Load tests

wallet-scale.js — Phase 6.2

Goal: confirm whether wallet-api can run with replicas > 1 without
breaking sessions or wallet state under round-robin load balancing.

How to run

1. Bring up the cluster: ./deploy/k8s/scripts/k8s-deploy.sh up waltid
2. Apply baseline (1 replica):

   kubectl -n waltid scale deploy waltid-walt-wallet --replicas=1
   k6 run --env BASE_URL=https://wallet.verifiably.local --env VUS=50 \
          deploy/k8s/test/load/wallet-scale.js | tee report-1replica.txt
   

3. Apply scale-out (3 replicas):

   kubectl -n waltid scale deploy waltid-walt-wallet --replicas=3
   k6 run --env BASE_URL=https://wallet.verifiably.local --env VUS=50 \
          deploy/k8s/test/load/wallet-scale.js | tee report-3replicas.txt
   

4. Compare cross_pod_drift rate between the two.

Interpretation

• cross_pod_drift = 0 in both runs → wallet-api is stateless w.r.t.
  sessions; safe to enable HPA. Update walt-wallet/values.yaml
wallet.hpa.enabled: true.
• cross_pod_drift jumps in the 3-replica run → wallet-api keeps
  in-process session state. Options:
  • sticky sessions via ingress-nginx nginx.ingress.kubernetes.io/affinity: cookie
  • move session state to Redis (walt.id supports this — see ktor-server-sessions-redis)
    Document the chosen path in docs/k8s/wallet-scaling.md.

Status

Not yet run — pending Phase 8.1 cluster-up + first up waltid end-to-end.