Skip to content

openstack-glance-image-download-failed.md

Latest commit

 

History

History
145 lines (109 loc) · 5.08 KB

openstack-glance-image-download-failed.md

File metadata and controls

145 lines (109 loc) · 5.08 KB
Raw
title OpenStack Glance Image Download Failed
slug openstack-glance-image-download-failed
technologies
openstack
glance
severity medium
tags
openstack
glance
image
download
store
production
related
openstack-glance-image-stuck-in-saving
last_reviewed 2026-06-27

OpenStack Glance Image Download Failed

Error Message

$ openstack image save --file ubuntu-2204.qcow2 7f0a3c1e-9b2d-4a11-8c3e-2f1a9d0b7e44
HTTP 404 Not Found: Image 7f0a3c1e-9b2d-4a11-8c3e-2f1a9d0b7e44 not found.
The image data is not available.

glance-api[1842]: ERROR glance.api.v2.image_data [req-...] Image data not found for image \
7f0a3c1e-9b2d-4a11-8c3e-2f1a9d0b7e44: glance_store.exceptions.NotFound: Image not found in store

Description

A request to retrieve image bits — openstack image save, a Nova compute pulling the base image, or a GET /v2/images/{id}/file — fails even though the image appears in the catalog. The image metadata exists in the Glance database, but the data cannot be served: either the object is missing from the backing store, the store is unreachable, or the image is in a state (queued, saving, deleted) that has no downloadable data.

Technologies

  • openstack (glance-api, glance_store driver, Nova compute as a consumer)

Severity

medium — instances depending on this image cannot boot and the image cannot be exported. Other images and running instances are unaffected.

Common Causes

  1. The store object was deleted out-of-band (manual rbd rm, file removed, Swift object purged) while the Glance metadata row survived.
  2. The backing store is unreachable — Ceph auth/connectivity failure, NFS mount missing, Swift endpoint down.
  3. The image never finished uploading — it is queued or stuck in saving, so there is genuinely no data to download.
  4. A multi-store migration left stores/locations metadata pointing at a store that no longer holds the object.
  5. Corrupt or truncated data fails the checksum verification during transfer.

Root Cause Analysis

On a download, glance-api looks up the image's locations (or default store), asks the glance_store driver to open the object, and streams it back. If the driver cannot find or read the object, it raises NotFound/StoreError, which the API surfaces as a 404 or 500. The key distinction is metadata-vs-data: the catalog row tells you the image should exist, while the store tells you whether the bytes are actually present and reachable. Mismatch between the two is the root cause.

Diagnostic Commands

# Status must be 'active' to have downloadable data; check size and locations
openstack image show <image-id> -c status -c size -c stores -f value

# Locations as the admin sees them (needs show_multiple_locations / admin)
openstack image show <image-id> -f json | python3 -m json.tool | grep -i location

# Does the object physically exist? (Ceph RBD example)
rbd -p images ls | grep <image-id>

# File store presence and size
ls -lh /var/lib/glance/images/<image-id>

# glance-api errors at download time
journalctl -u devstack@g-api --since "15 min ago" | grep -iE "not found|store|error"

Expected Results

status   active
size     2361393152
stores   ceph

# But the object is missing from the store -> data unavailable:
$ rbd -p images ls | grep <image-id>
(no output)

# Healthy comparison: the object exists and size matches the metadata.

Resolution

  1. If the image is queued/saving, there is no data — re-upload it (see the "stuck in saving" runbook) rather than chasing the store.
  2. If the store is unreachable, restore connectivity/auth (Ceph keyring, rbd_store_pool, NFS mount, Swift endpoint) and retry the download.
  3. If the object was deleted but metadata remains, the data is unrecoverable — delete the dangling record and re-import the image: 
    openstack image delete <image-id>
openstack image create --disk-format qcow2 --container-format bare \
  --file ubuntu-2204.qcow2 ubuntu-2204
  4. If locations point at the wrong store after a migration, fix the location metadata or copy the image into the active store: 
    openstack image import --method copy-image --stores ceph <image-id>

Validation

openstack image save --file /tmp/test.img <image-id>
ls -lh /tmp/test.img   # size should match 'openstack image show -c size'

Prevention

  • Never delete store objects directly — always go through the Glance API so metadata and data stay consistent.
  • Enable and monitor a periodic consistency check between Glance metadata and store contents.
  • Use checksums (os_hash_value) to detect truncation/corruption early.
  • Monitor store reachability and auth (Ceph health, NFS mounts, Swift endpoints).

Related Errors

References

Tags

openstack · glance · image · download · store · production