220433719e
The DatabaseConfigurationsTests.test_values_tab_discard_action test is unstable, failing and succeeding on subsequent runs. The root cause of the instability is the django local memory cache used by the test cases. This cache has defaults that will cull 33% of the cache entries when it reaches its max capacity of 300. During the test runs the cache becomes full and the culling will randomly choose to cull the cached configuratoin being used by the test case about 33% of the time. This leads to test case failure. In a production environment the cache used is memcached which has different behavior. If the cache entry were still culled, for the "discard changes" option, that would be equivalent to a discard since the configuration manager would re-fetch. In the test case flow it breaks because the 'changed values' fetch gets a cache miss. The fix is to set the cache configuration for the tests to use a higher max_entries value and change the cull frequency to only cull 5% of the entries every cull. Change-Id: I44bcf14c4ed17505de94b44b01ffcccd818399b7 |
||
---|---|---|
devstack | ||
releasenotes | ||
tools | ||
trove_dashboard | ||
.gitignore | ||
.gitreview | ||
babel-django.cfg | ||
babel-djangojs.cfg | ||
CONTRIBUTING.rst | ||
HACKING.rst | ||
LICENSE | ||
manage.py | ||
MANIFEST.in | ||
README.rst | ||
requirements.txt | ||
run_tests.sh | ||
setup.cfg | ||
setup.py | ||
test-requirements.txt | ||
tox.ini |
OpenStack Dashboard plugin for Trove project
How to use with Horizon on server:
Use pip to install the package on the server running Horizon. Then either copy or link the files in trove_dashboard/enabled to openstack_dashboard/local/enabled. This step will cause the Horizon service to pick up the trove plugin when it starts.
How to use with devstack:
Add the following to your devstack local.conf
file:
enable_plugin trove-dashboard git://git.openstack.org/openstack/trove-dashboard
To run unit tests:
./run_tests.sh
Editing Code
Apache
Make a change to trove-dashboard then goto to the horizon directory and compress the code with django and then restart apache.:
# rsync code to /opt/stack/trove-dashboard
# copy or link files from trove-dashboard/enabled/* to horizon/openstack_dashboard/local/enabled/
cd /opt/stack/horizon
python manage.py compress
python manage.py collectstatic --noinput
sudo service apache2 restart
Django
You can also speed up development time using the django test server instead of apache.:
/opt/stack/horizon/run_tests.sh --runserver
If you set COMPRESS_ENABLED and COMPRESS_OFFLINE to False in local_settings.py that allows you to bypass the compress and collectstatic as well.
Settings
The use of a cross-process cache such as Memcached is required.
Install Memcached itself and a Memcached binding such as python-memcached.
For a single horizon instance use the CACHES setting like the example below.
- CACHES = {
-
- 'default': {
-
'BACKEND': 'django.core.cache.backends.memcached.MemcachedCache', 'LOCATION': '127.0.0.1:11211',
},
}
For multiple horizon instances behind a load balancer configure each instance to use the same cache like the example below.
- CACHES = {
-
- 'default': {
-
'BACKEND': 'django.core.cache.backends.memcached.MemcachedCache', 'LOCATION': [u'10.2.100.133:11211', u'10.2.100.134:11211'']
},
}
NOTE:
As of the Mitaka release, the dashboard for trove is now maintained outside of the Horizon codebase, in this repository.
Links:
Trove project: https://git.openstack.org/openstack/trove
Trove at wiki.openstack.org: https://wiki.openstack.org/wiki/Trove
Launchpad project: https://launchpad.net/trove-dashboard