I was in the process of reviewing the patch
https://review.openstack.org/#/c/38275/ which fixes the bug I reported
https://bugs.launchpad.net/ceilometer/+bug/1193165 about errors in the
Ceilometer documentation due to the change from nose to testr. But the patch
was merged before I could submit my code review. I verified the config_dir
errors (mentioned by Gordon Chung) that occur when $ tox -e py27 -- api.v2 is
run.
I pulled down the latest changes in Ceilometer and rebuilt the testing environment with
$ tox -r -e py27,pep8
However, there's an error when it tries to install the package xattr. It
needs the dependency libffi-dev, which I installed using
$ sudo apt-get install libffi-dev
This appeared to fix the error. Later, I found out that this bug has been reported https://bugs.launchpad.net/openstack-ci/+bug/1203198
This is the sort error bug I was working on before https://bugs.launchpad.net/ceilometer/+bug/1193906. The patch I submitted, which was merged: https://review.openstack.org/#/c/36159/
Since some time had passed, I was curious if the connection errors I had previously had been resolved. Example of such a connection error: https://gist.github.com/terriyu/6004283
Ran tox 4 times on single test and didn't get any connection errors. This is the test I ran:
$ tox -e py27 -- tests.storage.test_impl_mongodb.RawSampleTest.test_get_samples_in_default_order
Out of curiosity, I also tried to see if I could get MongoDB to throw a sorting error if I force it to sort on a large database without an index. On 15 Jul 2013, I noted that, after recent changes to the Ceilometer codebase, I wasn't able to get MongoDB to throw the sorting error anymore.
The good news is that I did get it to throw the sorting error again and at database sizes that were consistent with my previous studies -- see the table in https://gist.github.com/terriyu/5954419
My test command was:
$ tox -e py27 -- tests.storage.test_impl_mongodb.RawSampleTest.test_get_samples_in_default_order
If I make the test database 12000 times larger, I don't get any sorting error. The test takes 256.869 sec to run, which is roughly consistent with my previous time of 240.318 sec. See tox output: https://gist.github.com/terriyu/6090520
If I make the test database 15000 times larger, I do get a sorting error and it takes 320.696 sec to run, which is roughly consistent with my previous time of 291.161 sec. See tox output: https://gist.github.com/terriyu/6090521
jd thinks that my test was fixed because his patch "Disable mongod prealloc, wait for it to start" was merged: https://review.openstack.org/#/c/37105/
The patch forces execution of the tests to wait until MongoDB is ready. jd's commit message:
No prealloc should make mongod starts faster, and the while loop will wait it is listening before running the tests.