horizon hangs for 1 minute if there are no compute nodes
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Dashboard (Horizon) |
New
|
Undecided
|
Unassigned |
Bug Description
The problem is essentially, that Horizon is unresponsive when nova-network is not available. The behavior in Folsom is different from Essex, where it worked just fine.
Setup:
- CentOS 6.3
- EPEL packages for Folsom
- Controller node without compute nodes
On Essex:
- works fine
On Folsom:
- Horizon front page loads fine
- when you submit username/password, it hangs for one minute and then finally loads. but there is a red HTML popup saying "Error: Unable to retrieve quota information."
- as long as you add a new compute node with nova-network, it starts to work just fine
Logs:
unable to retrieve service catalog with token
Traceback (most recent call last):
File "/usr/lib/
endpoint_
File "/usr/lib/
raise exceptions.
EndpointNotFound: Endpoint not found.
[[31;1mRecoverable error: n/a (HTTP 504)[[0m
Traceback (most recent call last):
File "/usr/lib/
self.quotas = api.nova.
File "/usr/lib/
value = self.func(*args)
File "/usr/lib/
floating_ips = tenant_
File "/usr/lib/
return novaclient(
File "/usr/lib/
return self._list(
File "/usr/lib/
_resp, body = self.api.
File "/usr/lib/
return self._cs_
File "/usr/lib/
**kwargs)
File "/usr/lib/
resp, body = self.request(url, method, **kwargs)
File "/usr/lib/
raise exceptions.
ClientException: n/a (HTTP 504)
This is essentially a duplicate of https:/ /bugs.launchpad .net/horizon/ +bug/961761 since it's the same timeout issue. I still have no fix in mind for what to do when nova network is down. It's a required service for Horizon and is documented as such.