DNS Performance Testing with Kubernetes Pods

Integrated Items

• DNSPerf 2.14.0

Preflight checklist

Does CoreDNS running expected version?

$ kubectl get deployments coredns -n kube-system -o jsonpath='{$.spec.template.spec.containers[0].image}'

Does CoreDNS running with expected Corefile?

$ kubectl get configmap coredns -n kube-system -o jsonpath='{$.data.Corefile}'

Does CoreDNS running with correct resources configuration?

$ kubectl get deployments coredns -n kube-system -o jsonpath='{$.spec.template.spec.containers[0].resources}'

Benchmark with Kubernetes Pods

Apply pre-configured testing deployment/pods

$ kubectl apply -f https://raw.githubusercontent.com/guessi/dnsperf-bench/master/k8s-dnsperf-bench.yaml
configmap/dns-records-config created
deployment.apps/dnsperf created

Find out your Kubernetes DNS service IP address

$ kubectl get service kube-dns -n kube-system -o jsonpath='{$.spec.clusterIP}'

If your DNS service address is not "10.100.0.10", you will need to change the value of predefined "DNS_SERVER_ADDR"

$ kubectl edit deployment dnsperf

Make sure the deployment is running as expected

$ kubectl get deploy dnsperf

NAME      READY   UP-TO-DATE   AVAILABLE   AGE
dnsperf   1/1     1            1           81s

$ kubectl get pods -l app=dnsperf

NAME                       READY   STATUS    RESTARTS   AGE
dnsperf-7b9cc5b497-d5nfs   1/1     Running   0          1m16s

To check benchmark results

$ kubectl logs -f deployments/dnsperf

DNS Performance Testing Tool
Version 2.14.0

[Status] Command line: dnsperf -f any -m udp -s 10.100.0.10 -p 53 -d /opt/records.txt -c 1 -T 1 -l 30 -t 5 -Q 100000
[Status] Sending queries (to 10.100.0.10:53)
[Status] Started at: Thu Apr 11 16:47:46 2024
[Status] Stopping after 30.000000 seconds
[Status] Testing complete (time limit)

Statistics:

  Queries sent:         517317
  Queries completed:    517317 (100.00%)
  Queries lost:         0 (0.00%)

  Response codes:       NOERROR 517317 (100.00%)
  Average packet size:  request 43, response 160
  Run time (s):         30.004152
  Queries per second:   17241.513774

  Average Latency (s):  0.004805 (min 0.000101, max 0.029327)
  Latency StdDev (s):   0.002537

Check resources utilization of the CoreDNS deployment

$ kubectl top pods -n kube-system -l k8s-app=kube-dns

NAME                       CPU(cores)   MEMORY(bytes)
coredns-79989457d9-fn2hc   1644m        15Mi
coredns-79989457d9-xss5l   925m         16Mi

Stress Test

You may gain more replicas to stress your CoreDNS even harder

$ kubectl scale deployments/dnsperf --replicas 4

deployment.apps/dnsperf scaled

After gaining workload, you should find it's CPU utilization even higher

$ kubectl top pods -n kube-system -l k8s-app=kube-dns

NAME                       CPU(cores)   MEMORY(bytes)
coredns-79989457d9-fn2hc   1839m        17Mi
coredns-79989457d9-xss5l   1344m        16Mi

If you try to gain too much stress without tuning CoreDNS configureation (e.g. CPU, Memory, Replicas of CoreDNS), you should find some TIMEOUT, packet losts. That's expected... don't report it as bug! you should give CoreDNS more resources to handle that stress.

Reference

• https://www.dnsperf.com/