On the night of March 9-10, 2021, a fire ravaged the OVHcloud SBG2 datacenter in Strasbourg. Within hours, 30,000 servers were destroyed, 3.6 million websites went down, and thousands of businesses discovered that their backups had burned along with their production servers. This disaster remains the largest datacenter incident in Europe. Here is what it taught us — and why we designed the Double Drive PBS offering so this scenario never repeats for our customers.
March 10, 2021: timeline of a disaster
The OVHcloud campus in Strasbourg comprised four buildings: SBG1, SBG2, SBG3, and SBG4. On the evening of March 9, abnormal humidity readings were detected in the UPS (uninterruptible power supply) room of SBG2. Shortly after midnight, the fire alarm triggered.
Fire alarm in SBG2's power supply room
Emergency services called — the fire was already out of control
SBG2 is completely destroyed; fire has spread to SBG1
Fire contained after 6 hours of fighting, 100 firefighters deployed
The outcome was devastating: SBG2 was entirely destroyed (5 floors, 500 m² of servers). SBG1 was partially burned — 4 out of 12 rooms destroyed — and was ultimately declared unrecoverable on March 22 after a second incident (smoking batteries). SBG3 and SBG4, shut down as a precaution, were spared but remained offline for days.
The BEA-RI report identified critical aggravating factors: no automatic fire suppression system in SBG2, a wooden ceiling with only one hour of fire resistance, and a chimney effect between floors that accelerated the spread.
The extent of the damage: far beyond a simple outage
30,000
physical servers destroyed
3.6M
websites unavailable
464,000
domain names affected
12,000+
clients directly impacted
Among the victims: data.gouv.fr (the French government's open data platform), the Centre Pompidou, the European Space Agency, dozens of French local authorities, the game Rust (25 European servers lost with no possible recovery), and thousands of SMBs and associations.
More than 140 clients filed a class action seeking over 10 million euros in damages. The Lille commercial court ordered OVHcloud to pay over EUR 400,000 to two companies, ruling that the hosting provider had breached its contractual obligation by keeping backups at the same physical location as production.
The real problem: production and backup in the same place
The most painful lesson from the Strasbourg fire is not technical — it is organizational. Thousands of clients had subscribed to OVH's backup option. The problem: the backup servers were hosted in another room within the same datacenter. When the fire swept through the building, production and backups burned together.
Quote from Octave Klaba, OVH founder, March 10, 2021:
"We recommend to activate your Disaster Recovery Plan." — A message that sent chills down the spine of everyone who didn't have one.
The 3-2-1 backup rule (3 copies, 2 different media types, 1 offsite copy) was followed by almost none of the affected clients. Many believed that "cloud" meant "protected." They discovered, too late, that a datacenter is a physical building, subject to the same risks — fire, flood, disaster — as any other infrastructure.
France's data protection authority (CNIL) also pointed out that the destruction of personal data, even accidental, constitutes a data breach under the GDPR, with mandatory notification within 72 hours. Liability falls on the data controller — that is, the client, not the hosting provider.
After the disaster, Octave Klaba stated that this incident would "change the industry standard" and that OVH would make backups "free and enabled by default." This is an admission: the market was not protecting its customers enough. We decided to do better from the start.
Our response: Double Drive PBS — 2 sites, 0 single point of failure
The OVH SBG2 fire is exactly the scenario that motivated the creation of our Double Drive PBS offering. The principle is simple: your Proxmox backups should never depend on a single geographic point of failure.
What we provide concretely
- 2 dedicated Proxmox Backup Servers, each on a separate physical site
- Separate power grids — no shared power supply between the two PBS
- Minimum distance of 5 to 6 km between the two sites, to eliminate any risk of a disaster affecting both locations simultaneously
- Need more distance? We can adapt the configuration to your DRP (Disaster Recovery Plan) requirements
In practice, if a fire, flood, or any other disaster destroys one of the two sites, all of your backups remain intact and accessible on the second PBS. No surprises, no "activate your DRP" — your data is already safe elsewhere.
Two usage modes: you choose
With the Double Drive PBS offering, you get two independent PBS servers. You decide how to use them, according to your backup strategy.
Mode 1: Two independent backups
Your Proxmox infrastructure sends two separate backup jobs, each to a different PBS. Each PBS receives its own complete copy, with its own retention points and its own pruning policy.
