Coin Brief ENDE

Geth 1.17.7 lets node operators pick the block where chain history starts

The go-ethereum team released Geth v1.17.7 on 30 September. According to the notes, it exists mostly because the Ubuntu packages for v1.17.6 could not be published, and it picks up a handful of fixes that landed afterwards.

For validators the key line is the fork date. The Amsterdam upgrade reaches the Sepolia testnet on 6 October 2026, and the team says either v1.17.6 or v1.17.7 will handle it, so Sepolia operators can run whichever they have.

The main new option is a user-chosen history pruning point. With --history.chain <block number>:<block hash>, an operator can pick where chain history starts. The flag applies to an existing node trimmed with geth prune-history as well as to snap sync, which then skips downloading history before that block. A predefined postosaka pruning point is also added.

Several fixes relate to Amsterdam features. The miner now accounts for both execution gas and state gas, as defined in EIP-8037, when selecting transactions. Block-level access lists are encoded and hashed once in a dedicated pipeline, are only fetched close to the chain head, and are applied after the state sync so the pivot of snap sync v2 keeps moving.

Other changes fix a memory leak when stopping the chain, release database caches when the database is disabled, skip peers whose announced block range does not cover a request, allow an interrupted snap sync to restart when a history cutoff is set, and restore a transaction-pool fallback during snap sync that regressed in v1.17.3. Five new bootnodes run by NodeOps are added, and test fixtures move to execution-spec-tests v21.

Geth 1.17.7 lets node operators pick the block where chain history starts
Geth 1.17.7 lets node operators pick the block where chain history starts — Coin Brief

Why it matters

Testnet forks are where client bugs surface before mainnet. Running a release that the client team has marked ready for the Sepolia date, and upgrading before 6 October, is the step that keeps a validator from missing the fork. The pruning option will matter to anyone trimming disk use on long-running nodes.