id: PG-35
category: compute
title: After a reboot, VMs stay off?
situation: >-
  A rebooted compute answers again, but its instances stay `SHUTOFF`, or run without network while Nova
  shows `ACTIVE`.

diagram:
  type: branches
  head: [You find, The cause is]
  split: 48
  rows:
    - see: "Instances `SHUTOFF`, nothing else wrong"
      means: Nova's default
      sub: Guests are not resumed at host boot
      tone: mute
    - see: A service or container did not come back
      means: A failed start
      sub: "If it is `ovs-vswitchd`: no network"
      tone: stop
    - see: "External interface or bond `DOWN`"
      means: The host's network
      sub: Not in the host's own configuration
      tone: warn
    - see: A smaller MTU than the other computes
      means: Set by hand, then lost
      sub: Ping works, large transfers stall
      tone: warn

checks:
  - label: Instances, once nova-compute is up
    hint: openstack server list --all-projects --host <host> --long
  - label: "Kolla units: did all come back?"
    hint: "systemctl list-units 'kolla-*' --state=failed,inactive,activating"
  - label: Does Nova resume guests?
    hint: grep -rn resume_guests_state /etc/nova/
  - label: Interfaces up, expected MTU?
    hint: "ip -br link ; ip link show <tunnel-if> | grep -o 'mtu [0-9]*'"

caution:
  - level: state-changing
    text: >-
      Check with the owner before starting a `SHUTOFF` instance: one stopped on purpose looks the same.
  - level: know
    text: >-
      Until `nova-compute` is up and has synchronised, Nova still shows the instances `ACTIVE`, from before
      the reboot.

lesson: A reboot brings back what is written down. Nova does not restart guests by default.

runbook:
  path: recovery/after-a-compute-reboot-instances-shutoff-or-without-network.md
  status: lab-verified
  release: "2026.1"
  also:
    - networking/ovs-vswitchd-down-on-compute.md
    - networking/tunnel-mtu-mismatch-large-packets-dropped.md

scope: >-
  Written from the Nova 2026.1 configuration reference and Kolla-Ansible 2026.1, which does not set
  resume_guests_state_on_host_boot. Observed on one all-in-one lab: instances SHUTOFF without the option and
  resumed with it, and an external interface taken down without a reboot. Not observed: a container that does
  not come back, a lost MTU, a host address on br-ex, a multinode platform. A failed unit of a service
  disabled after deployment is not the cause: compare with globals.yml. ML2/OVN and DPDK are not covered.

linkedin:
  title: "After a compute reboot, the instances stay SHUTOFF. That is the default."
  caption: |
    A compute was rebooted for a kernel update. It is back, it answers SSH, nova-compute is up. And every instance on it is SHUTOFF.

    Nothing is broken. resume_guests_state_on_host_boot defaults to False: Nova does not start the guests that were running before a host reboot. Someone has to.

    The other surprises after a reboot come from the same place: what was never written down.
    → a service that did not come back, and if it is ovs-vswitchd, every instance is cut off
    → an external interface or bond that is DOWN
    → an MTU that was set by hand, months ago, and is back to the default

    One caution before starting everything again: an instance that was stopped on purpose before the reboot looks exactly like the others.
  hashtags: [OpenStack, Nova, Maintenance, CloudOperations, SRE, OzarLand]
