Nethermind 2.1.0 is required for Sepolia's Glamsterdam fork on 6 October
Serguey Asael Shinder2 min readNethermind 2.1.0, released on 1 October, is the version every Sepolia node running the client must use before the Glamsterdam upgrade activates on that testnet on Tuesday 6 October at 13:53:36 UTC. The release collects 359 merged pull requests from 29 contributors since 2.0.0. For most nodes it is an in-place upgrade without a resync.
Glamsterdam on Sepolia. The release implements the final rules of EIP-8037 for state gas, a cap on the total gas a single transaction may use (TX_MAX_TOTAL_GAS_LIMIT), and Engine API changes for the Amsterdam execution upgrade.
A faster EVM. The interpreter loop is replaced by tail-call dispatch, stack words stay in a 256-bit limb layout, storage values stay as 256-bit integers until they are persisted, and JUMPDEST analysis is rewritten. The team's measurements: 3.8 to 4.5% faster block processing on its benchmark sets; SLOAD 9 to 18% faster, SSTORE 42% and TSTORE about 50%; and worst-case analysis of 64 KiB of initcode down from 0.4–3.2 seconds to 22–153 milliseconds.
Faster replies to the consensus client. engine_newPayload now answers as soon as the block has executed and its roots match, before the commit; engine_forkchoiceUpdated still waits for that commit.
For archive and tracing operators. An opt-in per-transaction index for flat archives lets a single-transaction trace start at its target instead of replaying the block before it. In the pull request's measurements, debug_traceTransaction went from about 7.1 seconds to about 28 milliseconds, and whole-block traces run 6 to 10 times faster. It is enabled with FlatDb.HistoryTransactionIndexEnabled=true.
Other operational changes. After snap sync on the flat backend, state is healed by replaying block access lists instead of downloading trie nodes, on by default. Networking listens on IPv4 and IPv6. A flat database that RocksDB had to auto-repair is now re-synced rather than trusted. An optional setting deletes the old trie store after a trie-to-flat conversion, about 227 GB on mainnet. The Ethereum Foundation's bootnode entries for mainnet, Sepolia and Hoodi are replaced with a new fleet of five dual-stack nodes. Discovery traffic on a warm-started Hoodi node fell by 90%.
Read first if you are affected. The release notes ask operators to read the breaking changes before upgrading if they run the flat backend with a custom sync configuration, run poacore, run a Flashbots relay, parse Parity-style traces or build plugins against Nethermind.
What it means
Sepolia is the testnet where applications rehearse a fork, so a Nethermind node there that misses this release will fall off the chain on Tuesday afternoon. For everyone else, the release is a preview of what Glamsterdam asks of clients, including the new per-transaction gas cap that contract deployers and bundlers will want to test against on Sepolia first.