Overview
Important
This is a recommended release for all op-reth operators. It updates reth to v2.7.0, which brings breaking RPC changes and new RPC and engine defaults. It also adds the sepolia-devnet-4, -5 and -6 chain configs and fixes the interop filter's endpoint.up metric.
Breaking changes
- Three RPC limit flags reject
0(#24024).--rpc.max-tracing-requests,--rpc.max-blocking-io-requestsand--rpc.proof-permitsno longer accept0. Operators setting any of them to0must pick a positive value before upgrading, or op-reth will not start. --rpc.gascapnow also capseth_estimateGas(#24024). The default stays at 50,000,000 gas. Estimates that need more gas than the cap now fail. Operators of chains whose block gas limit is above the cap, or whose users estimate large transactions, should raise--rpc.gascapto match.- New error for state queries on pruned history (#24024). A state query for a block whose history is pruned now fails with code
4444andpruned history unavailable: requested N, earliest available M. It used to fail with-32603and aninsufficient changesetsmessage. RPC clients that match the old code or text must be updated.
Chain Configuration
- sepolia-devnet-4, -5 and -6 added (#24054). These are OP Labs devnets on Sepolia used to rehearse upcoming network upgrades, listed in the superchain registry. op-reth can now start them by name with
--chain, without an explicit genesis. No other chain's config changes in this release.
Other changes
Upstream
- ! Updates reth to v2.7.0 (#24024). No flags are removed. Besides the breaking changes above, operators inherit these upstream changes:
- Lower RPC cache defaults:
--rpc-cache.max-blocksdrops from 5000 to 1000, and--rpc-cache.max-receiptsfrom 2000 to 500. These upstream defaults are not tuned for 2-second blocks: 1000 blocks is about 33 minutes of history. Operators who serve recent-history RPC traffic may want to set the old values explicitly. - New RPC cache flags:
--rpc-cache.max-blocks-bytes(default 2 GiB),--rpc-cache.max-receipts-bytes(default 1 GiB) and--rpc-cache.idle-timeout(default 1h). op-reth rejects--rpc-cache.max-bals-bytes, like every other block access list option. --engine.sender-recovery-cacheis now on by default. Pass--engine.sender-recovery-cache=falseto turn it off.- New
--engine.backfill-thresholdflag (default 32, the previous fixed value).
- Lower RPC cache defaults:
- reth is pinned to the tag
op-v2.7.0-0of the ethereum-optimism/reth fork: upstream reth v2.7.0 plus OP Stack patches. op-reth now builds only against tagged fork releases (#24071)
Metrics
- The interop transaction pool's per-endpoint
optimism_transaction_pool.interop.endpoint.upgauge now tracks the failsafe poll, which reaches every configured--rollup.interop-httpendpoint about once a second. It used to change only on interop transaction checks, so with no interop traffic it kept reporting 1 even when the filter was unreachable. Operators can now alert on it before interop traffic arrives. With traffic, a reachable but out-of-sync filter makes the gauge alternate between 0 and 1;rejected_no_quorumdecisions show that case (#23964)
Full Changelog: op-reth/v2.8.2...op-reth/v2.9.0
🚢 Docker Image: