Why Did Backblaze Stop Backing Up My Git, OneDrive, and Dropbox Folders?

# Why Did Backblaze Stop Backing Up My Git, OneDrive, and Dropbox Folders?

Backblaze didn’t “lose” your data so much as its desktop client changed what it considers safe—or even possible—to back up: recent versions began skipping cloud‑synced folders and special filesystem objects (including .git directories and files marked with reparse points/placeholder metadata). Because these items can look normal in Finder or File Explorer while not being “plain local file data,” many users only noticed the gap when they tried to restore and discovered whole folders (or critical repo history) weren’t in the backup.

Quick answer: what happened and why

In late 2025, users and independent bloggers reported that Backblaze’s desktop client started excluding data stored inside common cloud‑sync folders—especially OneDrive and Dropbox (with some reports also mentioning services like Google Drive and Box)—and that some users found .git directories were not being backed up. One observer traced the behavior to a client release observed as v9.2.2.878, and multiple accounts note the change was not broadly or prominently announced, leaving people with apparently “successful” backups that were incomplete.

Backblaze’s explanations point to platform-level constraints. On macOS, Backblaze’s help/support guidance has said (paraphrasing that guidance) that recent macOS/iCloud changes can prevent backing up iCloud‑managed files even when they appear downloaded. On Windows, similar issues arise from how cloud sync tools use NTFS features (notably reparse points) to represent cloud‑managed or “online‑only” files. In practice, the Backblaze client appears to treat these flagged items as excluded, so they never enter the backup set.

Technical background: reparse points, placeholders, and .git

To understand why a backup tool might skip “files you can see,” you need one key concept: a file listing doesn’t guarantee local file bytes are present.

• Reparse points are an NTFS filesystem attribute on Windows that can redirect or alter how file access works. Cloud sync clients use them to implement “online‑only” behavior.
• Placeholder files (often called “online‑only”) appear in folder views, but their full contents may not be stored locally. The sync client fetches data on demand.
• Some cloud tools can also mark files in ways that backup software treats as nonstandard, even when a file appears to be available.

Then there’s .git, the hidden directory inside a Git repository that stores history and metadata. If a backup agent skips .git, you may still have the working files, but you lose the local commit graph, branches, and other repository state. Users reported that .git directories were skipped—an especially sharp edge for developer workflows.

Why syncing folders were excluded: product and platform dynamics

From the reported behavior and Backblaze’s own help/support explanations, the underlying dynamic looks like this:

1. Cloud-sync folders aren’t just folders anymore. They can contain a mix of fully local files and placeholders governed by system flags.
2. Backup tools need stable, readable local bytes to produce a reliable archive. If a “file” is actually a placeholder, backing it up might capture incomplete content—or trigger downloads that don’t behave consistently across OS versions.
3. OS vendors keep changing the rules. Backblaze has specifically cited (in its help/support guidance) macOS/iCloud changes that interfere with backing up iCloud‑managed items, even if they appear downloaded in Finder. Windows has its own complexity: reparse semantics and placeholder behavior can lead backup apps to skip items rather than risk inconsistency.

The user-facing problem isn’t only the exclusion—it’s the visibility. Multiple reports emphasize that the change wasn’t surfaced loudly, so people discovered it only during restores or deep inspection. That is the worst time to learn a “backup” was effectively a partial index of what you had.

For broader context on how “silent changes” can surface only after recovery attempts, see Today’s Tech Mosaic: Moon Code, Safer Web, and Surprising Hardware Wins.

Practical risks: retention gaps and recovery surprises

The biggest risks described in user reports are about false confidence:

• Silent exclusion: Backups may look healthy in the UI while key folders were never captured. Users only learn the truth during restore.
• Retention mismatch: Many sync providers retain deleted/previous versions for a limited time (often roughly ~30 days, but this varies by vendor and plan), while Backblaze has historically offered longer retention options (e.g., one year). If Backblaze stops backing up those cloud folders, you lose the longer safety net you thought you had.
• Developer workflow breakage: If .git is skipped, you can lose local history/state. Even if your code is pushed to a remote, relying on a local backup for repo recovery becomes unreliable.

