Heartbeats and Self-Preservation

Registration in Eureka is a lease. Each client renews it with a heartbeat every 30 seconds. If the server gets no renewal for 90 seconds, the lease expires and the eviction task removes the instance. A clean shutdown deregisters immediately.

eureka:
  instance:
    lease-renewal-interval-in-seconds: 30     # heartbeat interval (default 30)
    lease-expiration-duration-in-seconds: 90  # expire after (default 90)

import java.util.*;

public class Main {
    public static void main(String[] args) {
        Registry registry = new Registry(90);
        registry.renew("order-service#1", 0);
        registry.renew("order-service#2", 0);
        registry.renew("product-service#1", 0);

        for (long t = 30; t <= 150; t += 30) {
            registry.renew("order-service#1", t);
            registry.renew("product-service#1", t);
            if (t <= 30) registry.renew("order-service#2", t); // #2 crashes after t=30s
            registry.evictExpired(t);
            System.out.println("t=" + t + "s registry=" + registry.names());
        }
    }
}

class Registry {
    private final long leaseSeconds;
    private final Map<String, Long> lastRenewal = new TreeMap<>();

    Registry(long leaseSeconds) { this.leaseSeconds = leaseSeconds; }

    void renew(String instance, long now) { lastRenewal.put(instance, now); }

    void evictExpired(long now) {
        lastRenewal.entrySet().removeIf(e -> {
            long silent = now - e.getValue();
            boolean expired = silent > leaseSeconds;
            if (expired) System.out.println("  EVICT " + e.getKey() + " (no heartbeat for " + silent + "s)");
            return expired;
        });
    }

    Set<String> names() { return lastRenewal.keySet(); }
}

t=30s registry=[order-service#1, order-service#2, product-service#1]
t=60s registry=[order-service#1, order-service#2, product-service#1]
t=90s registry=[order-service#1, order-service#2, product-service#1]
t=120s registry=[order-service#1, order-service#2, product-service#1]
  EVICT order-service#2 (no heartbeat for 120s)
t=150s registry=[order-service#1, product-service#1]

If a network problem stops many heartbeats at once, Eureka cannot tell "instances died" from "my network is broken". When renewals fall below 85% of the expected number, Eureka enters self-preservation: it stops evicting anyone. Better to keep a few dead entries than to wipe the whole registry.

open http://localhost:8761

EMERGENCY! EUREKA MAY BE INCORRECTLY CLAIMING INSTANCES ARE UP WHEN THEY'RE NOT.
RENEWALS ARE LESSER THAN THRESHOLD AND HENCE THE INSTANCES ARE NOT BEING EXPIRED JUST TO BE SAFE.