ALB Health Check Configuration for SAP

Quick Answer

When an Application Load Balancer keeps sending traffic to instances that are actually unhealthy, the root cause is almost always that the load balancer's own definition of healthy doesn't match reality, which is controlled entirely by the health check configuration - its thresholds and check interval. Configuring appropriate healthy and unhealthy thresholds along with a suitable check interval ensures the ALB actually detects a failing target within a reasonable time and stops routing new traffic to it, since those settings determine how many consecutive failed or successful checks it takes before the ALB changes its view of a target's health, and how frequently it re-evaluates that status. The other configuration options mentioned don't touch this detection mechanism at all: sticky sessions only control whether a client keeps getting routed to the same target across requests, which has nothing to do with whether that target is deemed healthy; deregistration delay only affects how long the load balancer keeps draining existing in-flight connections to a target before fully removing it, not how it detects unhealthy targets in the first place; and cross-zone load balancing controls how traffic is distributed across Availability Zones, not whether unhealthy targets are excluded from receiving traffic. The pattern worth remembering is that any time a load balancer appears to be routing to targets that should be considered down, the fix almost always lives in the health check settings themselves, rather than in session, deregistration, or zone-balancing settings.