Implement IBC index and PostgreSQL backend

[romac]

Background

This issue keeps track of the work that contributes to the objective to "Reduce relayer operators' costs". At a high level, the aim is to implement an IBC node indexer that will reduce the relaying pressure on validator nodes, thus reducing their relaying / infrastructure costs.

🚧 WIP in the anca/ibcnode branch.

Matrix

Status	Tendermint indexer	IBC Node	Runtime	Hermes supervisor/ chain handle	Comments
✅	kv		Cosmos - no cache	temp mem cache	Cache cannot be disabled
🚧	psql		CosmosPsql - IBC db, psql only	temp mem cache	Partially supported in anca/ibcnode
🚧	psql	(g)RPC proxy, no extra storage	Cosmos - no cache	In-memory cache
🔮	kv or psql		CosmosPsql or Cosmos - IBC cache in memory (configurable)	No or in-memory cache	Not supported but should look at the overall feature matrix and figure out what makes sense
🔮	kv or psql	(g)RPC proxy, IBC db psql	Cosmos - no cache

✅ Released - 🚧 WIP - 🔮 Future work

TODO

• [ ] on start:
  • [ ] on first event at H scans everything for H-1 and populates the snapshot at H-1:
  • [x] app state (mainly the timestamp)
  • [x] clients
    • [x] state
    • [x] consensus states
  • [ ] headers
  • [x] connections
  • [x] channels
  • [x] pending send packets
  • [ ] pending acknowledgements
  • [x] saves snapshot
• [ ] during normal operation
  • [ ] maintains based on events
  • [x] app state
  • [x] clients
    • [x] state
    • [x] consensus states
  • [ ] headers
  • [x] connections
  • [x] channels
  • [x] pending send packets
    • [x] add on SendPacket event
    • [x] remove on Acknowledgement event
  • [ ] pending acknowledgements
• [ ] queries moved to psql
  • [x] app state
  • [x] clients
  • [x] state
  • [x] consensus
  • [ ] headers
  • [x] connections
  • [x] client connections
  • [x] connection
  • [ ] connection channels
  • [x] channels
  • [x] channel
  • [ ] packet commitments
  • [ ] packet acknowledgments
  • [x] packet data
  • [x] via query tx (tendermint)
  • [x] via query IBC snapshots

Future work

• [ ] client updates and header related: It would be good if we were able to store readily available data that can reduce the number of RPCs light client performs when building headers for client updates
  • [ ] NewBlock raw event from tendermint-rs includes the block “Meta” data from chain. We drop it in the monitor before we create the NewBlock IbcEvent to be sent to the supervisor. We should include this in the event so it is accessible to the chain to store the block header.
  • [ ] Could query and store validator set if there new IBC events were seen at previous height (H-1) since it is very likely a client update at H will be required
  • [ ] Ensure the current light client APIs can be used, adapt if not. Romain suggested we can populate the memory store with the stored data we have available

---

For Admin Use

• [x] Not duplicate issue
• [x] Appropriate labels applied
• [ ] Appropriate milestone (priority) applied
• [x] Appropriate contributors tagged
• [x] Contributor assigned/self-assigned

[seanchen1991]

@romac or @ancazamfir, if you could figure out which milestone to tag this work under, that would be much appreciated 🙂