protolambda
commented
1 month ago PRs:
- op-supervisor changes by @axelKingsley : https://github.com/ethereum-optimism/optimism/pull/11956
- op-node lazy-rpc connection by @protolambda : https://github.com/ethereum-optimism/optimism/pull/11970
- op-geth PR was updated to lazy-dial the sequencer RPC too: https://github.com/ethereum-optimism/op-geth/pull/372
There is a cyclic dependency in RPCs:
The op-supervisor should decouple the notion of an "L2 endpoint" from an "L2 monitor / DB". If the endpoint is unavailable, offline, etc. then the existing DB should still be discovered, and the existing data should be served.
An L2 endpoint should be possible to add/remove from the op-supervisor without restarting the L2 monitor, or the backend as a whole.
Once it's more decoupled, the op-supervisor can be started, op-geth can connect to it, and then we can make the op-supervisor connect to op-geth when available.
Ideally we also make op-supervisor retry connection attempts of existing configured endpoints, without blocking startup of the op-supervisor, so that we don't have to call the op-supervisor RPC after all the things start up.