Slot time
—
the protocol's default is 400 ms
Timers, bots and docs that assume 400 ms run a third behind the chain.
Your browser just made a random identity and speaks the validators' gossip protocol itself: pull requests, pings, signatures, votes. The slot and the chain's clock below come straight from what validators tell each other, and they don't match the numbers most people quote.
waiting for votes
Voted tip—highest slot 3+ validators voted
Being produced—from turbine shreds reaching this server
one validator's vote, at the slot it voted for and the moment it was signed the fitted line through them: its slope is the slot time
Timers, bots and docs that assume 400 ms run a third behind the chain.
A swap retried after that with the same blockhash can never land. Fetch a fresh one.
Staking rewards and stake activations land at the boundary.
Validators stamp votes in whole seconds, and the on-chain clock is built from those stamps, so it trails real time.
The schedule comes from public RPC through this site's relay (browsers are refused it). The timing is your tab's own slot clock: the slot validators are voting on, which trails the leader producing now by a slot or two. Places are DB-IP's estimate for each leader's advertised gossip IP (IP geolocation by DB-IP, CC BY 4.0). Names are self-reported on-chain.
—
—
—
first shred to first shred, same leader
Each leader's own four slots against the median slot, in ms per slot. They share one network path to this server, so the spacing is the leader's. Slow slots can come from hardware or from deliberate delay; this doesn't say which.
| Client | Slots | Missed | Block | Pace |
|---|
From the turbine shreds reaching one unstaked node in Chicago, which gets about 63% of all shreds. Shred headers aren't signature-checked here, so a block's parent counts only when two different senders agree on it. Block size needs the shred flagged last-in-slot, which arrives for about 60% of blocks. Checked against RPC's finalized blocks over 5,341 slots: the missed list matched the chain's 12 empty slots exactly, with no false alarms. A rolling hour kept in memory: it starts empty whenever the relay restarts.
heard in the last 45 min
of all staked validators
since each validator last started
An unstaked gossip node on this server pulls a fixed 1/64 of the table (four 1/256 slices, each with its own filter). Every node re-signs its contact record every few seconds and each new copy lands in a random slice, so every node passes through ours: most within 15 minutes, nearly all within 30. It reads contact records, not every vote. Clients and start times come from each node's own gossip record; stake and versions from public RPC. Client names follow agave 4.3 (ids 0 to 13, Harmonic included); clients without a colour of their own are pooled as “other clients”.
Each one pinged your node with a random token and got back your signed pong, which gossip requires before it serves you anything.
| Validator | Stake (SOL) | Gossip socket |
|---|
Your node advertises … and answers pings there, so any validator can verify it. It still won't be listed in getClusterNodes: agave 4.2 adds a node to its table only when the node pushes its own record, and this tab only pulls. This server's own node pushes, and is listed.
| Validator | Slot | Behind | Signed | Kind |
|---|