Skip to content

Latest commit

 

History

History
257 lines (217 loc) · 8.68 KB

netdns.md

File metadata and controls

257 lines (217 loc) · 8.68 KB
Raw

netdns

简体中文 | English

Introduction

kdoctor-controller creates the necessary resources, including agent, based on the agentSpec. Each agent Pod sends DNS requests to a specified DNS server. By default, the concurrency level is set to 50, which can handle scenarios with multiple replicas. The concurrency level can be configured in the kdoctor configmap. The success rate and average latency are measured, and the results are evaluated based on predefined success criteria. Detailed reports can be obtained using the aggregate API.

  1. Use Cases:

    • Test the accessibility of the CoreDNS service from all corners of the cluster in production or E2E environments.
    • Adjust CoreDNS resources and replica count during application deployment to ensure it can handle the expected load.
    • Apply stress to CoreDNS for purposes like testing upgrades, chaos tests, bug reproduction, etc
    • Test external DNS services from within the cluster.

  2. For more detailed information about the NetDns CRD, refer to NetDns

  3. Features:

    • Support testing DNS servers both inside and outside the cluster
    • Supports typeA and typeAAAA records
    • Supports UDP, TCP, and TCP-TLS protocols

Steps

The following example demonstrates how to use NetDNS.

Install kdoctor

Follow the installation guide to install kdoctor.

Install Test Server (Optional)

The official kdoctor repository includes an application called "server" that contains an HTTP server, HTTPS server, and DNS server. This server can be employed to test the functionality of kdoctor. If you have other test servers available, you can skip this installation step.

helm repo add kdoctor https://kdoctor-io.github.io/kdoctor
helm repo update kdoctor
helm install server kdoctor/server -n kdoctor-test-server --wait --debug --create-namespace

Check the status of test server

kubectl get pod -n kdoctor -owide
NAME                                READY   STATUS    RESTARTS   AGE   IP            NODE                    NOMINATED NODE   READINESS GATES
server-7649566ff9-dv4jc   1/1     Running   0          76s   172.40.1.45   kdoctor-worker          <none>           <none>
server-7649566ff9-qc5dh   1/1     Running   0          76s   172.40.0.35   kdoctor-control-plane   <none>           <none>

Obtain the service address of the test server

kubectl get service -n kdoctor 
NAME               TYPE        CLUSTER-IP    EXTERNAL-IP   PORT(S)                                AGE
server   ClusterIP   172.41.71.0   <none>        80/TCP,443/TCP,53/UDP,53/TCP,853/TCP   2m31s

Create NetDns

Create NetDns object that will execute a 10-second continuous task. The task will send UDP requests to the cluster's internal DNS server at a rate of 10 QPS. It will request the typeA records whose domain name is kubernetes.default.svc.cluster.local.

cat <<EOF | kubectl apply -f -
apiVersion: kdoctor.io/v1beta1
kind: Netdns
metadata:
  name: cluster
spec:
  expect:
    meanAccessDelayInMs: 1500
    successRate: 1
  request:
    domain: kubernetes.default.svc.cluster.local
    durationInSecond: 10
    perRequestTimeoutInMS: 1000
    protocol: udp
    qps: 10
  schedule:
    roundNumber: 1
    roundTimeoutMinute: 1
    schedule: 0 1
  target:
    targetDns:
      serviceName: kube-dns
      serviceNamespace: kube-system
      testIPv4: true
EOF

Check Task Status

After completing a round of tasks, you can use the kdoctor aggregate API to view the report for the current round. When the FINISH field is set to true, it indicates that all tasks have been completed, and you can access the overall report.

kubectl get netdns
NAME             FINISH   EXPECTEDROUND   DONEROUND   LASTROUNDSTATUS   SCHEDULE
cluster           true     1               1           succeed           0 1
  • FINISH: indicate whether the task has been completed
  • EXPECTEDROUND: number of expected task rounds
  • DONEROUND: number of completed task rounds
  • LASTROUNDSTATUS: execution status of the last round of tasks
  • SCHEDULE: schedule rules for the task

View Task Reports

  1. View existed reports

    kubectl get kdoctorreport
NAME             CREATED AT
netdns-cluster   0001-01-01T00:00:00Z

