Andrew Schran
Andrew Schran
Is your local repo up to date? This should have been fixed by https://github.com/MystenLabs/sui/commit/a9cbe6ec174339dc6247ec24baef6e4d322c6553 When I run your repro steps with a fresh clone, it works
If we think it matters for performance, I could actually change this to have primary indicate in the request whether it needs to be sync (batches from a Header) or...
> I think we should probably stop using this db abstraction entirely and just use the sync db reads and writes directly. This abstraction currently serializes all reads and writes...
> No we can do this now, but as a note we just need to change all of narwhal's db usage. File an issue on me? :)
FYI: Left a few questions for PR reviewers in comments tagged with "DO NOT MERGE".
> Just curious, will there still be randomness-only checkpoints, or it will be another discussion / change? Yes there are still randomness checkpoints. I don't follow what other discussion you're...
> > > Just curious, will there still be randomness-only checkpoints, or it will be another discussion / change? > > > > > > Yes there are still randomness...
> Looks good but how the new numbers estimated? I did some back of the envelope math and assumed we'd want at least ~2Mbps per stream on a 1Gbps NIC...
> 1. Sui contacting its own worker makes a lot of sense to me (prior art: [MystenLabs/narwhal#715](https://github.com/MystenLabs/narwhal/pull/715)), but could you say more about the use case in which Narwhal itself...
Are there any debug logs on worker side we could also look at?