Device not shown if occupied by a task

[ruspaul013]

Nomad version

[root@base-machine ~]# nomad version
Nomad v1.9.6

Operating system and Environment details

Plugin "nomad-usb-device" v0.4.0
OS: AlmaLinux 9.5

Issue

When you inspect a node, for devices will appear something like this:

[root@base-machine ~]# nomad node status 1f5a5301
Device Resource Utilization
0627/usb/0001[001-002-001-0627-0001]  <none>

Even if the resource is in use, this is not shown. So it's not possible to know which resources are currently used. I know that allocations could be denied due to lack of resources, but cannot be tracked to the node level.

Reproduction steps

1. install a device plugin (usb/nvidia)
2. run a job with device block
3. check node status

Expected Result

I expect to see which task uses the resources.

[root@base-machine ~]# nomad node status 1f5a5301
Device Resource Utilization
0627/usb/0001[001-002-001-0627-0001]  <alloc_id>

Actual Result

It shows the device stats or none ( based on the device used ).

[root@base-machine ~]# nomad node status 1f5a5301
Device Resource Utilization
0627/usb/0001[001-002-001-0627-0001]  <none>

Job file (if appropriate)

job "redis" {
  datacenters = ["dc1"]
  type        = "service"

  group "redis" {
    network {
      port "redis" { to = 6379 }
    }

    task "redis" {
      driver = "podman"

        config {
          image = "docker://redis"
          ports = ["redis"]
        }

      resources {
                memory = 1024
                cores = 2
                device "0627/usb/0001" {}
            }
    }
  }
}

[Juanadelacuesta]

Hi @ruspaul013, Im looking at the issue you are reporting and I wanted to know if you are installing the Nomad USB Device Plugin for the correct distribution you are using under the plugin_dir in your nomad clients and configuring it as recommended?

[ruspaul013]

@Juanadelacuesta yes, This is my nomad config:

name="test"
data_dir  = "/var/lib/nomad"
plugin_dir = "/var/lib/nomad/plugins/"
log_level="info"

bind_addr = "0.0.0.0" # the default


client {
  enabled        = true
  cni_path       = "/opt/cni/bin"
  servers         = ["10.150.92.131:4647"]
}



plugin "nomad-driver-podman" {
    config {
        disable_log_collection = false
        socket_path = "unix://run/podman/podman.sock"
        client_http_timeout = "30s"
        extra_labels = ["job_name", "job_id", "task_group_name", "task_name", "namespace", "node_name", "node_id"]

        gc {
            container = true
        }

        volumes {
            enabled      = true
            selinuxlabel = "z"
        }
    }
}

plugin "nomad-driver-usb" {
    config {
        enabled = true

        included_vendor_ids = [0x1d6b,0x0627]
        excluded_vendor_ids = []

        included_product_ids = [0x0001,0x0001]
        excluded_product_ids = []

        fingerprint_period = "1m"
    }
}

server {
  enabled          = true
  bootstrap_expect = 1
}

plugin_dir:

[root@base-machine bp_agent]# ls /var/lib/nomad/plugins/
nomad-driver-nvidia  nomad-driver-podman  nomad-driver-usb

I may have labeled it wrong, I don't think it is a bug - maybe a feature.