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
When we restart whole cluster with eventual failover enabled, it doesn't have time to sync membership status. As a result, every instance thinks he's the only one alive, and becomes a leader an all 100 instances do call init({is_master = true}).
Instead, every instance could make first appointments as in disabled mode, and make decisions immune for some time.
The text was updated successfully, but these errors were encountered:
rosik
changed the title
Make first eventual failover decision disregard membership status
[4pt] Make first eventual failover decision disregard membership status
Aug 26, 2020
Proposal form @opomuc.
When we restart whole cluster with eventual failover enabled, it doesn't have time to sync membership status. As a result, every instance thinks he's the only one alive, and becomes a leader an all 100 instances do call
init({is_master = true})
.Instead, every instance could make first appointments as in disabled mode, and make decisions immune for some time.
The text was updated successfully, but these errors were encountered: