// uptime · 30-day strip
99.97% uptime, 2024
One 14-minute incident on Falkenstein node bts-3. Resolved without data loss.
- 2024 incidents
- 1
- Total downtime
- 14m 22s
- Source
- pingdom + Prometheus
probe set: 4 datacenters · scrape interval: 15s
// OPERATIONAL STATS
The index, in numbers.
A public readout of the same counters our 14-person team watches from Tallinn — magnet fetches, uptime, indexer throughput, and malware blocks. Refreshed manually. No JS. No CDN cache in front of you.
12,438,602
torrents indexed
99.97%
uptime, 2024
1.8s
avg time-to-magnet
6,103,847
active torrents
// 01 / MAGNET FETCHES
2.1B magnet fetches served in 2024.
Rolling 90-day window. Human fetches (the index page itself) are separated from mirror and API pulls. Median time-to-magnet across all 4 datacenters: 1.8s.
Peak day: 2025-12-31 — 11.4M fetches (year-end archival spike)
Trough day: 2025-10-14 — 4.1M fetches
Total 2024: 2,103,447,821
// 02 / UPTIME · THROUGHPUT · BLOCKS
Three numbers the team watches every morning.
Each chart is sourced from the same Prometheus scrape the on-call engineer reads at 09:00 UTC. Hover a bar for the exact figure.
// throughput · daily
~8,000 entries/day added
Net throughput once removals (DMCA, dead torrents) are subtracted. Removals hover around 1,200/day.
- Median added
- 8,041 / day
- Median removed
- 1,188 / day
- Net growth
- ~6,850 / day
queue depth: 6 indexers · avg curation: 4.2 min / entry
// malware blocks · monthly
1.4M malware hits blocked in 2024
Tri-engine pipeline (ClamAV + VirusTotal API + 47 custom YARA rules) blocks before any torrent reaches the public index.
- ClamAV hits
- 912,400
- VirusTotal hits
- 381,022
- YARA hits
- 108,503
avg scan time: 1.4s / entry · false-positive rate: 0.03%
// 03 / METHODOLOGY
How these numbers are measured.
Every dashboard value on this page comes from the same Prometheus scrape our on-call engineer reads at 09:00 UTC. Probes run from four datacenters — Falkenstein, Strasbourg, Helsinki, and Beauharnois — at a 15-second interval with a 90-day retention window. Magnet-fetch traffic is recorded by an Envoy sidecar on each ingress node, deduplicated by client IP-subnet, and split by User-Agent into "human" (browser-class) and "API / mirror" before it lands in the warehouse.
Seed and leech counts are sampled by a dedicated crawler that loops the 6.1M active torrents every 90 seconds and records the median across three trackers. Indexer throughput comes from the RabbitMQ work queue, where each "indexed entry" event carries a timestamp and engine attribution. Malware blocks are logged by the scan-runner service with engine-tagged verdicts — false-positive rate is audited weekly against a held-out test corpus of 12,000 known-good torrents.
The page is manual-refresh by design. There is no client-side polling, no WebSocket, and no caching layer in front of your request. What you see is what the database reports right now — which is also what our team sees. If a number looks wrong, it usually is, and we'd rather you tell us about it than paper over it with a smoother chart.
- 15sPrometheus scrape interval
- 4probe datacenters
- 90dretention window
- 0client-side JS on this page
// 04 / 2024 SNAPSHOT
Four numbers a journalist can screenshot.
2,103,447,821
magnet fetches served
41,302,914
unique visitors
1,401,925
malware blocks
12,438,602
torrents indexed (Q1 2026)
snapshot timestamp: 2026-01-14 09:42 UTC · measured via the four-datacenter probe set · report a discrepancy
// 05 / STATS FAQ
Questions the index gets asked.
How are these numbers measured?
Every figure on this page comes from a Prometheus scrape at 15s intervals across four probe datacenters — Falkenstein, Strasbourg, Helsinki, and Beauharnois. The same scrape feeds the internal Grafana board the on-call engineer reads at 09:00 UTC, so there is no separate "public" pipeline.
Why is there no auto-refresh?
Because a continuously-updating dashboard invites two failure modes we don't want: people citing half-loaded numbers, and a CDN edge accidentally masking a real outage for a few minutes until the cache TTL expires. Hitting refresh is two keystrokes (Ctrl-R) and gives you the truth.
What about my privacy when I hit this page?
We log your IP, User-Agent, and the page URL for 14 days for abuse and capacity planning — the same retention window as the rest of the index. We do not log the magnet links you fetch. For anonymity beyond that, pair the site with a separate VPN or Tor; this site is not, and does not claim to be, an anonymity tool.
Can I scrape these numbers for a story or research paper?
Yes, with attribution and a timestamp. The raw 90-day Prometheus snapshots are published daily at /stats/snapshots/ as CSV (no auth, rate-limited to 60 req/min per IP). If you need the full historical archive for academic work, drop a line to [email protected] and we can arrange a one-off transfer.
Where do I report a discrepancy?
Email [email protected] with the metric name, the time window, and the figure you saw vs. the figure you expected. We audit discrepancies monthly and publish corrections in the changelog at /about/.