Recently, we have been getting many questions about using the OneDrive sync app to pull back cloud-based data so it can be incorporated into on-premise backups.
In fact, many people have gone ahead, installed the app, and synced back their files, only to find that BackupAssist Classic won’t show these folders in the backup selection screen. No error. No warning. They’re just missing.
The short version is this: using the sync app for backup purposes is a bad idea. We will explain why by going into the technology. But if you need a quick summary, these are the three things you need to know.
- BackupAssist Classic can only back up data that is actually present on a machine or network share.
- Modern sync apps like OneDrive use a feature called Files On-Demand. What exists locally is a placeholder pointing towards the cloud copy — the data still sits in the cloud. A backup will only capture the pointer, not the data itself. Even forcing the download by marking files as always available isn’t always guaranteed, as we explain below.
- The real solution is to use BackupAssist 365 to pull your cloud data down to your on-premise machine and incorporate it into your backups.
How modern OneDrive actually stores files
When you use OneDrive with its default settings, you’re using a feature called Files On-Demand — implemented via the Windows Cloud Files API (CFAPI). Instead of downloading your entire OneDrive library to local storage, Windows creates placeholder files that look real but aren’t fully stored on your machine.
It’s worth pausing on what OneDrive actually is. It’s a sync and collaboration tool — not a backup application. It was designed to make files accessible across devices and to enable sharing, not to create independent, recoverable copies of your data. The Files On-Demand feature reflects that design intent: keep the appearance of everything being available locally, while actually storing the data in the cloud and only downloading it when needed.
Each file in your OneDrive sync folder has one of three states:
- Online-only — the file exists in the cloud; only a placeholder is on your local drive
- Locally available (cached) — the file has been accessed recently and is temporarily cached on disk
- Always available (fully pinned offline) — the file has been explicitly marked to always remain on the local drive
This is by design. Files On-Demand is a sensible feature — it lets you see your entire OneDrive library without using local disk space for files you rarely need. The problem arises when backup software tries to treat these placeholders as real files.
What a reparse point is — and why it matters
The placeholder files OneDrive creates are not ordinary files. They are implemented as NTFS reparse points — a special filesystem object with a tag that tells Windows:
“Don’t treat this like a normal file — hand it off to a filter driver.”
For OneDrive specifically, that tag is `IO_REPARSE_TAG_CLOUD`. When Windows sees a file with this tag, it knows to route any read request through the OneDrive filter driver, which then decides whether to serve the file from local cache or download it on the fly from the cloud.
BackupAssist Classic does not back up NTFS reparse points at a file level. This isn’t an oversight — reparse points require specific handling to interpret correctly, and without it, attempting to back one up produces either nothing useful or a meaningless stub.
What’s actually stored locally
When a file is online-only, here’s what exists on your local disk:
Stored locally: – File metadata (name, size, timestamps) – A small stub — often 0 bytes or minimal in size – A reparse tag and cloud file identity reference
Not stored locally: – The actual file contents (until the file is accessed and downloaded)
This creates a specific and important problem for backup. BackupAssist Classic can only back up data that is actually present on the machine — and for an online-only file, the data isn’t present. Windows Explorer reports the file as existing, with its correct name, size, and modified date. But that reported size is not backed by actual data on disk. A file that Windows reports as 2 GB may physically occupy just a few kilobytes of local storage — or less. The rest is in the cloud, retrieved on demand when you open the file.
Why “the file looks like it’s there” is misleading
This is the core of the problem.
Windows Explorer shows you: – The file name – The reported file size (from cloud metadata) – The modified date
For an online-only file, none of that means the data is on your disk. The size is a logical value pulled from the cloud — it tells you how large the file is in OneDrive, not how much of it is actually stored locally. A folder that appears to contain 50 GB of files may have almost nothing on local storage at all.
If backup software naively reads those placeholders without understanding what they are, it captures the stub and the metadata — not the file. You’d have a backup that appears complete but contains nothing recoverable.
Why BackupAssist Classic doesn’t show these folders
Since BackupAssist Classic does not back up NTFS reparse points, the only way to capture OneDrive files through Classic would be to force-download every file to local storage first — fully pinning the entire library offline. That may work for some environments, but it reintroduces exactly the storage overhead that Files On-Demand is designed to avoid, and it’s not reliable at scale (as we’ll explain below).
Rather than show these sync folders in the file selection screen and risk giving you the impression that your OneDrive data is protected when it isn’t, we exclude them. An empty selection is more honest than a misleading one.
This is not a gap we left by accident. It reflects a deliberate decision: if we can’t back something up reliably, we won’t claim that we can.
Why the sync app is not a reliable backup path
We do not recommend using a sync app for backup purposes. If sync meant mirror, then in general it would be okay. But the problem is sync apps do unexpected things under the hood.
It is worth understanding the history here. The previous OneDrive for Business sync client — Groove.exe, which Microsoft retired in 2021 — had a hard limit of 20,000 items across all synced libraries. Hit that ceiling, and sync simply stopped. Because servers generally run unattended, this means that if you were relying on the sync folder for your backup, you could have an incomplete backup you did not even know about.
The current OneDrive sync client does not specify what its hard limits are. But equally worrying is the fact that Windows’ Storage Sense feature can silently dehydrate locally cached files when disk space runs low — meaning files that appeared to be fully local get converted back to online-only stubs without any notification.
For backing up business data, this is simply unacceptable. It does not provide the level of predictability that a backup strategy requires. The backup may be incomplete, it happens silently, and the first time you find out is when you try to do a restore.
The solution: BackupAssist 365
The goal of pulling down Microsoft 365 data and including it in the on-premise backups is sound. The problem is that the OneDrive sync folder is not the path to get there.
The right solution is to bypass the local sync folder entirely and go directly to the source. Pull back the data so that it lives locally, and then decide how many additional replicas or backups are required based on the business requirements.
BackupAssist 365 does exactly that. It connects to your Microsoft 365 tenant via Microsoft’s own APIs and downloads your data directly to your chosen local backup destination — no sync folder required, no dependency on what happens to be cached on a particular machine. The backup reflects what’s actually in the cloud, not the state of whichever device ran the backup job.
How it works
BackupAssist 365 is a separate product from Classic — its own installer, its own interface, its own licensing. If you’re running Classic for Windows Server and workstation backup, you run both products alongside each other. They address different problems and don’t overlap.
Key things to understand about BackupAssist 365:
It covers mailboxes, SharePoint, and OneDrive. All three are downloaded to your local storage — mailboxes to PST files (a standard format recoverable without BackupAssist), and SharePoint and OneDrive to a plain folder structure that mirrors your cloud layout. You can browse the lot directly in Windows Explorer without any BackupAssist tool installed.
No special storage infrastructure required. The backup destination is a plain file system — no S3-compatible storage, no proprietary appliance, no vendor-managed cloud destination. If you have a local drive or a network share, you have everything you need. And because there’s no proprietary format to decode, you can access your files directly in Windows Explorer without BackupAssist installed.
Selective restore is included. A restore tool is available for both bulk and granular restores. You don’t need to restore everything to recover a single file or folder.
Microsoft’s retention policies are not a substitute for this. Microsoft’s built-in retention and recovery features have real limitations in duration and scope, and they don’t give you an independent copy of your data that you control. BackupAssist 365 gives you that copy, stored on your own infrastructure, accessible without Microsoft.
MSP visibility is included. If you’re managing multiple tenants, BackupAssist 365 can be monitored through the Managed Backup Console alongside your Classic installations — one place to see the backup status of both your on-premise and Microsoft 365 environments.
For more information on BackupAssist 365, visit the product page.
