I used to think data loss was something that happened to other people — the ones who didn't know better, who didn't have backups, who clicked the wrong button. Turns out, I was exactly that kind of person.

Before the disasters, my critical data followed the "1+1" pattern: one local copy, one OneDrive copy. And SOMETIMES, the ever-benevolent Microsoft would remove my local copy for no discernible reason. So mostly, there was only one copy of my data — floating somewhere on OneDrive's servers. What's worse, the OneDrive I own is an E5 subscription, and the stupid behemoth has started banning accounts' SharePoint access at random. Just a few days ago, my account was banned. Archives, photos, chat histories, source code — all of it, quite literally, dead.

The Loss of OneDrive

The story begins in October 2025. I came up with a freaking cool, wild, and frankly stupid idea: a hybrid storage system. OneDrive as cold storage, my Storage VPS as hot storage. I built a whole toolkit — a Mover, a VFS, a WebDAV server. Data encrypted with Rclone, mounted as a local drive, accessible from anywhere.

It worked beautifully. For about two weeks.

Then SharePoint went READ ONLY.

With Gemini's help, I unlocked the site and decided to ignore the warning signs. Perhaps that was the first mistake. Perhaps I was just too tired to care.

You may try this script which I used to unlock the site:

# 1. 配置你的管理中心 URL (根据你之前的截图)
$AdminUrl = "https://akiwork-admin.sharepoint.com"

# 2. 连接服务 (如果已经连接可以跳过，但为了保险建议重连)
Write-Host "正在连接到管理中心..." -ForegroundColor Cyan
try {
    Connect-SPOService -Url $AdminUrl -ErrorAction Stop
    Write-Host "连接成功！" -ForegroundColor Green
}
catch {
    Write-Error "连接失败，请检查 URL 或网络。错误信息: $_"
    return
}

# 3. 获取租户内所有站点（包括 OneDrive 和 SharePoint 站点）
Write-Host "正在扫描租户内的所有站点，这可能需要一点时间..." -ForegroundColor Cyan
# -IncludePersonalSite $true 表示包含 OneDrive 站点
# -Limit All 表示不限制数量
$AllSites = Get-SPOSite -Limit All -IncludePersonalSite $true

# 4. 循环检查并解锁
$LockedCount = 0
foreach ($Site in $AllSites) {
    # 只处理状态不是 "Unlock" 的站点
    if ($Site.LockState -ne "Unlock") {
        Write-Host "发现锁定站点: $($Site.Url)" -ForegroundColor Yellow
        try {
            # 执行解锁命令
            Set-SPOSite -Identity $Site.Url -LockState Unlock -ErrorAction Stop
            Write-Host "  -> 已成功解锁" -ForegroundColor Green
            $LockedCount++
        }
        catch {
            Write-Host "  -> 解锁失败: $_" -ForegroundColor Red
        }
    }
}

# 5. 总结
Write-Host "----------------------------------------"
if ($LockedCount -eq 0) {
    Write-Host "扫描完成。没有发现其他被锁定的站点，你的租户现在是健康的。" -ForegroundColor Green
}
else {
    Write-Host "操作完成。共尝试解锁了 $LockedCount 个站点。" -ForegroundColor Green
}

A couple of days later, my laptop died. Cold solder joints on the CPU and southbridge. I couldn't find anyone in Wuhan to fix it, so I hauled it back home for an acquaintance to repair.

After it was fixed, OneDrive still couldn't write. Error `0x8004deef`. I Googled it.

Banned.

Not "temporarily suspended." Not "please verify your identity." Just... banned.

I assumed my hybrid system was to blame — 600GB+ uploaded in a short burst. But after browsing the Nodeseek forum, I found the real culprit: logging in from a Chinese IP is apparently enough to get your account nuked. So, fuck it. I gave up the data on OneDrive, copied whatever remained on my laptop, and reinstalled Windows.

Two days later, the laptop died again. Same symptoms. Again. The universe, apparently, had notes.

I've lost everything from 2019 to 2025. The only hope is my old NAS — and even that feels like a long shot.

Building a New NAS

After the Read-Only incident, even though I wanted to pretend nothing happened, I bought a new NAS. QNAP this time. A secondhand TS-464C2, snagged for less than 1k RMB just before RAM prices rocketed. But SSDs and HDDs had already gone wild by the time I was making up my mind to build a NAS in my dorm. A recertified 14TB HC530 for 1.4k RMB squeezed my wallet dry.

This is my first prebuilt NAS. I've run DIY setups for years — Unraid, hacked Synology, the works. It's also my first time using QTS.

QTS is slow. Click a setting? Wait. Save it? Wait again. Every interaction feels like negotiating with an unseen bureaucracy, each action requiring its own form in triplicate.

But here's the thing — file transfers are fast. Really fast. 2.5 GbE and kernel-space SMB do their job. The OS is a nightmare; the hardware is a dream.

The HARD Way

So how do I store data the HARD way?

Hard as in hard drives. Hard as in hard to lose. Hard as in... this has been really fucking hard.

Multiple copies matter. I've got the hard disk. I've got infrastructure. But QTS's backup feature — HBS — is a pain in the ass. Small backups? Fail. Large backups? Fail harder. HBS insists on running database operations on the remote end, as if latency and packet loss are myths. I'm trying to deploy Duplicati instead.

For now, I'm running a single drive without RAID. RAIDZ comes later, when drive prices drop — whenever that may be.

Speaking of RAID failures: [Jimmy Wei](https://lo-li.icu)'s DIY NAS broke down recently too. His system died mid-upgrade, transitioning from RAID 1 to RAID 5 with BTRFS. I suspect his "brand new" secondhand drives were the culprit. NVM.QTS runs on ext4, but I really want CoW. And I also want RAID 5.

Can't stop, won't stop. I want to migrate to QuTS Hero, embrace ZFS, set up RAIDZ properly. But all of that requires more drives — and more money.

So for now, I wait. Still that person who doesn't know better — just slightly more aware of it.