Solana slot vs block height comes down to skipped slots. A slot is a numbered time window that may or may not produce a block; block height goes up only when a block is really added. Every slot without a block widens the gap. Store the slot as the key, because shreds, RPC and explorers all name blocks by slot.
Two counters on one chain#
The slot number goes up by one for every window a leader is given, whether or not a block comes out of it. Block height goes up by one only when a block is actually added to the chain. Both start at the beginning of the chain, so the difference between them is the number of slots in history that never became a block.
The two also answer different questions. A slot tells you when a block was due and which leader was scheduled. A block height tells you how many blocks sit beneath this one, which is what the network uses for transaction expiry: lastValidBlockHeight is a block height, as how long a transaction stays valid explains.
What the numbers looked like on mainnet#
| Reading (18 November 2026) | Value |
|---|---|
| Finalized slot | 470,935,639 |
| Block height at that slot | 448,950,235 |
| Slots in history with no block | 21,985,404 |
| Slots from 470,835,639 to 470,935,639 | 100,001 |
| Blocks in that range | 99,935 |
| Slots in that range with no block | 66 |
So in recent traffic the gap grows slowly, about 66 slots in 100,001, but over the chain's whole history it has reached almost 22 million. Any join that mixes the two numbers will be off by millions.
A skipped slot, step by step#
Slot 470,839,179 produced a block at height 448,853,841. Slots 470,839,180 to 470,839,183 produced none. The next block came in slot 470,839,184, with height 448,853,842 and parent slot 470,839,179. The slot jumped by five, the height by one, and the parent field points back over the gap.
The same pattern marks the change to the new consensus. The block in slot 459,652,999 has height 437,687,746; slots 459,653,000 to 459,653,007 have no block; the block in slot 459,653,008 has height 437,687,747 and names 459,652,999 as its parent.
Asking getBlock for a slot without a block returns an error, not an empty block. A backfill loop should list the slots that have blocks with getBlocks first rather than treat every error as a fault.
What a shred tells you#
The slot number sits in the first header of every shred, data and coding alike. A data shred also carries a two-byte parent offset: the slot minus the parent's slot, so 1 when the previous slot produced a block and more after a gap. Once the block is reassembled, its header marker states the parent slot again, together with the parent's block id.
Neither the shred headers nor the block header give a block height. A receiver that wants one has to count blocks along the chain of parents, or ask RPC later with getBlock. That is one more reason the slot is the natural key for rows decoded from shreds: it is there in the first bytes of every packet. Decoding a block from shreds shows where each field is read.
Which one to store#
- Use the slot as the block key in every table. It is what shreds,
getBlock,getSignaturesForAddressand the status returned bygetSignatureStatusesall report. - Pair it with the transaction's position in the block when you need order, and with the first signature when you need identity.
- Store block height only as an extra column, filled from
getBlockonce the block is confirmed, if you compare againstlastValidBlockHeightor count blocks. - Never compute one from the other. The gap depends on every skipped slot before that point.
Questions about slots and block height#
Is a skipped slot an error?#
No. A leader can miss its window, or its block can lose out to another fork. The chain simply continues from an earlier block, and the next block names that earlier slot as its parent.
Can a block height belong to two different blocks?#
Before a block is final, yes: two competing forks can each have a block at the same height. After finalization only one chain remains, so slot and height both identify a single block.
Which number do explorers show?#
Explorers link blocks by slot, and many also show the block height on the block page. When they say "block 470,935,639" they mean the slot.
Where the slot key is not enough#
A slot alone does not prove a block is on the final chain. A receiver can see shreds for a slot whose block is later skipped, so rows keyed on that slot stay provisional until reconciliation confirms them. And the slot says nothing about expiry. That still needs block height, which you only learn from RPC.
Where to use the slot key next#
The reconcile guide settles slot-keyed rows against confirmed blocks, and checking a decoder against RPC compares them block by block. To read slots straight off live packets, start with the quickstart.
Sources#
- Mainnet getEpochInfo at finalized commitment, slot 470,935,639, read 2026-11-18 (checked 2026-11-18)
- Mainnet getBlocks over slots 470,835,639 to 470,935,639, read 2026-11-18 (checked 2026-11-18)
- Mainnet getBlock blockHeight and parentSlot for slots 470,839,179, 470,839,184, 459,652,999 and 459,653,008 (checked 2026-11-18)
- Agave ledger shred.rs: data shred header with parent_offset (u16) (checked 2026-11-18)
- Agave entry block_component.rs: block header marker with parent slot and block id (checked 2026-11-18)
Raw shreds on your own server
Ultra Low Latency Shreds is $200 and Standard Shreds is $100. You pay in crypto, tell us which IPv4 address and UDP port should receive the stream, and we wire it up by hand.