🚨 Mandatory Testnet (atlantic-2) Upgrade - v6.8

⚠️ This is a MANDATORY upgrade for all atlantic-2 testnet node operators and validators. Nodes that do not upgrade will halt at the upgrade height and fall out of consensus.

📋 Upgrade Summary

Note: The on-chain Upgrade Name (v6.8) is intentionally different from the git tag (v6.8.0-rc2). Cosmovisor matches on the Upgrade Name, so the directory must be named v6.8 — see below.

⚠️ Check Before Upgrading

v6.8 removes several modules and servers and changes some RPC defaults. If any of these affect your node, act before the upgrade height:

• The Cosmos gRPC (:9090) and gRPC-Web (:9091) servers are removed. Use Tendermint RPC or the REST/LCD API, which serve the same queries and broadcast. The [grpc] and [grpc-web] sections of app.toml are no longer read. --grpc.* and --grpc-web.* flags are accepted as no-ops, but --grpc-only now fails as an unknown flag, so remove it from your service unit.
• RPC rate limiting is on by default for nodes whose config files predate v6.7 and for fresh seid init nodes: EVM JSON-RPC allows 200 rps per IP (burst 200) and batch_request_limit drops to 100, and CometBFT RPC allows 200 rps per IP (burst 400). Config files generated by v6.7 already set rate_limiting_enabled = false and keep it. If your node sits behind a reverse proxy or load balancer, set trusted_proxy_cidrs under [evm] in app.toml and trusted-proxy-cidrs in config.toml. Otherwise all traffic is charged to the proxy's IP.
• ABCI /store state proofs are disabled by default. Proof queries return ErrNotSupported (code 37). If you still serve them, set sc-enable-state-proof = true under [state-commit] in app.toml.
• The oracle, tokenfactory and vesting modules are removed. New oracle and tokenfactory transactions are rejected at CheckTx, and CosmWasm contracts can no longer use the tokenfactory messages or query route. Remaining vesting accounts become plain accounts, so their locked coins become spendable. Historical transactions from these modules still decode.
• Legacy precompile versions are removed. A v6.8 node runs debug_trace* and historical eth_call only with the current precompiles. Send requests for blocks before the upgrade to a node running the earlier release in freeze mode.
• seid export is removed.
• Do not swap the binary before the halt. The v6.8 upgrade deletes the retired module stores using the upgrade-info.json that v6.7 writes when it halts. If a v6.8 node halts with an error that retired stores survive, write {"name":"v6.8","height":<upgrade height>} to data/upgrade-info.json and restart.

Full details are in the v6.8 section of the CHANGELOG.

✅ What Operators Need to Do

1. Build the new binary from the v6.8.0-rc2 tag (git checkout v6.8.0-rc2 && make install).
2. Wait for the chain to halt automatically at height TBD. Do not stop your node or switch to v6.8.0-rc2 before this height.
3. Swap in the new binary and restart your node once the halt occurs.
4. Confirm your node resumes producing/syncing blocks past the upgrade height.

💡 Cosmovisor users: place the new binary under upgrades/v6.8/bin/ (using the upgrade name, not the git tag) ahead of time, and the switch will happen automatically at the upgrade height.

What's Changed

• Backport release/v6.8: Return [] from eth_accounts on keyless node by @masih in #4731
• Backport release/v6.8: perf(evm): key transientAccounts by address by @seidroid[bot] in #4724
• Backport release/v6.8: Update v6.8 changelog in prep to cut v6.8.0-rc2 by @seidroid[bot] in #4779
• Bump version to v6.8.0-rc2 in prep for release by @amir-deris in #4778

Full Changelog: v6.8.0-rc1...v6.8.0-rc2

Experimental Pre-built Binaries

Pre-built statically-linked seid binaries attached below for linux/amd64
and linux/arm64.

Verify your download:

sha256sum -c checksums.txt --ignore-missing

For Docker images, see ghcr.io/sei-protocol/sei.