Swallowing "Not Found" errors

[mikeal]

I'm curious why this line is in the code:

https://github.com/mafintosh/hyperlog/blob/master/lib/replicate.js#L131

It caused me some serious headaches, I couldn't figure out why replication was failing. I actually still don't know why replication is failing cause I don't know why I'm getting this error, but it was really tough tracking it down.

[mafintosh]

Its actually not swallowing the error but a not found at this point indicates a node was received out of order and will be buffered for a bit.

Can you share a failing test case where replication doesnt work?

[hackergrrl]

@mikeal I've seen this happen in cases where one of the sides' hyperlog is corrupted: out of order or conflicting entries in one of the sublog's append-only log. This was happening in certain deduplication cases until around Nov 20 2016 and may explain the replication failures.

I recommend wiping your hyperlogs (if possible), upgrading to the latest version of the module, and trying again. You can use hyperlog-doctor on your hyperlogs' leveldbs to see if there is any (detectable) corruption.