[Bug]: Gateway cannot serve queries

[tmigone]

Indexer address

Operator address

Indexer agent version

Indexer service version

TAP agent version

Description

Gateway is not serving queries, erroring out with

 {"timestamp":"2025-10-29T13:39:47.546389Z","level":"INFO","fields":{"message":"indexer_request","indexer":"0xb0188c4d02eab4d444c1678f1ede9f790ffc838e","deployment":"QmPVJW8VbtU5RDs7kpBdY8JJ3mhkoHzLiEm6zDeq67jDPB","allocation":"0xc2186cd1e8fec170d03c6d638ab2c9b4e591cbd6","url":"https://capitanbeto.xyz/","result":"Err(BadResponse(\"bad attestation: recovered signer is not expected\"))","response_time_ms":170,"seconds_behind":0,"fee":0.00015374742434627302},"target":"graph_gateway::client_query","spans":[{"request_id":"996314bcea22e7fd-ORD","selector":"QmPVJW8VbtU5RDs7kpBdY8JJ3mhkoHzLiEm6zDeq67jDPB","name":"client_request"}]}

Relevant Logs

No response

[tmigone]

Looks like the attestation signer being recovered by the gateway does not match the allocation id for the subgraph deployment. Investigating...

[neithanmo]

I remember that We hit this error during testing in our local network and fixed that at that time, if not mistaken it was caused by using the wrong eip712 domain?

[tmigone]

Probably a network subgraph bug!

Indexer service gets disputeManager address from the network subgraph with a query:

{
  graphNetworks {
		disputeManager
  }
}

This is currently returning 0x7c9b82717f9433932507df6eda93a9678b258698 which is the old dispute manager, it should return 0x96e1b86b2739e8A3d59F40F2532caDF9cE8Da088 .

[tmigone]

Network subgraph initializes disputeManager with the correct horizon value: https://github.com/graphprotocol/graph-network-subgraph/blob/8a08240bae6b4695424873e11c56997b1010de65/src/mappings/helpers/helpers.ts#L670

But this is then overriden by this handler: https://github.com/graphprotocol/graph-network-subgraph/blob/8a08240bae6b4695424873e11c56997b1010de65/src/mappings/controller.ts#L32

when the original DisputeManager is registered in the controller here: https://sepolia.arbiscan.io/tx/0x7cc8fab8c0d3a8bcffc1191798dc5f6e77c45da0fbc6a7e143c7e91517149399

[tmigone]

Proposed fix

Have the governor update the DisputeManager to the new contract address. However this needs to be done during Phase 4, otherwise during Phase 3 indexers will use the incorrect value for signing attestations.

[tmigone]

Ended up uncovering some issues with query serving continuity during the transition to horizon.

Solved with a couple different changes:

• Ensuring during Phase 4 we override the DisputeManger value in the Controller
• Deploying an intermediate gateway during Phase 3 that accepts attestations for both dispute managers (legacy and horizon)
• Ensuring horizon tap-escrow-manager creates accounts for indexers based on legacy allocations. Previously it was ignoring them.

Related PRs:

• [Horizon] Phase 3 gateway edgeandnode/gateway#1144
• feat: register new dispute manager in controller overriding old one contracts#1248
• edgeandnode/tap-escrow-manager@30bd197