tests, and -t 0x00 specifies that only zero is written. By default, -w without -t will write 0xaa, 0x55, 0xff, and 0x00 modes to verify once. 4T requires two days and nights to run, which is unnecessary for home use. Writing zero once is enough to force out weak sectors. -v shows progress and error count, -s gives a progress bar; otherwise, if the screen stays black for over ten hours, you feel uneasy.

Step 3, etc. I wrote and read the 4TB CMR drive for 17 and a half hours, AFK at 10 p.m., and finished at 3:30 the next afternoon. Tile-stacked SMR disks are directly discouraged. Writing that scaling up can let you run for more than 30 hours, and when buying, you should the two checkpoints in the replacements article confirm the CMR. During the drive, I turned off the hard drive for sleep; otherwise, the drive would crash as soon as it fell asleep. If the machine already has the old problem of stealing the writing drive, this time it's a good time to check it out together.

finish the run, check the final output: "Pass completed, 0/0/0 errors" means clearing all. Even if you do report an error, don't panic. Note the block number, run smartctl -a again to see if Reallocated and Pending have increased—modern hard drive firmware automatically remaps bad sectors, how to interpret red and yellow tags I wrote about this in the article about bad disk signs. New releases can be remapped, so don't hesitate or go straight for after-sales replacement.

I've stepped through these pitfalls for you during the inspection

the first pitfall, I checked it with a USB hard drive enclosure. Many people first insert a new drive into the enclosure to test on their computer, which is a valid approach. However, USB bridge chips often block SmartCTL commands halfway, resulting in a bunch of Unsupported messages. In this case, adding a -d SAT parameter to convert SCSI to ATA is sufficient. Badblocks running inside a USB box can easily disconnect midway. Don't rush to slaughter an error halfway through; just disconnect it and connect directly to SATA and run again. It's probably the box's fault.

second pitfall: during testing, they were reluctant to stop their business. Both the long and write zeros are at full load for read/write, with the magnetic head arm scanning the whole time. At this moment, you can use Jellyfin to watch 4K, and compete for both transcoding and disc reading. Don't blame the disk for the lag in the PPT. My approach is to hang it up before bed and check the results in the morning, so the NAS doesn't interfere at all with what it should do. At zero time, the disk temperature is about seven or eight degrees higher than idle. The case is stuffed inside the cabinet, so open the side panel first.

Synology Feiniu's built-in scanner—is it enough

don't want to touch the command line, the system also has ready-made ones. Synology has two levels: quick test and full test in the storage manager, HDD/SSD, selected hard drive, health info, and test paths. The full test layer is smartctl long, running the 4TB in the background for six to seven hours, during which the NAS is still frequently used. Feiniu selects the hard drive in storage management for health checks, with the same logic.

but be clear about the graphical ceiling: they only read, not write. It can catch the bad things you can't read, but the darkest weak sector of 'writing and reading back' is helpless. So my approach is a compromise: when you get a blank drive, run it overnight with the system's built-in test. If it reads fine, and the drive is for installing renewable movie downloads, just go straight in; If you want to put photos and documents—things that are lost and gone once lost—make up for another night and write zeros. It sounds exaggerated, but AFK and running it is actually doing whatever you need to do, much easier than pending on the third day and messing with data again—after all, important data is always backup detection is just less trouble for yourself.

do it or not: Take it personally

one more thing about building RAID: a sick disk sneaks into the array. Normally, when there's nothing going on, you wait for a disk to retire and trigger rebuild. If the sick disk runs under high load for over ten hours, it's easy to crash in a chain of events. How exhausting is the rebuild window I did the math in that RAID article. If you're buying several units at once, don't skip this step.

my complete order compressed into five lines: unboxing smartctl info, checking power hours and firmware, and installing a refurbished drive; Empty disk hangs overnight with a long thread, reading block error; For important uses, hang it overnight to write zeros, block weak sectors; After everything is done, format mount the data ; Prepare an extra copy of important documents for the first two weeks. 18 hours sounds scary, but in reality, it's just two nights of AFK and less than one yuan for electricity. Compared to staring blankly at the pending on the third day, this money was definitely worth it.