CSI volumes links on the job page lead to a 404

[apollo13]

Nomad version

0.12.5

Issue

I managed to mount a CSI volume and it shows up properly under the job volume requirements in the UI:

Now the link in the first column leads to: https://nomad.infra.domain/ui/csi/volumes/test while I think it should lead to https://nomad.infra.domain/ui/csi/volumes/wiki

The first name (test) is the named as specified in the job file:

    volume "test" {
      type      = "csi"
      read_only = false
      source    = "wiki"
    }

but the CSI name is the name from the source column.

Reproduction steps

Create a job using a volume that has a different name than it's source

Job file (if appropriate)

job "infra.test" {
  datacenters = ["dc1"]

  group "test" {
    volume "test" {
      type      = "csi"
      read_only = false
      source    = "wiki"
    }


    task "ubuntu" {
      driver = "docker"
      volume_mount {
       volume      = "test"
       destination = "/test"
       read_only   = false
      }

      config {
        image = "ubuntu:20.04"
        command = "sleep"
	args = ["infinity"]
      }
    }

  }
}

[tgross]

Hi @apollo13!

I was able to confirm this behavior on the task group page (ex. http://localhost:4646/ui/jobs/example/cache) for both 0.12.5 and the current master

I was also able to confirm this behavior on the individual allocation page (ex. http://localhost:4646/ui/allocations/09965ad4-a4b8-261b-6b01-03dcb77b473e ), in which case the screenshot for that section looks like this:

[tgross]

Ok, so I've narrowed down what's happening here. Here's my volume registration:

id = "test-volume0"
type = "csi"
external_id = "EXTERNAL_ID"
plugin_id = "hostpath-plugin0"
access_mode = "single-node-writer"
attachment_mode = "file-system"

And the relevant part of the jobspec:

job "example" {

  group "cache" {
    volume "myvolume" {
      type   = "csi"
      source = "test-volume0"
    }

    task "redis" {
      volume_mount {
        volume      = "myvolume"
        destination = "${NOMAD_ALLOC_DIR}/test"
      }
    }
  }
}

But when I hit the allocations API that's populating this data:

$ curl -s http://localhost:4646/v1/allocation/:alloc_id | jq '.Job.TaskGroups[0].Volumes'
{
  "myvolume": {
    "MountOptions": null,
    "Name": "myvolume",
    "ReadOnly": false,
    "Source": "test-volume0",
    "Type": "csi"
  }
}

So the API is returning myvolume as the name of the volume within the task group. But that's not how you query for a volume (ref API). These two UI sections should both be populating the URL of the link with the source and not the name of the volume. I think the allocation page has the nicest way of showing that -- it's showing both the bolded name and the link to the real volume, but the link we construct is wrong.

I definitely tested this out a bunch... using a volume name in the task group that was a valid prefix for the real volume name! (ex. test is a valid prefix of test-volume). 🤦

[tgross]

See also: #9297, which suggests the issue is somewhere in the API and not the UI.

I'll self-assign this issue and follow-up if it turns out to coincidentally be a problem in both the CLI and the UI by accident, and I can't manage to do the tiny UI PR myself.

[tgross]

I have a patch up for both this and the nomad alloc status -verbose fix (#9354 #9355), which turned out to be coincidentally the same bug in both the CLI and UI and not the API after all.

[tgross]

Fixed in #9355 and this will ship in 1.0

[apollo13]

Lovely, thanks!

[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.