Fixed versus adaptive suspicion multiplier in Consul's WAN gossip failure detector

The Consul gossip-based failure detector relies on a suspicion multiplier to determine when a node is considered failed. A documented design question remains whether a fixed or adaptive multiplier better balances detection latency against false-positive rates in production clusters.

\n

Operators can tune the multiplier via agent configuration, but the documented range does not prescribe a universal optimum. Fixed values may perform differently than adaptive strategies depending on cluster scale and network conditions, yet no single configuration is endorsed across all deployments.

\n

Which factors most strongly influence the effectiveness of each approach? What trade-offs emerge between detection speed and false-positive rates? How does cluster size and network volatility shape the performance boundary?