  2. View specific task reports

    The reports are aggregated from the agents running on both the kdoctor-control-plane node and the kdoctor-worker nodes after performing two rounds of stress testing respectively,Report name consists of ${TaskKind}-${TaskName}

    root@kdoctor-control-plane:/# kubectl get kdoctorreport netdns-cluster -oyaml
apiVersion: system.kdoctor.io/v1beta1
kind: KdoctorReport
metadata:
  creationTimestamp: null
  name: netdns-cluster
spec:
  FailedRoundNumber: null
  FinishedRoundNumber: 1
  Report:
  - NodeName: kdoctor-control-plane
    PodName: kdoctor-agent-ntp9l
    ReportType: agent test report
    RoundDuration: 11.025723086s
    RoundNumber: 1
    RoundResult: succeed
    StartTimeStamp: "2023-08-01T09:09:39Z"
    EndTimeStamp: "2023-08-01T09:09:50Z"
    TaskName: netdns.cluster
    TaskType: Netdns
    netDNSTask:
      detail:
      - FailureReason: null
        MeanDelay: 0.2970297
        Metrics:
          DNSMethod: udp
          DNSServer: 172.41.0.10:53
          Duration: 11.002666395s
          EndTime: "2023-08-01T09:09:50Z"
          Errors: {}
          FailedCounts: 0
          Latencies:
            MaxInMs: 0
            MeanInMs: 0.2970297
            MinInMs: 0
            P50InMs: 0
            P90InMs: 0
            P95InMs: 0
            P99InMs: 0
          ReplyCode:
            NOERROR: 101
          RequestCounts: 101
          StartTime: "2023-08-01T09:09:39Z"
          SuccessCounts: 101
          TPS: 9.179593052634765
          TargetDomain: kubernetes.default.svc.cluster.local.
        Succeed: true
        SucceedRate: 1
        TargetName: typeA_172.41.0.10:53_kubernetes.default.svc.cluster.local
        TargetProtocol: udp
        TargetServer: 172.41.0.10:53
      succeed: true
      targetNumber: 1
      targetType: kdoctor agent
      MaxCPU: 30.651%
      MaxMemory: 97.00MB
    netDNSTaskSpec:
      ...
  - NodeName: kdoctor-worker
    PodName: kdoctor-agent-krrnp
    ReportType: agent test report
    RoundDuration: 10.024533428s
    RoundNumber: 1
    RoundResult: succeed
    StartTimeStamp: "2023-08-01T09:09:39Z"
    EndTimeStamp: "2023-08-01T09:09:49Z"
    TaskName: netdns.cluster
    TaskType: Netdns
    netDNSTask:
      detail:
      - FailureReason: null
        MeanDelay: 0.58
        Metrics:
          ...
        Succeed: true
        SucceedRate: 1
        TargetName: typeA_172.41.0.10:53_kubernetes.default.svc.cluster.local
        TargetProtocol: udp
        TargetServer: 172.41.0.10:53
      succeed: true
      targetNumber: 1
      targetType: kdoctor agent
      MaxCPU: 30.651%
      MaxMemory: 97.00MB
    netDNSTaskSpec:
      ...
  ReportRoundNumber: 1
  RoundNumber: 1
  Status: Finished
  TaskName: cluster
  TaskType: Netdns

If the reports do not align with the expected results, check the MaxCPU and MaxMemory fields in the report to verify if there are available resources of the agents and adjust the resource limits for the agents accordingly.

Test an External DNS Server

Below are examples of HTTP and HTTPS requests with body:

  1. Create NetDns object that will execute a 10-second continuous task. The task will send UDP requests to the cluster's internal DNS server at a rate of 10 QPS. It will request the typeA records whose domain name is kubernetes.default.svc.cluster.local.

    We are using the service address of the test server. If you have a different server address available, feel free to use it instead.

    Create NetDns

    SERVER="172.41.71.0"
apiVersion: kdoctor.io/v1beta1
kind: Netdns
metadata:
  name: user
spec:
  expect:
    meanAccessDelayInMs: 1500
    successRate: 1
  request:
    domain: kubernetes.default.svc.cluster.local
    durationInSecond: 10
    perRequestTimeoutInMS: 1000
    protocol: udp
    qps: 10
  schedule:
    roundNumber: 1
    roundTimeoutMinute: 1
    schedule: 0 1
  target:
    targetUser:
      port: 53
      server: ${SERVER}
EOF

Environment Cleanup

kubectl delete netdns cluster user