The network stanza is not documented in resources

[remilapeyre]

The output of nomad job init -short is:

job "example" {
  datacenters = ["dc1"]

  group "cache" {
    task "redis" {
      driver = "docker"

      config {
        image = "redis:3.2"

        port_map {
          db = 6379
        }
      }

      resources {
        cpu    = 500
        memory = 256

        network {
          mbits = 10
          port "db" {}
        }
      }
    }
  }
}

with a network stanza in job > group > task > resources but it is not documented at https://www.nomadproject.io/docs/job-specification/resources.

[remilapeyre]

Also in the documentation for the network stanza the mbits attribute is not documented.

[remilapeyre]

volume_mount is also not in https://www.nomadproject.io/docs/job-specification/task, and I think it's missing restart too but I've not tried to use it in a task yet.

[notnoop]

Hi @remilapeyre ! Very sorry for the confusion here. We will update the sample docs.

FWIW, network under task > resources as well as mbits are being deprecated - but our documentation is a bit confusing there, as we didn't update all the locations they appear in. We will clarify that. Thanks for raising.

[Legogris]

@notnoop I can't find if there's a repo for learn.hashicorp.com (or a proper place to report that) but I noticed there's a good amount of guides and examples there explicitly relying on now deprecated (but without runtime warnings or errors to that end) behavior or configuration.

[shantanugadgil]

given that the network inside resources is being deprecated, what would be the right way to be able to allocate a single job to any node?

Assuming that I can force set the network_speed of all my worker nodes to say 1000, what can I use to be able to block an entire machine?

Previously I used to do it using mbits = 999 so no other job could get the machine as "free"

[nickethier]

This was fixed in #8911

@shantanugadgil that is an interesting unintended use of that! I think its also another example that it was not a useful resource from a networking perspective 😅

This usecase is being planned and is currently tracked in #2299

[shantanugadgil]

@nickethier I am confused ... I don't want jobs to land up on any specific node(s).
If I do want specific nodes, I can already achieve it via contraints or meta tags.

All I wanted to do using the network "mbits = 999" thing was to schedule (and allocate) 1-and-only-1 job (i.e. task) onto any node.

[shantanugadgil]

There was another ugly hack which I had discovered, which was to use a static port for the task, which would cause any other job/task trying to come onto the same node to fail, and eventually go elsewhere (or fail).

I keep mentioning job/task, as for now, all I have is 1-job-1-group-1-task scenario.

[github-actions]

I'm going to lock this issue because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.