lighthouse

History

realbigsean 42da392edc fix deneb sync bug (#4869 ) ## Issue Addressed I observed our forward sync on devnet 9 would stall when we would hit this log: ``` 250425:Oct 19 00:54:17.133 WARN Blocks and blobs request for range received invalid data, error: KzgCommitmentMismatch, batch_id: 4338, peer_id: 16Uiu2HAmHbmkEQFDrJfNuy1aYyAfHkNUwSD9FN7EVAqGJ8YTF9Mh, service: sync, module: network::sync::manager:1036 ``` ## Proposed Changes `range_sync_block_and_blob_response` [here](`1cb02a13a5/beacon_node/network/src/sync/manager.rs (L1013)`) removes the request from the sync manager. later, however if there's an error, `inject_error` [here](`1cb02a13a5/beacon_node/network/src/sync/manager.rs (L1055)`) expects the request to exist so we can handle retry logic. So this PR just re-inserts the request (withthout any accumulated blobs or blocks) when we hit an error here. The issue is unique to block+blob sync because the error here is only possible from mismatches between blocks + blobs after we've downloaded both, there's no equivalent error in block sync Co-authored-by: realbigsean <seananderson33@gmail.com>	2023-10-31 21:04:18 +00:00
..
src	fix deneb sync bug (#4869 )	2023-10-31 21:04:18 +00:00
Cargo.toml	Enable BLS portable feature on all CI tests (#4868 )	2023-10-20 07:30:27 +00:00

realbigsean 42da392edc fix deneb sync bug (#4869 )

## Issue Addressed

I observed our forward sync on devnet 9 would stall when we would hit this log:
```
250425:Oct 19 00:54:17.133 WARN Blocks and blobs request for range received invalid data, error: KzgCommitmentMismatch, batch_id: 4338, peer_id: 16Uiu2HAmHbmkEQFDrJfNuy1aYyAfHkNUwSD9FN7EVAqGJ8YTF9Mh, service: sync, module: network::sync::manager:1036
```

## Proposed Changes

`range_sync_block_and_blob_response` [here](1cb02a13a5/beacon_node/network/src/sync/manager.rs (L1013)) removes the request from the sync manager. later, however if there's an error, `inject_error` [here](1cb02a13a5/beacon_node/network/src/sync/manager.rs (L1055)) expects the request to exist so we can handle retry logic. So this PR just re-inserts the request (withthout any accumulated blobs or blocks) when we hit an error here.

The issue is unique to block+blob sync because the error here is only possible from mismatches between blocks + blobs after we've downloaded both, there's no equivalent error in block sync



Co-authored-by: realbigsean <seananderson33@gmail.com>

2023-10-31 21:04:18 +00:00

src fix deneb sync bug (#4869 ) 2023-10-31 21:04:18 +00:00

Cargo.toml Enable BLS portable feature on all CI tests (#4868 ) 2023-10-20 07:30:27 +00:00