Proxmox VE (your infra)
--- Backup job 1 → PBS Site A (datacenter 1)
--- Backup job 2 → PBS Site B (datacenter 2, 5+ km away)
Why this is our recommendation:
- No dependency between the two PBS — if one goes down, the other continues to receive backups normally
- Two different retention policies possible — for example, 7 days on site A (fast restore) and 30 days on site B (archiving)
- Double ransomware protection — PBS keeps previous snapshots (point-in-time restore), and with 2 independent PBS, a compromise of one does not affect the other
- Cross-verification — you can compare the two copies if you have doubts about a backup's integrity
Mode 2: Backup + PBS replication
Your Proxmox sends its backups to a primary PBS. The second PBS is configured for replication: it automatically syncs the content of the first, without any action on your part.
Proxmox VE (your infra)
--- Backup job → PBS Site A (datacenter 1)
↓ automatic replication
PBS Site B (datacenter 2, 5+ km away)
When to choose this mode:
- Maximum simplicity — only one backup job to configure on the Proxmox side
- Limited bandwidth — data only transits once from your infra; PBS-to-PBS replication happens between our datacenters
- Existing infrastructure — you already have a PBS backup job configured and don't want to duplicate it
In both modes, PBS native deduplication applies on each server. You don't pay double for storage: only actually different blocks consume space.
If OVH clients had Double Drive PBS
Let's replay the March 2021 scenario and apply it to our architecture:
OVH SBG2 scenario
- Production and backup in the same building
- Fire → everything burns at the same time
- Data irretrievably lost
- No DRP possible without external copy
- Mandatory CNIL notification within 72 hours
With Double Drive PBS
- 2 PBS on sites 5+ km apart
- Fire at site A → site B intact
- Immediate restore from the surviving PBS
- RTO of a few hours, RPO based on your frequency
- PBS snapshots: roll back in time if ransomware
- Business continuity preserved
Going further: air-gapped, LTO, and bank vault
Double Drive PBS protects against physical disasters. And thanks to the snapshots kept by Proxmox Backup Server, you can restore a point prior to a ransomware attack — within your retention policy limits. The longer the retention, the more leeway you have to detect and recover.
To go even further (physical isolation, 30+ year archiving, bank vault), NimbusBackup offers a complete range of 6 PBS offerings:
Single Drive PBS — EUR 12/TB/month
PBS backup on mechanical disk, 1 site in a French datacenter. The entry-level option for offsite Proxmox backups outside your infrastructure.
Double Drive PBS — EUR 22/TB/monthTopic of this article
Geo-replication on 2 separate sites (5+ km). The direct answer to the OVH SBG2 scenario: if one site burns, the other is intact.
AirGapped Drive PBS — EUR 34/TB/month
Physical isolation through offline disk rotation. Anti-ransomware protection via air-gap — even a network attacker cannot touch disconnected disks.
Drive Bank PBS — EUR 69/TB/month
Air-gapped disk stored in a bank vault. Maximum physical security, NIS2 compliance.
Magnetic PBS — EUR 89/TB/month
HDD + automatic archiving to LTO tape retained for 30+ years. Dual protection: mechanical disk + magnetic tape.
Magnetic Bank PBS — EUR 149/TB/month
The highest level of protection: LTO archiving in a bank vault. Designed for critical infrastructure operators and sectors with maximum regulatory requirements.
Check out our complete guide to Proxmox backup pricing to compare all offerings in detail, or discover all PBS offerings on our dedicated page.
Don't replay the OVH SBG2 scenario
Protect your Proxmox backups on 2 sites from EUR 22/TB/month. AES-256 encryption, French support, no egress fees.
The OVH fire changed the rules: follow them
Five years after the Strasbourg fire, the lessons are clear. A local or single-site backup is not a backup — it's a copy that will burn with the rest. The 3-2-1 rule is not a luxury, it's the bare minimum.
With Double Drive PBS, you apply this rule natively: two PBS, two sites, two power grids, zero single point of failure. And if you need to migrate to multi-datacenter Proxmox, RDEM Systems handles everything — including the on-call cost, far less than a disaster. And if tomorrow you need to increase protection — air-gapped, immutable, bank vault — the offering scales with you.
Sources and references
- Data Center Knowledge — Fire has destroyed OVH's Strasbourg Data Center SBG2
- The Register — OVH data centre destroyed by fire in Strasbourg
- The Register — OVHcloud datacenter 'lacked' automatic fire extinguishers (BEA-RI report)
- Data Center Dynamics — OVHcloud ordered to pay EUR 250k+ to customers
- Wikipedia EN — OVHcloud Strasbourg fire
- Le Monde Informatique — Panic among OVH clients lacking backups
- BEA-RI Report — Analysis of the OVHcloud Strasbourg fire (PDF)