Nomad CLI exec and logs randomly fails when used with -task and -job combination

[Ranjandas]

Nomad version

$ nomad version
Nomad v1.9.5
BuildDate 2025-01-14T18:35:12Z
Revision 0b7bb8b60758981dae2a78a0946742e09f8316f5+CHANGES

Operating system and Environment details

OS Independent (happens on macOS and Linux)

Issue

The nomad CLI subcommand exec and logs fails randomly when used with -task and -job combinations, when used without explicitly specifying allocation ID.

Reproduction steps

(ref: https://asciinema.org/a/0oRoEEcho4rs78NosVvOgm8jn)

• Create the countdash example job

  nomad init -short -connect
  

• Run the job against the cluster

  nomad run example.nomad.hcl
  

• Try to exec or use logs subcommand against a task without explicitly specifying the allocation ID

  nomad exec -task web -job countdash sh 
  

  or

  nomad logs -task web -job countdash sh
  

Both the above subcommands would randomly fail to pick the right task.

Expected Result

❯ nomad exec -task web -job countdash sh                                                                                                                                                  
/opt 

Actual Result

❯ nomad exec -task web -job countdash sh                                                                                                                                                  
Could not find task named: web, found:                                                                                                                                                       
dashboard                                                                                                                                                                                    
connect-proxy-count-dashboard

However, repeating the command multiple times will finally work.

Job file (if appropriate)

nomad init -short -connect

Nomad Server logs (if appropriate)

Nomad Client logs (if appropriate)

[tgross]

The challenge here is that alloc exec picks a random allocation and the job you're targeting has multiple groups. So it's picking an allocation at random, but not all allocations will have the same tasks. We probably should add a -group flag to cover this case.