Brennan
Brennan
Some A vs. B performance comparison data:   Baseline was taking ~7 seconds for dirty store processing and 10 seconds for overall clean. This code was taking ~800ms for...
> Whats the 7-4 deep slots? The 7-4 deep votes on the tower. We have the constraint that our 8th deep slot/vote needs to be OC in order to vote,...
First cut at adding another vote stake threshold constraint: https://github.com/solana-labs/solana/pull/34120
#34120 has been merged into master and should help mitigate the negative impacts of these events. So far, it has not been backported and would be part of a v1.18...
Can we also add a warn or err log message inside `entries_to_shreds`. Something like this: ``` if let Some(slot_meta) = blockstore.meta(slot).unwrap() { if slot_meta.received > 0 || slot_meta.consumed > 0...
> Well if it is an "invariant", don't we want a loud panic here if someone introduces a bug and breaks that invariant? Generally speaking, I agree. In this specific...
Everything still appears to be working after the latest changes to use cluster slots and take a weighted sampling: ``` [2024-01-18T22:07:05.109321073Z WARN solana_core::repair::repair_service] BW: successfully sent repair request to 7dKiopJkRwh6yrsBUnhEX5zTNCDWQwBRsvbD4fTdrUVc...
Put together a proof of concept for downloading from multiple RPC nodes in parallel: https://github.com/solana-labs/solana/compare/master...brennanwatt:solana:snapshot_torrent
I believe the proof of concept is mature enough to have a productive discussion on the right path forward. See draft PR #29507 EDIT: See comment in #29507 describing flaw...
Activated on testnet at ~2024/01/23 14:10 UTC The cluster was spending ~19% of the time hashing before and now is spending ~27% of the time, so increasing linearly with hash...