You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Fixed
Recovering many failed partitions of one observer is no longer quadratic. Every save of an observer's failed partitions rewrote all of them, and the resolved ones again, one round trip at a time, so with hundreds of failed partitions each recovery took over a second of the observer's time. The MongoDB storage now sends only what changed since it last wrote, in a single bulk request. 659 failed partitions that took about an hour to recover now recover in under a minute.