Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Backup and Recovery

Strategies for backing up MediaGit repositories and recovering from failures.

What to Back Up

A MediaGit repository consists of:

.mediagit/
├── objects/          # Content-addressable object database (chunks, trees, commits)
├── manifests/        # Chunk manifests (postcard format) — index of all chunks per file
├── refs/             # Branch and tag references
├── HEAD              # Current branch pointer
├── config.toml       # Repository configuration
└── stats/            # Operation statistics (non-critical)

The objects/ directory is the most important — it contains all file content. manifests/ and refs/ are smaller but required for correct operation.

Backup Strategies

1. Full Repository Archive (Local)

The simplest backup: archive the entire .mediagit/ directory.

# Create timestamped backup
tar czf mediagit-backup-$(date +%Y%m%d-%H%M%S).tar.gz .mediagit/

# Verify backup integrity
tar tzf mediagit-backup-*.tar.gz | head -20

Schedule daily backups with cron:

# /etc/cron.d/mediagit-backup
0 2 * * * /usr/bin/tar czf /backups/mediagit-$(date +%Y%m%d).tar.gz /path/to/repo/.mediagit/

The safest backup is a live remote that always has the latest state:

# Configure a backup remote (different server or cloud account)
# In .mediagit/config.toml:
[remotes.origin]
url = "http://primary-server.example.com/my-project"

[remotes.backup]
url = "http://backup-server.example.com/my-project"

# Push to both
mediagit push origin main
mediagit push backup main

For cloud storage backends, the provider handles redundancy. S3 and Azure Blob have 99.999999999% (11 nines) durability by default.

3. Sync to Cloud Storage

If you use a local filesystem backend and want to also back up to cloud storage, sync the .mediagit/objects/ directory:

# Sync to S3
aws s3 sync .mediagit/objects/ s3://my-backup-bucket/mediagit-objects/

# Sync to Google Cloud Storage
gsutil -m rsync -r .mediagit/objects/ gs://my-backup-bucket/mediagit-objects/

4. Incremental Backup

Because MediaGit uses content-addressable storage, objects are immutable once written. Incremental backup is straightforward — only new files need to be copied:

# rsync is efficient: only copies new objects
rsync -av --checksum .mediagit/objects/ backup-host:/backup/mediagit-objects/

Before Critical Operations

Always verify repository integrity and create a backup before major operations:

# Verify integrity
mediagit fsck

# Create safety snapshot
tar czf .mediagit-backup-$(date +%Y%m%d-%H%M%S).tar.gz .mediagit/

# Proceed with operation
mediagit rebase main

Recovery Procedures

Restore from Archive

# Remove damaged repository
rm -rf .mediagit/

# Restore from backup
tar xzf mediagit-backup-20260101-020000.tar.gz

# Verify restored repo
mediagit fsck
mediagit log --oneline -5

Restore from Remote

If your local repository is damaged but the remote is intact:

# Re-clone from remote
mediagit clone http://media-server.example.com/my-project my-project-restored
cd my-project-restored
mediagit fsck

Repair Corrupt Objects

MediaGit’s fsck can attempt automatic repair:

# Check what's damaged
mediagit fsck

# Attempt repair
mediagit fsck --repair

# If objects are missing, fetch from remote
mediagit fetch
mediagit fsck

If the remote has the correct objects, a fetch followed by fsck will usually restore a damaged local repository.

Recover Deleted Branch

Deleted branches may still be reachable via the reflog:

# Show recent ref history
mediagit reflog

# The deleted branch's last commit will appear
# Recreate the branch
mediagit branch create recovered-branch <commit-hash>

Lost-Found Recovery

If objects become dangling (unreferenced) after a failed operation:

# Find dangling objects
mediagit fsck --dangling

# Save them to lost-found
mediagit fsck --lost-found

# Inspect what was saved
ls .mediagit/lost-found/

Disaster Recovery Plan

For production environments managing large media repositories:

  1. Regular pushes to a remote on a different server or cloud provider
  2. Daily archives of .mediagit/ directory stored offsite
  3. Weekly mediagit fsck to detect corruption early
  4. Monthly restore test — actually restore from backup to confirm it works
  5. Document recovery procedures for your team

Estimating Backup Size

The backup size equals the repository’s stored object size. Check with:

du -sh .mediagit/objects/
du -sh .mediagit/manifests/
du -sh .mediagit/   # total

For cloud backends, the objects are already stored remotely — only the .mediagit/refs/, .mediagit/HEAD, and .mediagit/config.toml need local backup (these are typically a few KB).

See Also