Horizon Not Starting in IPv6 SingleStack OpenShift Cluster #478
Description
While trying to enable the deployment of Horizon in a Single Stack IPv6 cluster I came across the following error:
ALLOWED_HOSTS = [get_pod_ip(), "horizon-openstack.apps.lab._redacted_"]
File "/usr/lib/python3.9/site-packages/openstack_dashboard/local/local_settings.py", line 72, in get_pod_ip
s.connect(hostport)
OSError: [Errno 101] Network is unreachable
From what I can tell the IPv6 supported added in #315 only works within DualStack setups, but I am not able to verify this. The problem to me seems to stem from the attempt in the try catch block to use IPv4 first:
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
try:
s.connect(("{{ .horizonEndpointUrl }}", 80))
return s.getsockname()[0]
except socket.gaierror:
[...]As there would not be any IPv4 Network available it would raise the aforementioned OSError and never actually attempt to connect via IPv6. My crude attempt for a fix was to add a config map that overwrites the file with the get_pod_ip() function changed to the following:
def get_pod_ip():
import socket
import errno
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
hostport = ("{{ .horizonEndpointUrl }}", 443)
try:
s.connect(hostport)
return s.getsockname()[0]
except OSError as e:
if e.errno == errno.ENETUNREACH:
s.close()
s = socket.socket(socket.AF_INET6, socket.SOCK_DGRAM)
s.connect(hostport)
return "[{}]".format(s.getsockname()[0])
else:
raise
except socket.gaierror:
s.close()
s = socket.socket(socket.AF_INET6, socket.SOCK_DGRAM)
s.connect(hostport)
return "[{}]".format(s.getsockname()[0])
finally:
s.close()However, this resulted in other problems, as I assume the kolla config running on startup tries to change the file. If the lack of an IPv4 address is the actual problem here and the "fix" /workaround is acceptable I would submit a PR accordingly.