Using Scheduler

Introduction

Using the TFGrid scheduler enables users to automatically get the nodes that match their criterias. We present here some basic information on this resource.

How the Scheduler Works

To better understand the scheduler, we summarize the main process:

• At first if farm_id is specified, then the scheduler will check if this farm has the Farmerbot enabled
  • If so it will try to find a suitable node using the Farmerbot.
• If the Farmerbot is not enabled, it will use grid proxy to find a suitable node.

Quick Example

Let's take a look at the following example:

terraform {
  required_providers {
    grid = {
      source = "threefoldtech/grid"
      version = "1.8.1-dev"
    }
  }
}
provider "grid" {
}

locals {
  name = "testvm"
}

resource "grid_scheduler" "sched" {
  requests {
    farm_id = 53
    name = "node1"
    cru  = 3
    sru  = 1024
    mru  = 2048
    node_exclude = [33] # exlude node 33 from your search
    public_ips_count = 0 # this deployment needs 0 public ips
    public_config = false # this node does not need to have public config
  }
}

resource "grid_network" "net1" {
  name        = local.name
  nodes       = [grid_scheduler.sched.nodes["node1"]]
  ip_range    = "10.1.0.0/16"
  description = "newer network"
}
resource "grid_deployment" "d1" {
  name         = local.name
  node         = grid_scheduler.sched.nodes["node1"]
  network_name = grid_network.net1.name
  vms {
    name       = "vm1"
    flist      = "https://hub.grid.tf/tf-official-apps/base:latest.flist"
    cpu        = 2
    memory     = 1024
    entrypoint = "/sbin/zinit init"
    env_vars = {
      SSH_KEY = file("~/.ssh/id_rsa.pub")
    }
    planetary = true
  }
  vms {
    name       = "anothervm"
    flist      = "https://hub.grid.tf/tf-official-apps/base:latest.flist"
    cpu        = 1
    memory     = 1024
    entrypoint = "/sbin/zinit init"
    env_vars = {
      SSH_KEY = file("~/.ssh/id_rsa.pub")
    }
    planetary = true
  }
}
output "vm1_ip" {
  value = grid_deployment.d1.vms[0].ip
}
output "vm1_planetary_ip" {
  value = grid_deployment.d1.vms[0].planetary_ip
}

output "vm2_ip" {
  value = grid_deployment.d1.vms[1].ip
}
output "vm2_planetary_ip" {
  value = grid_deployment.d1.vms[1].planetary_ip
}

From the example above, we take a closer look at the following section:

resource "grid_scheduler" "sched" {
  requests {
    name = "node1"
    cru  = 3
    sru  = 1024
    mru  = 2048
    node_exclude = [33] # exlude node 33 from your search
    public_ips_count = 0 # this deployment needs 0 public ips
    public_config = false # this node does not need to have public config
  }
}

In this case, the user is specifying the requirements which match the deployments.

Later on, the user can use the result of the scheduler which contains the [nodes] in the deployments:

resource "grid_network" "net1" {
  name        = local.name
  nodes       = [grid_scheduler.sched.nodes["node1"]]
  ...
}

and

resource "grid_deployment" "d1" {
  name         = local.name
  node         = grid_scheduler.sched.nodes["node1"]
  network_name = grid_network.net1.name
  vms {
    name       = "vm1"
    ...
  }
  ...
}