[[inputs.diskio]] not honoring HOST_MOUNT_PREFIX / Missing HOST_DEV environment variable

[Tschebbischeff]

Relevant telegraf.conf

[[inputs.net]] # https://docs.influxdata.com/telegraf/v1/input-plugins/net/
  interfaces = ["eth*", "enp*", "wlan*"]

Logs from Telegraf

W! Strict environment variable handling is the new default starting with v1.38.0! If your configuration does not work with strict handling please explicitly add the --non-strict-env-handling flag to switch to the previous behavior!
I! Loading config: /etc/telegraf/telegraf.conf
I! Starting Telegraf 1.38.2 brought to you by InfluxData the makers of InfluxDB
I! Available plugins: 244 inputs, 9 aggregators, 35 processors, 26 parsers, 68 outputs, 8 secret-stores
I! Loaded inputs: cpu disk diskio docker mem mqtt_consumer (2x) net
I! Loaded aggregators:
I! Loaded processors:
I! Loaded secretstores: docker_secrets
I! Loaded outputs: health influxdb_v3
I! Tags enabled: host=__REDACTED__
I! [agent] Config: Interval:10s, Quiet:false, Hostname:"__REDACTED__", Flush Interval:10s
I! [outputs.health] Listening on http://[::]:8080
I! [inputs.mqtt_consumer] Connected [tcp://__REDACTED___mosquitto:1883]
I! [inputs.mqtt_consumer] Connected [tcp://__REDACTED___mosquitto:1883]
W! [inputs.diskio] Unable to gather disk name for "mmcblk0p1": error reading /dev/mmcblk0p1: no such file or directory
W! [inputs.diskio] Unable to gather disk name for "mmcblk0p2": error reading /dev/mmcblk0p2: no such file or directory
W! [inputs.diskio] Unable to gather disk name for "zram0": error reading /dev/zram0: no such file or directory
W! [inputs.diskio] Unable to gather disk name for "mmcblk0": error reading /dev/mmcblk0: no such file or directory

System info

Telegraf 1.38.2, Linux Host (NixOS 25.11), Docker Image: telegraf (latest)

Docker

  telegraf:
    # build: # Removed for minimal diskio-issue-check config, note that I am running telegraf as root to access secrets properly, so I needed a custom entrypoint that just starts telegraf ignoring the root check in the original entrypoint, but this should not affect the diskio issue
    #   context: ./telegraf
    image: telegraf:latest
    container_name: telegraf
    # restart: always # Removed for minimal diskio-issue-check config
    # user: "root" # Removed for minimal diskio-issue-check config
    # env_file: # Removed for minimal diskio-issue-check config
    #   - ./.env
    #   - path: ${ENV_DIR}/.env
    #     required: false
    #   - path: ${ENV_DIR}/telegraf.env
    #     required: false
    environment:
      HOST_PROC: '/hostfs/proc'
      HOST_SYS: '/hostfs/sys'
      HOST_ETC: '/hostfs/etc'
      HOST_VAR: '/hostfs/var'
      HOST_RUN: '/hostfs/run'
      HOST_DEV: '/hostfs/dev'
      HOST_MOUNT_PREFIX: '/hostfs'
      HOST_ROOT: '/hostfs'
    # secrets: # Removed for minimal diskio-issue-check config
    #   - INFLUXDB_ADMIN_TOKEN
    volumes:
      - ./telegraf/config/telegraf.conf:/etc/telegraf/telegraf.conf:ro
      - /proc:/hostfs/proc:ro
      - /sys:/hostfs/sys:ro
      # - /etc:/hostfs/etc:ro
      # - /var:/hostfs/var:ro
      # - /run:/hostfs/run:ro
      - /run/udev:/hostfs/run/udev:ro
      - /dev:/hostfs/dev:ro
      # - /etc/os-release:/hostfs/etc/os-release:ro
    # networks: # Removed for minimal diskio-issue-check config
    #   internal:
    #     ipv4_address: '172.18.0.11'
    # healthcheck: # Removed for minimal diskio-issue-check config
    #   test: [ "CMD", "wget", "--spider", "-q", "http://127.0.0.1:8080/health" ]
    #   interval: 30s
    #   timeout: 5s
    #   retries: 5
    #   start_period: 15s
    # depends_on: # Removed for minimal diskio-issue-check config
    #   socket-proxy:
    #     condition: service_healthy

Steps to reproduce

1. Start a docker image with the [[inputs.diskio]] input without giving the container privileged access and make sure to override all env vars needed my psutils/ telegraf (I've pasted my config and commented out what I feel isn't necessary for the test, I did not mount /etc, /var and /run)
2. Observe how the telegraf container uses /dev/$deviceName instead of /hostfs/dev/$deviceName and reports missing devices (requires host to have devices the container doesn't have of course)

Expected behavior

Devices are searched for in /hostfs/dev according to the environment variables $HOST_DEV and/ or $HOST_MOUNT_PREFIX and/or $HOST_ROOT

Actual behavior

Host devices are searched for in /dev and reported as not being found

Additional info

This line of code hardcodes /dev/{deviceName} and should use the environment variable or a function returning the path from the environment variable similar to getProcPath():

telegraf/plugins/inputs/diskio/diskio_linux.go

Line 25 in ec770d9

path := "/dev/" + devName

HOST_DEV and HOST_ROOT are variable names I found in the psutils documentation, since the other plugins use the same environment variables and diskio uses psutils like them, I think HOST_DEV should be used here, but if you do not want to support this variable (it's not mentioned in your documentation) it should at least honor HOST_MOUNT_PREFIX.

Notably I still receive data about these devices in my InfluxDB so it is not breaking the metrics.