id: PG-32
category: vm-lifecycle
title: Boot from volume fails?
situation: >-
  Local-disk instances build fine. Those whose root disk is a Cinder volume are refused, go to `ERROR`, or
  boot nothing.

diagram:
  type: branches
  head: [What happens, Look at]
  split: 52
  rows:
    - see: Refused at once, no instance created
      means: The request
      sub: Size, bootable flag, boot index, zone
      tone: mute
    - see: "`ERROR`, the volume `error` or still `creating`"
      means: Cinder
      sub: It failed, or was not ready in about 3 min
      tone: stop
    - see: "`ERROR` after retries, the volume `available`"
      means: The compute
      sub: It cannot connect to the volume
      tone: stop
    - see: "`ACTIVE`, but nothing boots"
      means: What the volume holds
      sub: No system on it, or firmware and bus properties
      tone: warn

checks:
  - label: Status and fault of the instance
    hint: openstack server show <id> -c status -c fault
  - label: The volume of the build
    hint: openstack volume show <volume> -c status -c bootable
  - label: Cinder's own explanation
    hint: openstack --os-volume-api-version 3.3 volume message list
  - label: Are the volume services up?
    hint: openstack volume service list

caution:
  - level: avoid
    text: >-
      Retrying blindly. Each retry from an image creates another volume, which counts in the quota and
      fills the backend.
  - level: destructive
    text: >-
      Deleting the failed instance also deletes its volume when "Delete On Termination" is true.

lesson: The fault names the volume. Read its status and Cinder's message before you retry.

runbook:
  path: scheduling/instance-fails-to-boot-from-volume.md
  status: lab-verified
  release: "2026.1"
  also:
    - volumes/volume-stuck-in-creating-or-error-after-create.md

scope: >-
  Written from the Nova and Cinder 2026.1 sources; its five causes were partly reproduced on an all-in-one lab
  with Cinder LVM over iSCSI. The three minutes are Nova's default wait (block_device_allocate_retries). Cinder
  2026.1 refuses a qcow2 image declared raw: the volume goes to error, it is not written as it is. Fixing an
  image property and rebooting does not help: the instance keeps the properties copied at build time. Fault
  wording varies with the client and microversion. Ironic and encrypted volumes are not covered.

linkedin:
  title: "Boot from volume fails: Nova reports it, Cinder explains it"
  caption: |
    Instances with a local disk build fine. The ones that boot from a volume do not.

    What happens tells you where to look:
    → refused at once, no instance: the request itself. Volume smaller than the image's min_disk, a volume that is not bootable, no boot index 0.
    → ERROR, the volume in "error" or still "creating": Cinder. It failed, or the volume was not ready when Nova stopped waiting, after about three minutes by default.
    → ERROR with "Exceeded maximum number of retries", the volume "available": the compute. It could not connect to the volume, and neither could the others.
    → ACTIVE, but nothing boots: what the volume holds. A blank volume marked bootable, or an image that needs a firmware or disk bus the instance did not get.

    For a volume in error, Cinder usually says why in one line: openstack volume message list.

    What not to do: retry in a loop. Each retry from an image creates another volume, which counts in the quota.
  hashtags: [OpenStack, Nova, Cinder, Storage, CloudOperations, OzarLand]
