Measured Uptime
How often rpc.mnr.network answered, measured from outside the relay. Every minute a Cloudflare Worker asks the relay for the chain height with a token of our own, and reads its health endpoint. This is a beta with no uptime promise: these are measurements, not a guarantee.
Loading…
Last 90 days
Recent incidents
| Start (UTC) | State | Measured minutes | Reason |
|---|---|---|---|
| Loading… | |||
This pool, measured the same way
Every 5 minutes the same probe goes to the relay and, directly, to each public node in its pool: one get_info, 4-second timeout, one retry. A target is up if it answers, reports itself synchronized, and is no more than one block behind the median of the round. The relay row here uses this probe, so it is comparable with the nodes; the figures at the top use the per-minute probe.
| Target | 24 h | 7 UTC days | 30 UTC days | Coverage, 30 d | Median latency |
|---|---|---|---|---|---|
| Loading… | |||||
These nodes are run for free by volunteers and companies, and mnr depends on every one of them. Each run measures from a single Cloudflare location, so a low figure can reflect that network path, or a node limiting Cloudflare addresses (counted as refused, which lowers coverage, not availability); it is not a verdict on the operator. The relay's latency includes the Cloudflare proxy and its quorum cache, unlike the direct node probes. Our own node and the onion and I2P nodes are not measured here. The relay comes first, then the nodes alphabetically; coverage counts from when this table began measuring, so a node added later shows the days it was not measured.
How this is measured
Up: the relay answered get_height with a valid height. Degraded: it answered, but its health endpoint did not report healthy (usually a degraded node pool). Stale: it answered, but the height had not moved for 60 minutes. Down: no valid answer twice in a row, 5 seconds apart. Availability counts up, degraded and stale minutes as available, out of all measured minutes. The 7- and 30-day figures are whole UTC days including today; the 24-hour figure is a rolling window.
A minute in which our own monitor token was refused is our mistake, not the relay's, and is not counted. A minute with no measurement (the cron did not run) is left out of the percentage; the coverage under each figure says how many of the expected minutes were measured. Measurements start on the date shown above, and nothing earlier is implied. The onion and I2P addresses are not measured here.
The raw numbers are at /status.json. Node-by-node health is on Upstreams.