Learn the vocabulary of preventing two independently elected leaders from emerging on either side of a network partition.
0 / 5 completed
1 / 5
A teammate explains that a network partition split a cluster into two halves, and each half, unable to see the other, independently elected its own leader and kept accepting writes, so the cluster now has two leaders disagreeing about the true state of the data. What failure mode is being described?
A split-brain condition is exactly this: a network partition divides a cluster so that two subsets of nodes can no longer see each other, and if both subsets independently elect their own leader and keep accepting writes, the cluster ends up with two leaders disagreeing about the true state of the data, risking conflicting, unreconcilable writes on each side. A DNS zone transfer is an unrelated concept about replicating name server records. This two-independent-leaders-during-a-partition failure is exactly why consensus protocols require a majority quorum before electing a leader, specifically to prevent split brain.
2 / 5
During a design review, the team requires a strict majority quorum before any leader election can succeed, specifically so that during a network partition, at most one side of the split can ever gather enough votes to become leader. Which capability does this provide?
Requiring a strict majority quorum here provides prevention of split brain, since a strict majority requirement mathematically guarantees at most one partition side can ever gather enough votes to become leader at the same time, because two disjoint node sets cannot both contain a majority of the same cluster. Letting either side of a partition elect a leader independently with no quorum requirement is exactly the scenario that produces two simultaneous, conflicting leaders. This majority-required-for-election guarantee is exactly why quorum-based consensus protocols prevent split brain by design.
3 / 5
In a code review, a dev notices a cluster's leader-election logic lets any single node that cannot reach the current leader immediately promote itself to leader, with no requirement to first confirm it has majority support from the rest of the cluster. What does this represent?
This is a missed split-brain-prevention requirement, since promoting without confirming majority support can let two disconnected sides each produce their own leader simultaneously. A cache eviction policy is an unrelated concept about discarded cache entries. This promote-without-majority-check pattern is exactly the kind of correctness gap a reviewer flags once network partitions are a realistic failure mode.
4 / 5
An incident report shows a cluster ended up with two active leaders accepting conflicting writes after a network partition, because the leader-election logic let any node promote itself the moment it lost contact with the previous leader, without first confirming it had majority support. What practice would prevent this?
Requiring a strict majority quorum before any node can be promoted to leader ensures at most one side of any partition can ever gather enough votes to become leader. Continuing to let any node promote itself to leader the moment it loses contact with the previous leader regardless of whether it has majority support is exactly what caused the dual-leader incident described here. This majority-quorum-required approach is the standard fix once self-promotion without majority checks is confirmed to produce split brain.
5 / 5
During a PR review, a teammate asks why the team enforces a strict majority quorum for leader election instead of simply letting whichever node notices the leader is unreachable first take over immediately, given that immediate takeover minimizes failover time. What is the reasoning?
A majority quorum trades a small amount of failover latency, since a candidate must first collect votes, for a mathematical guarantee that split brain cannot occur, while immediate takeover minimizes latency but risks two simultaneous leaders if a partition splits the cluster. This is exactly why a strict majority quorum is required for leader election in any cluster that must tolerate network partitions, while immediate takeover remains dangerous specifically because it skips that guarantee.