Skip to content

openstack-horizon-internal-server-error-500.md

Latest commit

 

History

History
155 lines (117 loc) · 5.6 KB

openstack-horizon-internal-server-error-500.md

File metadata and controls

155 lines (117 loc) · 5.6 KB
Raw
title OpenStack Horizon Internal Server Error (HTTP 500)
slug openstack-horizon-internal-server-error-500
technologies
openstack
horizon
severity high
tags
openstack
horizon
dashboard
500
django
production
related
openstack-keystone-401-unauthorized
openstack-keystone-could-not-find-service
last_reviewed 2026-06-27

OpenStack Horizon Internal Server Error (HTTP 500)

Error Message

Internal Server Error
The server encountered an internal error or misconfiguration and was unable to
complete your request.

[wsgi:error] [pid 22418] Internal Server Error: /horizon/auth/login/
ImportError: No module named 'openstack_dashboard.local.local_settings'
DisallowedHost at /  Invalid HTTP_HOST header: 'dashboard.example.com'. \
You may need to add 'dashboard.example.com' to ALLOWED_HOSTS.

Description

Horizon (a Django app served under Apache/mod_wsgi or uWSGI) returned an HTTP 500 instead of rendering a page. With DEBUG = False in production the browser shows only a generic error; the actual Python traceback is written to the web server error log. The cause is almost always a configuration or environment problem in the dashboard host — a bad local_settings, a missing static/cache backend, an unreachable Keystone, or a Django setting like ALLOWED_HOSTS.

Technologies

  • openstack (horizon / openstack_dashboard, Django, Apache mod_wsgi or uWSGI)

Severity

high — the web UI is unusable for all tenants and operators. The API and CLI remain functional, so it is a UI outage rather than a control-plane outage.

Common Causes

  1. ALLOWED_HOSTS does not include the hostname/IP the browser used (DisallowedHost), so Django 500s on every request.
  2. local_settings.py is missing, has a syntax/import error, or references an undefined OPENSTACK_HOST/OPENSTACK_KEYSTONE_URL.
  3. The session/cache backend is unreachable — Memcached down, or CACHES/SESSION_ENGINE misconfigured.
  4. Static files were never collected/compressed, or their directory is not writable, breaking page render.
  5. Keystone is unreachable from the Horizon host, so login/catalog calls raise.
  6. Wrong file permissions or SELinux/AppArmor denial on the WSGI process.

Root Cause Analysis

Apache/mod_wsgi loads the Django openstack_dashboard application using the settings module. If settings import fails, or a request violates a Django guard (ALLOWED_HOSTS), or a downstream dependency (Memcached, Keystone) raises during view processing, Django returns 500 and logs the traceback. Because DEBUG is off in production, the only way to find the true cause is the web server error log. The traceback's final exception line names the precise subsystem at fault.

Diagnostic Commands

# The real traceback lives here (Apache); pick your distro's path
journalctl -u apache2 --since "15 min ago" | grep -iE "error|traceback|horizon"
tail -n 100 /var/log/apache2/error.log        # or /var/log/httpd/error_log

# uWSGI deployments
journalctl -u horizon --since "15 min ago"

# Is Memcached (session/cache backend) reachable?
echo stats | timeout 2 nc 127.0.0.1 11211 | head

# Can the Horizon host reach Keystone?
curl -s -o /dev/null -w "%{http_code}\n" http://<keystone-host>:5000/v3

# Confirm settings load cleanly (read-only check)
python3 -c "import openstack_dashboard.settings" 2>&1 | head

# Verify the configured ALLOWED_HOSTS and OPENSTACK_HOST
grep -nE "ALLOWED_HOSTS|OPENSTACK_HOST|OPENSTACK_KEYSTONE_URL" \
  /etc/openstack-dashboard/local_settings.py

Expected Results

# Smoking-gun traceback lines:
DisallowedHost: Invalid HTTP_HOST header: 'dashboard.example.com'.
  -> ALLOWED_HOSTS is too narrow.

pylibmc.ConnectionError: error 3 from memcached_get
  -> Memcached / cache backend is down.

ConnectionError: HTTPConnectionPool(host='keystone', port=5000) Max retries exceeded
  -> Keystone unreachable from Horizon.

# Healthy: settings import returns nothing, nc to memcached prints stats,
# and the Keystone curl returns 200/300.

Resolution

  1. For DisallowedHost, add the hostname(s) and restart the web server: 
    ALLOWED_HOSTS = ['dashboard.example.com', 'localhost', '127.0.0.1']
    systemctl restart apache2     # or: systemctl restart horizon
  2. Fix local_settings.py syntax/imports; ensure OPENSTACK_HOST and OPENSTACK_KEYSTONE_URL point at a reachable Keystone.
  3. Restore the cache/session backend (start Memcached) or correct CACHES.
  4. Recollect and compress static assets, then fix ownership: 
    python3 manage.py collectstatic --noinput
python3 manage.py compress --force
  5. Resolve permission/SELinux denials on the WSGI user.

Validation

curl -s -o /dev/null -w "%{http_code}\n" http://<horizon-host>/auth/login/
# Expect: 200 (login page renders), not 500

Prevention

  • Keep DEBUG = False in production but ship error logging to a central place so 500 tracebacks are searchable.
  • Manage local_settings.py with config management and validate import in CI.
  • Monitor Memcached and Keystone reachability from each Horizon node.
  • Run collectstatic/compress as part of deployment, not by hand.

Related Errors

References

Tags

openstack · horizon · dashboard · 500 · django · production