How do you transfer data from a new NAS hard drive? The full process of moving from 4T to 8T, I was so careless that I almost wasted my effort when I shut down

Last week, a red alert popped up in Feiniu's backend: 85% storage was used. That 4TB Cool Wolf ST4000VN006 had been running in the N100 case for over three years, with photos, video library, and download catalogs all packed together and about to fill up. Gritting my teeth, I bought an 8T Cool Wolf ST8000VN004, just over 900 yuan. Only after getting the disk did I seriously think about it: How do you transfer 4T data to the new project? How to deal with old disks? I spent two days researching and hands-on practice, falling into two pitfalls. This article will go through the entire process from the beginning.

first distinguish which method you switch; the route is completely different

Before

the first screw, first check the state of your machine; the three scenarios can be vastly different:

I have a dual-drive machine that only occupies one slot, which is the most stable first route. If your slots are full, be sure to the most important items before you start copy out according to the backup rules. Every second during the move is a copy. For RAID players, note that rebuilding an 8TB drive takes about eight hours per session. Swapping two old drives twice means waiting two rounds. Sequence and patience matter more than skill. I wrote about how to select modes based on the number of drives in the RAID article.

new property gets your hands on a new property, get through two hurdles first—don't rush to plug in and get

the first checkpoint: confirm that you are not buying a tile-stacked tray. Mechanical disks are divided into CMR (vertical) and SMR (shingle). To maximize capacity, SMR stacks tracks like tiles, causing severe slowdowns in random writes. When warehouse drives are barely managed, but when NAS main drives are used, it's a disaster. SMRs of the same capacity can be several dozen yuan cheaper, but product pages usually don't mention it; you have to search for "model + SMR" yourself. Before I bought ST8000VN004, I specifically confirmed it was CMR. The 8T capacity range is vertical across the Cool Wolf lineup, so it's fine to buy with your eyes closed. But be cautious with Western Digital Purple and some smaller capacity models from Cool Wolf.

the second hurdle, you get a physical exam before you get it. SSH runs one run into the machine:

smartctl -a /dev/sdb

look at two numbers: Power_On_Hours power-on hours, a brand-new drive should last within a few dozen hours (factory testing plus shipping), mine shows 9 hours after delivery; Electricity costs thousands or tens of thousands. Congratulations, you got a refurbished unit—get a direct refund. Looking at Current_Pending_Sector and Reallocated, both should be 0. This step takes two minutes and can block a market with a high chance of failing the next day, so don't mind the hassle. By the way, if your old home has already shown signs of abnormal noises or speed drops, you'll need to speed up your move. the five signals in the article about warning signs of a bad home, compare them.

the right path: first insert a new drive, copy it inside the machine

shut down and cut off power, insert the new drive into the empty bay, secure the SATA cable and power cable, and power on. In Feiniu's storage management, you can see an uninitialized disk. Initialize, format, and mount it. I detailed steps in the article about mounting the second hard drive. Synology creates a new storage pool in the storage manager, with the same logic.

next is the most critical decision: how to move the 3.4T data there. My first thought was to mount the NAS into a Windows computer and copy and paste from the PC—luckily, I didn't do that. Data comes from the old drive, passes through the network cable to the PC, then from the PC to the new drive via the cable. Gigabit network runs twice, but the actual speed is only 40 to 60MB/s. For 3.4TB, copying takes at least twenty hours. The correct setup is to enter the NAS via SSH and let data flow inside the machine, SATA to SATA. Tested stable at around 170MB/s, 3.4TB in just over five hours:

rsync -avh --progress /vol1/warehouse/ /vol2/warehouse/

use rsync instead of CP; the biggest advantage is that you can resume if you disconnect. The -a parameter keeps permissions and timestamps. Multiple shared directories on the NAS have different permissions, which is much more important than speeding up the minutes.

During

more than five hours of copying data, I made two mistakes

the first pit, dormant. Previously, I set the hard drive to sleep for ten minutes without reading or writing, but during copying, rsync was performing verification and there was a gap between actual read/write tests. Halfway through, both drives fell asleep, and the task got stuck for half an hour, so I only realized it. Before copying the large disk, turn off the hibernation or temporarily set it to a few hours, then switch back after the copying. If your machine already has problem of repeatedly waking up after sleep failure, this is a good opportunity to check it out.

second pit, the big player who writes the chart. qBittorrent was still running three tasks downloading at the time, copying and writing old drives at the same time. Some files changed halfway through, and rsync had to redo the second round of verification. Before starting the copy, stop downloading the container, and pause background tasks like the album index, letting the old drive quietly serve as a read-only source. If the Docker container has key data directories on the old drive, do docker stop first and then copy it, so the container doesn't run and modify files while running.

here's a little pitfall: the night I copied it to 78%, I looked at the progress bar and thought my speed wasn't affected, so I casually restarted the router—both the NAS and PC restarted, and the rsync task was gone. Fortunately, the rsync breakpoint resume wasn't just for nothing. The next day, the same command restarted, and the verified files were skipped directly, only filling in the remaining 22%. So if you disconnect midway, don't panic, and don't start over by copy-pasting—just rerun the command.

must do three inspections before removing old properties; skipping one inspection will not touch screws

copying and running doesn't mean you can disassemble the disk. Before disassembly, I checked three things:

check the capacity. the used space in the new drive's directory matches the old drive. I have 3.4TB versus 3.4TB. If there's a few hundred GB difference, quickly find the cause—most likely there's a directory you didn't copy.

Second, check the shared directory for directions. Feiniu's shared folders are bound to specific storage spaces, and my 'Warehouse' directory still points to the old drive, so I need to switch the path to the new drive in the settings or delete the old directory and create a new one. If you skip this step, the shared directory of the old drive will lose contact immediately.

Docker mount on the third check. I have a Jellyfin container that maps the media library path to an old drive. Before removing the disk, either change the container configuration to point to the new path or confirm that the path in the configuration file matches correctly. All three inspections passed: shut down the machine, cut off the power, dismantle the old drive, and wrap up.

old property, don't rush to list it as second-hand; the final step is to clean it thoroughly

I didn't sell the 4TB drive I took apart; I put it back in the enclosure as a cold backup and it hung up for two weeks—if a new drive has hidden issues, the first two weeks are the easiest to expose, and that's when the old drive becomes your last copy. If nothing happens for two weeks, just back up a separate family photo, and the drive can be handled. Before you make a move, you must wipe the entire disk; formatting is useless, but data recovery software can scan and get it right. On Windows, using diskpart's clean all command to fill in zero means the 4TB drive will run for six to seven hours; On Linux, dd if=/dev/zero takes about the same time. After wiping, I listed it on a second-hand platform. A 4TB second-hand disk costs over a hundred yuan. Isn't it better to buy a memory stick to recover your money?

Finally, compress the entire sequence into a table and follow the instructions:

the most valuable lesson in the entire process: leave data a way out. Before the new property is verified and stabilized, the old property is not moved; before the old property is cleaned up, the data does not count as the move completed. It's a bit slow, but the NAS is filled with years of photos and resources accumulated over half a lifetime. Speeding up for half a day is pointless.