IBC Relayer Infrastructure: Nodes, Monitoring, and Reliable Packet Delivery
Plan IBC relayer infrastructure around reliable nodes, funded accounts, packet delivery, and client monitoring.
An IBC transfer depends on more than the two blockchains involved. Someone also has to observe both chains and submit the transactions that move packets, acknowledgements, and client updates between them.
That job belongs to an IBC relayer.
A relayer can be running while transfers remain stuck: its wallet may be empty, its RPC endpoint may be behind, or the light client on one side may need attention. Reliable relaying therefore requires reliable nodes, funded accounts, and monitoring of the entire path.
What an IBC relayer does
IBC-enabled chains use on-chain clients to verify information from their counterparties. An off-chain relayer observes the chains and submits the messages needed to move an IBC interaction forward.
For a transfer, this commonly includes:
- Observing a packet on the source chain.
- Submitting the packet and proof to the destination chain.
- Relaying the acknowledgement back to the source chain.
- Handling a timeout when a packet cannot be delivered within its allowed conditions.
The relayer pays transaction fees using its own accounts. Under the security assumptions of the IBC clients and chains, a relayer cannot simply invent a valid packet. However, a relayer that is offline or misconfigured can delay delivery.
Multiple independent relayers can serve a path, improving resilience, but their responsibilities and operating costs still need to be clear.
What infrastructure does a relayer need?
The exact requirements depend on the relayer software and chain configuration, but a production setup commonly includes:
- RPC access to both chains.
- gRPC access where required by the relayer.
- Access to recent blocks, events, and proof-related data.
- A funded relayer account on each chain.
- Monitoring for packets, clients, transactions, and node health.
- A process for upgrades and incident response.
An endpoint that responds to a basic health check is not necessarily ready for relaying. It may be behind the chain tip, missing a required interface, or unable to serve data at the heights the relayer needs.
A dedicated node can make these dependencies easier to control, especially when a path carries production traffic and the team needs predictable access to specific interfaces.
Common causes of stuck IBC packets
RPC or gRPC problems
If a node falls behind, drops connections, or cannot serve required data, the relayer may miss events or fail to construct transactions.
Monitor the endpoints used by the relayer, including:
- Block height and lag.
- RPC and gRPC error rates.
- Transaction submission failures.
- WebSocket connection health, if subscriptions are used.
- Access to the historical heights needed for proofs and recovery.
Use a second independent endpoint or node where practical, and test switching before an incident.
Insufficient relayer funds
The relayer needs enough balance on each chain to pay transaction fees. A low balance on either side can interrupt part of the packet lifecycle.
Alert before the balance reaches zero. A useful threshold depends on recent spending, expected packet volume, and the time it takes the team to replenish the account.
Fee settings should also be reviewed after network upgrades or changes to chain requirements.
Expired or inactive clients
A relayer may need to submit client updates as well as packets. If a client is not updated within its configured trusting period, normal updates may no longer be possible.
Recovery depends on the client, chain configuration, and deployed IBC version. It can require coordinated action, so client age deserves its own alert rather than being buried in process logs.
Running Hermes
Hermes is one relayer implementation used in the Cosmos ecosystem. A production configuration defines the chains, endpoints, accounts, gas settings, and operating modes for the paths being served.
A few settings deserve particular attention:
- Client refresh: ensure someone is responsible for keeping clients updated.
- Packet clearing: configure and test how pending packets are discovered and retried.
- Event collection: choose a supported polling or subscription method for the installed version.
- Channel filtering: limit the relayer to the paths it is responsible for.
- Gas configuration: use values appropriate for each chain.
- Key management: protect relayer accounts and limit the funds exposed.
Before deployment, compare the configuration with the Hermes configuration reference for the installed version. Test the full packet lifecycle on a suitable test network before relying on the setup in production.
What to monitor
A useful relayer dashboard answers a business question: are packets moving within an acceptable time?
Track at least:
- Pending packets and acknowledgements by path and channel.
- Packet age, not just packet count.
- Successful and failed relay transactions.
- Wallet balances and estimated time to depletion.
- IBC client update age and trusting-period margin.
- RPC and gRPC availability and block-height lag.
- Process restarts, version changes, and error rates.
A pending packet count can briefly rise during normal traffic. An increasing packet age, combined with failed transactions or endpoint errors, is a stronger signal that intervention is needed.
Who should operate the relayer?
A chain team can operate its own relayer, rely on community operators, or agree on dedicated coverage with an infrastructure provider.
The decision should account for:
- Which channels are covered.
- Who funds and replenishes the accounts.
- Who watches client expiry.
- Which node interfaces and data retention are required.
- Who responds when packets stop moving.
- How coverage is tested after upgrades.
If more than one operator serves the same path, coordination can reduce unnecessary transaction costs and clarify incident ownership.
Dedicated nodes for IBC relaying
DTEAM can discuss dedicated node infrastructure for IBC relayers and the chains they connect. If you already operate a relayer, dedicated RPC and gRPC access can help you control the node layer behind it.
Contact DTEAM with the two chains, relevant channels, expected traffic, and required interfaces. We can discuss the nodes, monitoring, and retention needed for your setup.