Middleware is the backbone of the whole profiler facility. It adds an
encoded payload to each request that should be profiled with
osprofiler library. This library is embedded into other OpenStack
services where it tries to decode the message with a known set of keys
(so Horizon middleware should use one of these keys) and if
successful, sends a message to MongoDB (via osprofiler driver). Every
message has its own id, the base id (root message id) and parent
id. Using these 3 ids a tree of trace messages is assembled.
Actually, Horizon Django application uses 2 middleware classes:
ProfilerClientMIddleware and
ProfilerMiddleware. ProfilerClientMiddleware is used to enable Horizon
self-profiling (profiling from the UI): if a specific key is found in
cookies, then 2 standard osprofiler headers are added to a
request. These headers are processed by ProfilerMiddleware which
should always be the last middleware class in a Django config, since
it's defines `process_view` method which returns HttpResponse object
effectively terminating the middleware process_view chain. Assuming
that all API calls happen during rendering a view, Horizon sets a
tracepoint there which becomes a root node of the trace calls tree.
Implements-blueprint: openstack-profiler-at-developer-dashboard
Change-Id: Ib896676e304f2984c011bd1b610c86d1f24d46b9
Renamed bootstrap theme preview page to theme preview to clarify
that it is a preview of both bootstrap and horizon specific elements.
Change-Id: I88281b674ba5fdee735cc1091bb9767949eb8318
"openstack_dashboard.api.keystone: Unable to retrieve Domain: default"
incessant warning logging when switching Projects while being on the
Identity>Project panel.
Retrieving domain information is a Keystone admin URL operation. As a
pre-check, such operations would be Forbidden if the logon user does not
have an 'admin' role on the current project.
Since this is a common occurence, and can cause incessant warning
logging in the horizon logs, we recognize this condition and return the
user's domain information instead.
Signed-off-by: Kam Nasim <kam.nasim@windriver.com>
Closes-Bug: #1638662
Change-Id: Iadd5184a16a73da1da5a7230c89e996248f1eba7
Currently many panels are missing appropriate policy rules.
This patch adds policy rules to the admin images panel.
Change-Id: I8b25b420417387ef1d0ede1169243e30da343067
Partial-bug:#1534495
I don't really know what this was meant to do, but what it actually
does doesn't make any sense -- when checking a quota for project
'foo' we really don't care about how many instances are used
in the entire cloud, only in project 'foo'.
Until Liberty, the all_tenant flag was broken in the nova
api, which probably explains why this was harmless until
Liberty.
Change-Id: I9e12c10951e7518ab1c622a4daac962bad332da5
Closes-Bug: #1610693