Some anecdotes in the reporting are stark—like a OneDrive folder of roughly 350 GB discovered missing from Backblaze’s backup set.

Backup vs sync: why a sync folder isn’t automatically “covered”

A sync service and a backup serve different purposes:

• Sync mirrors a current state across devices. Deletions and overwrites can propagate, and retention policies may be limited.
• Backup should preserve independent point‑in‑time copies for recovery, ideally with predictable retention.

The strategy “I’ll keep everything in Dropbox/OneDrive and Backblaze will back it up too” only works if the backup tool copies real local file content. If the desktop client skips placeholders or treats cloud-managed files as non-backupable, the “independent copy” never exists.

Immediate steps: audit whether your files were actually backed up

Given the reports that users learned of gaps only when restoring, the most practical response is to validate restore-ability, not just backup status.

1. Inspect your Backblaze file listings / restore view. Check whether your cloud-sync folders appear with expected contents and sizes, and explicitly check for .git directories inside repos.
2. Look for exclusions or clues in client behavior. Users have reported patterns: skipping cloud-provider folders, skipping items with reparse-point/placeholder flags, and skipping .git.
3. Run a controlled test file. Create a small file inside the sync folder, ensure it is fully local (not online-only), force a backup run, and then verify it appears in Backblaze’s archive. Repeat with a small Git repo and confirm .git presence.

The key is to validate with a restore test, not just “it says backup completed.”

Reconfigure or switch strategies to avoid silent data loss

Based on the reporting and Backblaze’s platform-constraint framing, mitigations fall into three buckets:

• Make important sync data fully local before backup. Use your sync client’s setting to keep critical folders “always available” / downloaded so they’re ordinary local files—though the reporting notes that even “locally available” may still be flagged in ways backup tools treat specially.
• Back up via tools that explicitly handle placeholders or cloud APIs. Reported experiences vary by product: one account noted CrashPlan also stopped backing up Dropbox for some users, while Arq was noted as still able to back up in reported findings. The lesson is to verify behavior for your exact setup, not assume.
• For Git: don’t rely on a single local backup path. Push important branches/tags to a remote and consider exporting repositories (for example, using a bare repo) so your history isn’t hostage to a skipped .git directory.

If you’re re-evaluating your backup posture generally, it’s also worth building a habit of testing restores after client updates, since behavior can change underneath you.

Why It Matters Now

This issue is urgent because it combines two failure modes: silent change plus delayed discovery. People aren’t learning about missing coverage when the update installs—they’re learning when they need a restore, after a disk failure, accidental deletion, or migration.

It also matters because the drivers aren’t going away. Backblaze attributes at least part of the problem to ongoing platform changes (macOS/iCloud behavior, Windows placeholder/reparse semantics). That suggests other backup tools can be affected too, especially as OS vendors evolve cloud-managed storage features.

In other words: if your “backup plan” implicitly depends on cloud-sync folders being treated like normal folders, you should assume this could break again and design checks around it—particularly ahead of audits, device upgrades, or any moment when you’d bet real money on your ability to restore.

What to Watch

• Backblaze release notes and support guidance: Look for clearer documentation and any reversal/refinement of how the desktop client handles cloud-managed files, reparse points, and .git.
• OS platform changes: Monitor Apple and Microsoft changes affecting iCloud-managed files and Windows placeholder/reparse behavior, since these can directly impact backup reliability.
• Community verification + your own restore drills: Treat user reports as early warning, but trust only what you can validate with regular test restores on your own machine and folders.

Sources: the-online.com , geekzone.co.nz , mjtsai.com , dropboxforum.com , help.backblaze.com , byteiota.com