Allocation scaling up to max limit and not scaling down when using Autoscaler and Nomad as apm

Nomad
1

Allocation is being scaled to max limit (here 3) defined under the scaling block, and it’s not scaling down although the average cpu and memory utilization is much below target value when using nomad apm. I am attaching my job template can anyone correct me if I’m wrong anywhere.

job "scaling-example" {
  datacenters = ["aws-us-east-1"]
  region      = "aws-us-east-1"
  namespace   = "devops"
  type        = "service"
  priority    = 50

  constraint {
        attribute = "${node.class}"
        value     = "general"
    }

    group "nomad-scaling" {
        count = 1
        scaling {
            min     = 1
            max     = 3
            enabled = true

            policy {
                evaluation_interval = "5s"
                cooldown            = "1m"

                check "cpu_allocated_percentage" {
                    source       = "nomad-apm"
                    query        = "avg_cpu"
                    query_window = "1m"
                    strategy "target-value" {
                        target = 50
                    }
                }

                check "memory_allocated_percentage" {
                    source       = "nomad-apm"
                    query        = "avg_memory"
                    query_window = "1m"
                    strategy "target-value" {
                        target = 50
                    }
                }
            }
        }
        restart {
        attempts = 3
        interval = "5m"
        delay = "30s"
        mode = "fail"
        }
        volume "ca-certificates" {
        type      = "host"
        read_only = true
        source    = "ca-certificates"
        }
        network {
        mode = "bridge"
        port "http" { to = 80 }
        }

        task "scaling-test" {
            driver = "docker"
            user   =  "root"
            volume_mount {
                volume      = "ca-certificates"
                destination = "/etc/ssl/certs"
            }

            config {
                image     = "nginx:latest"
            }
            service {
                name = "nomad-dynamic-scaling"
                port = "http"
                check {
                    name     = "alive"
                    type     = "http"
                    path     = "/"
                    interval = "60s"
                    timeout  = "5s"

                    check_restart {
                    limit           = 5
                    grace           = "90s"
                    ignore_warnings = false
                    }
                }
            }
            resources {
                cpu    = 100
                memory = 100
            }
        }
    }
}
2

Hi @Puskal07,

Do you have the nomad-autoscaler logs available (ideally at debug level)? We cannot tell what is happening from the job specification, whereas the autoscaler logs should be able to tell what calculations are being made.

Thanks,
jrasell and the Nomad team