-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Obsolete warning about even number of validators? (Authority Round) #11107
Comments
Yeah maybe we should reword that (cc @tomusdrw); the reason it's there is we didn't want to waste time on debugging odd bugs when running with unsupported setups (like e.g. two nodes). 2/3s quorum is not mandatory though right? Recommended but not mandatory? |
Yes, it's a transition and if that's not configured then the default is 50%. |
We don't recommend running with even number of validators, cause in the past we have seen splits, where With 2/3 majority and small number of validators the issue might be even worse, cause we may see very long forks and never reach finality (with 5 validators you need 4 to reach finality). I'm fine with removing the warning in case 2/3 majority is enabled (as it doesn't make much sense), but we might consider adding a warning when the number of validators is below 6 that you need to make sure that validators are (very) well connected and online. |
Thank you for the explanation! It rather sounds like the proper solution to this issue would be some sort of a tie-breaker, to make sure that in a situation where the nodes are split 50-50, they all switch to the same alternative? Anyway, feel free to close this issue! |
@afck the whole point of the warning is to give users a heads-up that what they are doing is potentially dangerous, so if you have a suggestion on better wording here I'd be happy to change it. I mean, if its meaning is unclear to you who are very knowledgeable about this stuff, perhaps it's not very good as is today? How can we improve it? |
I actually think the formulation is okay.
Not really. 😅 I wasn't involved in the original development of Authority Round, and still trying to wrap my head around it. |
SimpleList::new
prints a warning about a "risk of network split
" when the number of validators is even.Does #10909 make that warning obsolete after transitioning to a 2/3 quorum?
I'm not sure I understand the warning in the first place, though: Since even with a 1/2 quorum finality requires strictly more than 50%, why does an even number of validators increase the risk of a network split?
The text was updated successfully, but these errors were encountered: