id: PG-39
category: storage
title: Extended, same size inside?
situation: >-
  An attached volume was extended. The command is refused, the volume goes to `error_extending`, or Cinder
  shows the new size and the guest still sees the old one.

diagram:
  type: chain
  caption: An extend has four stages
  nodes:
    - step: "1"
      label: Cinder API
      sub: In-use volumes need volume API 3.42
      tone: ink
    - step: "2"
      label: Backend
      sub: "Grows the volume, or `error_extending`"
      tone: ink
    - step: "3"
      label: Nova
      sub: Gets the event, refreshes the device
      tone: ink
    - step: "4"
      label: Guest
      sub: "Partition and file system: its owner's job"
      tone: ink
  note: A volume cannot be shrunk afterwards, and the quota is charged for the new size.

checks:
  - label: "Cinder's view: size and status"
    hint: openstack volume show <volume> -c size -c status
  - label: Cinder's messages for the volume
    hint: openstack --os-volume-api-version 3.3 volume message list
  - label: Did Nova receive the event?
    hint: openstack server event list <server>
  - label: What the guest sees
    hint: lsblk (inside the guest)

caution:
  - level: state-changing
    text: >-
      Resetting `error_extending` changes Cinder's record only. Check first that the backend volume still has
      its old size.
  - level: state-changing
    text: >-
      A hard reboot shows the guest the new size, but stops it abruptly. Only if the event was lost or the
      refresh failed, and agreed with the owner.

lesson: Find the last stage that happened before you repeat the first.

runbook:
  path: volumes/extend-in-use-volume-guest-sees-old-size.md
  status: draft
  release: "2026.1"

scope: >-
  The runbook is a draft, written from the Cinder and Nova 2026.1 sources. Lab-run on an all-in-one with
  Cinder LVM over iSCSI: the refusals, a backend failure and Nova not refreshing the device. Not run: growing
  a file system in the guest, a refresh failure on the compute, Ceph RBD and Fibre Channel. When the pool has
  too little space the scheduler refuses the extend: the volume keeps its old size and status.

linkedin:
  title: "The volume was extended. The guest still sees the old size."
  caption: |
    "I extended the volume and nothing changed." Extending an attached volume is not one operation. It is four stages, and it can stop after any of them:

    1. The Cinder API accepts it. For an in-use volume that needs volume API 3.42.
    2. The backend grows the volume, or the volume goes to error_extending.
    3. Nova receives a volume-extended event and refreshes the device on the compute.
    4. Inside the guest, the partition and the file system are grown. That one is the owner's job.

    So before extending again, find the last stage that actually happened: Cinder's size and status, Cinder's messages for the volume, the instance's events, and lsblk inside the guest.

    And remember that a volume cannot be shrunk afterwards.

    Source runbook status: draft.
  hashtags: [OpenStack, Cinder, Nova, Storage, CloudOperations, OzarLand]
