Both are just different applications for running tests, they have their own rules for discovering tests and their own way of executing them.

The difference you are seeing in the number of tests executed using testr vs nosetests is basically based on the way each one discovers or identifies what is a test. To explain this easier let's use an example.

Let's say we want to run all tests in this package: tempest/api/compute/flavors

First using nosetests:

ad_cjmarti2@cas-devstack:/opt/stack/tempest$ nosetests -v tempest.api.compute.flavors
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.create_test_server ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.create_test_server_group ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_get_flavor[id-1f12046b-753d-40d2-abb6-d8eb8b30cb2f,smoke] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors[id-e36c0eaa-dff5-4082-ad1f-3f9a80aa3f59,smoke] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_filter_by_min_disk[id-3df2743e-3034-4e57-a4cb-b6527f6eac79] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_filter_by_min_ram[id-09fe7509-b4ee-4b34-bf8b-39532dc47292] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_limit_results[id-b26f6327-2886-467a-82be-cef7a27709cb] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_using_marker[id-6db2f0c0-ddee-4162-9c84-0703d3dd1107] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_filter_by_min_disk[id-10645a4d-96f5-443f-831b-730711e11dd4] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_filter_by_min_ram[id-935cf550-e7c8-4da6-8002-00f92d5edfaa] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_limit_results[id-8d7691b3-6ed4-411a-abc9-2839a765adab] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_using_marker[id-e800f879-9828-4bd0-8eae-4f17189951fb] ... ok
tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_with_detail[id-6e85fde4-b3cd-4137-ab72-ed5f418e8c24] ... ok

----------------------------------------------------------------------
Ran 13 tests in 16.461s

OK

And now using testr:

ad_cjmarti2@cas-devstack:/opt/stack/tempest$ testr run tempest.api.compute.flavors
Ran 11 tests in 2.366s (+0.587s)
PASSED (id=9)

To make the output of testr to look similar to that of nosetests let's us the subunit output and pipe it to the subunit formatter:

 ad_cjmarti2@cas-devstack:/opt/stack/tempest$ testr run --subunit tempest.api.compute.flavors | subunit-trace -n -f
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_get_flavor [0.549721s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors [0.248651s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_filter_by_min_disk [0.165985s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_filter_by_min_ram [0.156681s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_limit_results [0.066102s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_detailed_using_marker [0.164474s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_filter_by_min_disk [0.195904s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_filter_by_min_ram [0.176698s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_limit_results [0.076589s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_using_marker [0.191877s] ... ok
{0} tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.test_list_flavors_with_detail [0.203010s] ... ok

======
Totals
======
Ran: 11 tests in 9.0000 sec.
 - Passed: 11
 - Skipped: 0
 - Expected Fail: 0
 - Unexpected Success: 0
 - Failed: 0
Sum of execute time for each test: 2.1957 sec.

==============
Worker Balance
==============
 - Worker 0 (11 tests) => 0:00:02.201494

Viewing both executions we can clearly see that nosetests ran 13 tests while testr ran 11 tests. Taking a closer look at the test results we can see that these two are the tests that were ran by nosetests and were not run by testr:

tempest.api.compute.flavors.test_flavors.FlavorsV2TestJSON.create_test_server ... ok
tempest.api.compute ...