Anubis Market: the live mirror reference
Three production mirrors on rotation. The Primary handles bulk traffic, Backup A absorbs spillover, Backup B is the explicit failover.
Anubis runs three production v3 onion mirrors in parallel. The architecture is intentional. The Primary mirror takes the bulk of buyer traffic and lives behind the platform's primary anti-DDoS challenge layer. Backup A absorbs spillover during peak windows. Backup B is the explicit failover, on lower-throughput guard relays with intentional latency budget. Live table with copy buttons is on the Anubis profile page.
Why three
Two reasons. First, congestion isolation. A flood targeting one address does not affect sessions on the other two. Second, rotation continuity. The platform can stage and verify a new mirror before retiring an old one, which means zero-gap rotation across operator-side changes.
Which to use
Primary first. If the challenge layer times out, Backup A. If both fail, Backup B. If all three fail, the platform is dealing with an active flood; rotate your Tor circuit and try again in a few minutes. The atlas does not republish addresses that fail our prober; entries you see on the profile page are operationally current.