op-reth replica (AltDA) using docker-compose.yml with:
- op-reth
public.ecr.aws/i6b2w2n6/op-reth:v2.3.1 - op-node
public.ecr.aws/i6b2w2n6/op-node:1.16.1-celestia-e9ec322-altda(AltDA mode) - op-alt-da
public.ecr.aws/i6b2w2n6/op-alt-da:v0.15.0-11cc587(celestiaorg/op-alt-da)
Celestia namespace: ca1de12a2f7443cbfbb5 (29-byte v0 form in config: 00000000000000000000000000000000000000ca1de12a2f7443cbfbb5).
Two independent data paths, both preconfigured:
Following the chain head. op-node receives new blocks over the OP Stack p2p network from Caldera's replica (--p2p.static, port 9003). op-reth pulls the block bodies and state it is missing over execution-layer p2p from the same replica (--trusted-peers, port 30303). op-node runs --syncmode=execution-layer, so after a restart or when starting from a snapshot, op-reth catches up to the head through this path rather than re-executing from L1.
Verifying against L1. op-node independently derives the chain from Sepolia batch data. Batch contents live on Celestia. op-alt-da fetches them, from Celestia if you configure a bridge, otherwise from Caldera's public S3 cache. This path advances the safe head. It is slow and is not what makes the replica usable; the p2p path is.
Both need outbound TCP from the host.
Download and extract into ./datadir before the first start:
https://constellationlabs-dashboard-beta.s3.amazonaws.com/bedrock-manta-sepolia-reth-2026-May-14.tar
Manta Sepolia changed its data-availability format early in its life. The current op-node cannot derive the early blocks, so a sync from genesis will not complete. The snapshot starts after the change.
By default op-alt-da reads batch data from Caldera's public S3 cache. No credentials are needed and the cache holds the chain's full history. This is the recommended setup.
You can instead read from your own Celestia Mocha node by uncommenting and setting bridge_addr and bridge_auth_token in op-alt-da-config.toml. Only do this with an archival node. Celestia nodes prune blobs older than the sampling window, and op-alt-da treats a Celestia "not found" as final without consulting S3, so a pruned node breaks derivation of anything older than that window.
L1_RPC_URL must point at Ethereum Sepolia (chain id 11155111).
cp .env.example .env # set L1_RPC_URL
make up
# or without make: bash ./up.shup.sh generates jwt-secret.txt and p2p-node-key.txt on first run. Keep them out of version control.
Stop:
make downL2 RPC (default op-reth port 8545):
RPC_URL=http://localhost:8545
curl $RPC_URL -X POST -H "Content-Type: application/json" --data \
'{"jsonrpc":"2.0","method":"eth_chainId","params":[],"id":1}'Rollup sync status (default op-node port 7545):
RPC_URL=http://localhost:7545
curl $RPC_URL -X POST -H "Content-Type: application/json" --data \
'{"jsonrpc":"2.0","method":"optimism_syncStatus","params":[],"id":1}' | jq .unsafe_l2 tracks the chain head via p2p. safe_l2 tracks L1 derivation and lags behind; that is expected.
Catch-up progress: bash progress.sh
The images are amd64 only and run under Rosetta. Rosetta miscomputes one of the ciphers libp2p uses, which breaks every op-node peer connection. The compose file sets GODEBUG=cpu.avx2=off on op-node to work around it. It is harmless on other hosts.
Please refer to celestia docs for network upgrades: https://docs.celestia.org/how-to-guides/participate#network-upgrades
The deployed Manta Sepolia network runs op-alt-da:0.12.0. This guide pins the newer v0.15.0-11cc587, which is the version the guide was written and tested against. If op-alt-da fails to start or cannot read blobs, try public.ecr.aws/i6b2w2n6/op-alt-da:0.12.0 in docker-compose.yml. The config file format may differ between the two versions.