BlockDAG network status

Each row gets one verdict, and it is the only thing on the row coloured strongly. The columns after it are measurements, not judgements. An earlier version of this table judged state and chain in two separate columns: both were true, and an endpoint frozen for three days still collected two green labels, because it was answering and it was on this chain. Green reads as usable, so that table was accurate and misleading at once.

The rules are applied in this order, and the first one that fits is the verdict:

• Down — the last two probes in a row went unanswered.
• Fork — it reports this chain ID like every other endpoint, but its block at a height both nodes have is not the block this node holds.
• Stalled — its height has not moved in ten minutes while this node kept advancing.
• Behind — more than 375 blocks behind this node, or it reports itself as still syncing.
• Lagging — between 21 and 375 blocks behind.
• Unverified — it answers and is current, but the block lookup failed this round, so its chain could not be checked. Not an accusation; the next round usually settles it.
• OK — answering, on this chain, and within 20 blocks of this node.

Which chain, and how deep. A height on its own settles nothing: a node can be a million blocks behind and still be on this chain, and it can sit at the same height on a different one. So each endpoint is asked for the hash of one block at a height both it and this node have. Near the head of a node that is catching up the order has not settled, so it disagrees there while being on the same chain — comparing 64 blocks back once labelled an endpoint 350,000 behind as a fork, wrongly. The comparison is therefore taken 2,000 blocks back from the lower of the two heads, past the unsettled window on both sides. Being behind is not being forked, and an endpoint that is offline or syncing is neither.

When this node is itself behind, every lag figure here is measured against a head that is not current. A note appears above the table and the verdicts that rest on lag are shown grey rather than red, because the right response is to distrust the comparison, not the endpoint.

The Answered column counts probes that got a reply, and nothing else. An endpoint on a fork can answer every single time and still be useless to a wallet, so a high figure there is not a recommendation — read it together with the verdict, not instead of it. The heading says the span it actually covers, which is shorter than a day whenever the history was recently reset.