m-lab / mlab-ns / 1174 / 1
64%
master: 64%

DEFAULT BRANCH: master
Ran
Files 23
Run time 1s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 66.502% (+0.3%) from 66.211%
PYTHONPATH=$PYTHONPATH:$HOME/google-cloud-sdk/platform/google_appengine

push

travis-ci

web-flow 
Support per-experiment ReverseProxyProbability  (#199)

* Change get_reverse_proxy to accept an experiment name.

* Make pylint happy.

* Make pyflakes happy.

* Add log message.

* Change reverse_proxy.py to not use class methods.

Apparently memcache (and the underlying cPickle) doesn't like when the
class it's trying to serialize has been patched. The former code used the
get_by_key_name class method and was then patching it with MagicMock in
the test, and this made memcache.set() fail.

I reverted the code back to the previous behavior of reading and caching
*all* the entities in Datastore whenever the probability for any experiment
is requested, which makes sense to me since:

1. they all have the same expiration and
2. they are both requested very often when mlab-ns is running in production

So, the total number of calls to Datastore will be (slightly) smaller this way.

* Change tests to use the emulated Datastore/memcache instead of mocking.

* Make get_reverse_proxy return probability if memcache.set fails.

* Add whitespace between test setup and execution.

* Fix logic to avoid the extra call to memcache.get().

* Change get_reverse_proxy to return a default entity with the requested name.

Also, set probability for this default entity to 0.0.

* Update comments.

* s/records/record/

1076 of 1618 relevant lines covered (66.5%)

0.67 hits per line

Source Files on job 1174.1 (PYTHONPATH=$PYTHONPATH:$HOME/google-cloud-sdk/platform/google_appengine)