| title | OpenStack Nova Build of Instance Aborted | ||||||
|---|---|---|---|---|---|---|---|
| slug | openstack-nova-build-of-instance-aborted | ||||||
| technologies |
|
||||||
| severity | high | ||||||
| tags |
|
||||||
| related |
|
||||||
| last_reviewed | 2026-06-27 |
nova-compute[1843]: ERROR nova.compute.manager [req-9a4f...] [instance: 4d8e...] \
Build of instance 4d8e... aborted: Volume 6c1b... did not finish being created \
even after we waited 187 seconds or 61 attempts. And its status is downloading.
BuildAbortException: Build of instance 4d8e... aborted: Failed to allocate the \
network(s), not rescheduling.
BuildAbortException is raised by nova-compute when a build hits a condition
that there is no point retrying — so Nova deliberately stops and does not
reschedule to another host. This distinguishes it from a transient spawn failure
that triggers a retry. The most common triggers are a dependent resource that
never became ready (a Cinder volume stuck in downloading/creating, or a
Neutron network that could not be allocated). The instance goes to ERROR.
- openstack (nova-compute, nova-conductor, cinder, neutron)
high — the instance build fails permanently with no automatic retry; if the dependency (Cinder/Neutron) is broken, all similar builds will abort the same way.
- Boot-from-volume where the Cinder volume never reaches
availablewithin the timeout — backend slow, image conversion stuck, orcinder-volumedown. Failed to allocate the network(s)— Neutron could not create or bind ports.- Block Device Mapping is invalid (bad volume/snapshot/image reference).
- The maximum number of build retries was exhausted, so Nova aborts.
- Port quota or IP exhaustion in the requested network.
During _build_resources, nova-compute waits for block devices and networks
before calling spawn(). It polls Cinder until the volume is available (up to
block_device_allocate_retries × interval) and waits on Neutron for port
allocation. If those waits time out or return a hard error, Nova raises
BuildAbortException rather than rescheduling, because retrying on another host
would not fix an unavailable volume or a network it cannot allocate. The fault
text names the exact unmet dependency and how long Nova waited.
# The abort reason recorded on the instance
openstack server show <instance> -c fault -f value
# If boot-from-volume: is the volume stuck?
openstack volume show <volume-id> -c status -f value
# cinder-volume service health
openstack volume service list
# Neutron port allocation for the instance
openstack port list --device-id <instance-id>
# nova-compute build log
journalctl -u devstack@n-cpu --since "15 min ago" | grep -iE "aborted|build of instance"fault | {'message': 'Build of instance ... aborted: Volume ... did not finish
being created even after we waited 187 seconds...'}
# volume show
status | downloading # stuck — never reached 'available'
A volume stuck in downloading/creating, an empty port list, or a
Failed to allocate the network(s) line pinpoints the unmet dependency.
- Stuck volume: fix the Cinder backend (see the Cinder error in Related),
then recreate. Increase
block_device_allocate_retriesif the backend is merely slow under load. - Network allocation: resolve the Neutron failure (port binding, quota, IP exhaustion) before retrying.
- Bad BDM: correct the volume/snapshot/image references in the request.
- Delete the
ERRORinstance and re-create; aborted builds are not retried.
openstack server create --flavor m1.small --image cirros \
--boot-from-volume 10 --network private build-test
openstack server show build-test -c status -f value # Expect: ACTIVE- Monitor Cinder backend latency and
cinder-volumehealth. - Alert on Neutron port and IP quota approaching exhaustion.
- Validate volume/image references before bulk launches.
- Keep
block_device_allocate_retriesrealistic for your backend speed.
- OpenStack Nova Instance Failed to Spawn
- OpenStack Cinder Volume in Error State
- OpenStack Neutron Port Binding Failed
openstack · nova · build · conductor · rescheduling · production