7dbc59efeb
## Issue Addressed #3302 ## Proposed Changes Move the `reqwest::Client` from being initialized per-validator, to being initialized per distinct Web3Signer. This is done by placing the `Client` into a `HashMap` keyed by the definition of the Web3Signer as specified by the `ValidatorDefintion`. This will allow multiple Web3Signers to be used with a single VC and also maintains backwards compatibility. ## Additional Info This was done to reduce the memory used by the VC when connecting to a Web3Signer. I set up a local testnet using [a custom script](https://github.com/macladson/lighthouse/tree/web3signer-local-test/scripts/local_testnet_web3signer) and ran a VC with 200 validator keys: VC with Web3Signer: - `unstable`: ~200MB - With fix: ~50MB VC with Local Signer: - `unstable`: ~35MB - With fix: ~35MB > I'm seeing some fragmentation with the VC using the Web3Signer, but not when using a local signer (this is most likely due to making lots of http requests and dealing with lots of JSON objects). I tested the above using `MALLOC_ARENA_MAX=1` to try to reduce the fragmentation. Without it, the values are around +50MB for both `unstable` and the fix. |
||
|---|---|---|
| .. | ||
| account_utils | ||
| clap_utils | ||
| compare_fields | ||
| compare_fields_derive | ||
| deposit_contract | ||
| directory | ||
| eth2 | ||
| eth2_config | ||
| eth2_interop_keypairs | ||
| eth2_network_config | ||
| eth2_wallet_manager | ||
| fallback | ||
| filesystem | ||
| hashset_delay | ||
| lighthouse_metrics | ||
| lighthouse_version | ||
| lockfile | ||
| logging | ||
| lru_cache | ||
| malloc_utils | ||
| monitoring_api | ||
| sensitive_url | ||
| slot_clock | ||
| target_check | ||
| task_executor | ||
| test_random_derive | ||
| unused_port | ||
| validator_dir | ||
| warp_utils | ||
| README.md | ||
eth2
Common crates containing eth2-specific logic.