[reconfig] Tolerate redundant committee insert #6926
Conversation
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎ |
lxfind
requested review from
mystenmark,
sadhansood,
longbowlu,
andll and
bmwill
| ); | ||
| if let Some(old_committee) = self.get_committee(&new_committee.epoch)? { | ||
| // If somehow we already have this committee in the store, they must be the same. | ||
| assert_eq!(&old_committee, new_committee); |
There was a problem hiding this comment.
Should we return from the function after this assert?
There was a problem hiding this comment.
It's an assert, which would panic?
There was a problem hiding this comment.
I mean do we still need the insert below if the existing committee in the db is same as what we are going to insert again.
There was a problem hiding this comment.
When this happens, we probably have a fork. So it probably doesn't matte which committee we keep for this epoch.
There was a problem hiding this comment.
I think @sadhansood is referring to a fact that if we did have a committee and it does match current value, we could have a separate branch for this case and just skip call to insert(which will save some serialization/io).
We could be inserting the same committee several times from different sources, e.g. from state sync and from reconfiguration flow. This should be ok.