Cluster Autoscaler project study

Github Repository: https://github.com/kubernetes/autoscaler

1. cluster-autoscaler start up commands

    // I0624 09:24:10.787568       1 main.go:333] Cluster Autoscaler 1.13.5
   
    --address=":8085"
    // Logs are written to standard error as well as to files.
    --alsologtostderr="false"
    --balance-similar-node-groups="false"
    --cloud-config=""
    --cloud-provider="azure"
    --cloud-provider-gce-lb-src-cidrs="130.211.0.0/22,209.85.152.0/22,209.85.204.0/22,35.191.0.0/16"
    --cluster-name=""
    --cores-total="0:320000"
    --estimator="binpacking"
   
    // determine how to choose a vmss to expand nodes.
    --expander="random"
   
    --expendable-pods-priority-cutoff="-10"
    --gke-api-endpoint=""
    --gpu-total="[]"
    --httptest.serve=""
    --ignore-daemonsets-utilization="false"
    --ignore-mirror-pods-utilization="false"
    --kubeconfig=""
    --kubernetes=""
    --leader-elect="true"
    --leader-elect-lease-duration="15s"
    --leader-elect-renew-deadline="10s"
    --leader-elect-resource-lock="endpoints"
    --leader-elect-retry-period="2s"
    /*
      When set to a file and line number holding a logging statement, such as
        -log_backtrace_at=gopherflakes.go:234
      a stack trace will be written to the Info log whenever executionhits that statement. 
      (Unlike with -vmodule, the ".go" must be present.)
    */
    --log-backtrace-at=":0"
    // Log files will be written to this directory instead of the
    //        default temporary directory.
    --log-dir=""
    --log-file=""
   
    // Logs are written to standard error instead of to files.
    --logtostderr="true"
    --max-autoprovisioned-node-group-count="15"
    --max-empty-bulk-delete="10"
    --max-failing-time="15m0s"
    --max-graceful-termination-sec="600"
    --max-inactivity="10m0s"
    --max-node-provision-time="15m0s"
    --max-nodes-total="0"
    --max-total-unready-percentage="45"
    --memory-total="0:6400000"
    --min-replica-count="0"
    --namespace="kube-system"
    --new-pod-scale-up-delay="0s"
    --node-autoprovisioning-enabled="false"
    --node-group-auto-discovery="[]"
    --nodes="[3:10:aks-mynodepool-41521705-vmss]"
    --ok-total-unready-count="3"
    --regional="false"
    --scale-down-candidates-pool-min-count="50"
    --scale-down-candidates-pool-ratio="0.1"
    --scale-down-delay-after-add="10m0s"
    --scale-down-delay-after-delete="10s"
    --scale-down-delay-after-failure="3m0s"
    // Flag to enable scale down or not.
    --scale-down-enabled="true"
    --scale-down-non-empty-candidates-count="30"
    --scale-down-unneeded-time="10m0s"
    --scale-down-unready-time="20m0s"
    --scale-down-utilization-threshold="0.5"
    // The time interval to scan nodes (executing RunOnce()).
    --scan-interval="10s"        
    --skip-headers="false"
    --skip-nodes-with-local-storage="false"
    --skip-nodes-with-system-pods="true"
   
    // Log events at or above this severity are logged to standard
    //        error as well as to files.
    --stderrthreshold="2"
    --test.bench=""
    --test.benchmem="false"
    --test.benchtime="1s"
    --test.blockprofile=""
    --test.blockprofilerate="1"
    --test.count="1"
    --test.coverprofile=""
    --test.cpu=""
    --test.cpuprofile=""
    --test.failfast="false"
    --test.list=""
    --test.memprofile=""
    --test.memprofilerate="0"
    --test.mutexprofile=""
    --test.mutexprofilefraction="1"
    --test.outputdir=""
    --test.parallel="2"
    --test.run=""
    --test.short="false"
    --test.testlogfile=""
    --test.timeout="0s"
    --test.trace=""
    --test.v="false"
    --unremovable-node-recheck-timeout="5m0s"
   
    // klog level. Enable V-leveled logging at the specified level.
    --v="3"                                            
    /*
      klog module level. 
      The syntax of the argument is a comma-separated list of pattern=N,
          where pattern is a literal file name (minus the ".go" suffix) or "glob" pattern and N is a V level. For instance,
            -vmodule=gopher*=3
          sets the V level to 3 in all Go files whose names begin "gopher".
    */
    --vmodule=""
    --write-status-configmap="true"
   

2. autoscaler.go

   struct AutoscalerOptions is the whole set of options for configuring an autoscaler. The struct is mainly initialised in initializeDefaultOptions().

    type AutoscalerOptions struct {
      config.AutoscalingOptions
      KubeClient             kube_client.Interface
      AutoscalingKubeClients *context.AutoscalingKubeClients
      CloudProvider          cloudprovider.CloudProvider
      PredicateChecker       *simulator.PredicateChecker
      ExpanderStrategy       expander.Strategy
      EstimatorBuilder       estimator.EstimatorBuilder
      Processors             *ca_processors.AutoscalingProcessors
      Backoff                backoff.Backoff
    }
   

   • config.AutoscalingOptions: options coming from startup command.
   • ExpanderStrategy: Determine how to choose a vmss to expand nodes.
   • EstimatorBuilder: build a edtimator which implements approximation algorithm to calculate pod score according to cpu and memory.
   • Backoff: ???