Replication Setup

Atlas replicates completed snapshots to one or more secondary S3-compatible storage targets, so backup data can exist on multiple storage systems in multiple locations (a 3-2-1 strategy).

Replication covers all workloads. Outlook mailboxes, OneDrive owners, and SharePoint sites share the same encrypted object store and the same replication engine.

Deployment Patterns

Local MinIO primary, offsite MinIO replica

Run a primary MinIO instance on your local backup server and a second MinIO instance at a remote site (colocation, another office, or a VPS). After each backup, replicate to the offsite target:

atlas outlook backup -m user@company.com
atlas replicate -m user@company.com --target-config ./offsite.json

atlas onedrive backup -o user@company.com
# OneDrive snapshots replicate with the same tenant bucket and DEK

atlas sharepoint backup --site https://contoso.sharepoint.com/sites/Engineering
atlas replicate --site https://contoso.sharepoint.com/sites/Engineering --target-config ./offsite.json

Local MinIO primary, cloud S3 replica

Run MinIO locally for fast backups, then replicate to AWS S3, Backblaze B2, Wasabi, or any S3-compatible cloud storage for geographic redundancy:

{
  "target_id": "aws-us-east",
  "s3_endpoint": "https://s3.us-east-1.amazonaws.com",
  "s3_access_key": "AKIA...",
  "s3_secret_key": "...",
  "s3_region": "us-east-1"
}

Scheduling Replication

Pair replication with backup scheduling by adding a second cron entry that runs after the backup completes:

# Nightly Outlook backup at 2 AM, replicate at 4 AM
0 2 * * * /usr/bin/atlas outlook backup >> /var/log/atlas-outlook-backup.log 2>&1
0 4 * * * /usr/bin/atlas replicate -m user@company.com --target-config /etc/atlas/offsite.json >> /var/log/atlas-replicate.log 2>&1

The two-hour gap gives a large backup time to finish before replication starts. Adjust it to your observed backup durations.

Bandwidth Considerations

Replication transfers encrypted objects over the network. The first replication of a mailbox, OneDrive owner, or SharePoint site transfers all historical data, similar in volume to the initial backup. Subsequent replications only transfer new snapshots, because the same content-addressed deduplication that saves space on primary also means unchanged objects are never re-copied.

Plan bandwidth and scheduling accordingly: replicating 50 GB of backup data to a remote site over a 10 Mbps upload link takes approximately 11 hours.

Full Operational Detail

This page covers the self-hosting setup perspective. For the complete reference, including the shared encryption model, DEK mismatch protection, replica markers, the atlas rehydrate disaster recovery procedure, and replication status tracking, see Replication.