Assuming we're single digit failures combined between the two, I think a single test failure email would be manageable.
On Tue, Aug 16, 2016 at 2:46 AM, Joel Knighton <joel.knigh...@datastax.com> wrote: > =================================================== > testall: 1 failure > org.apache.cassandra.io.compress > .CompressedRandomAccessReaderTest.testDataCorruptionDetection > New flaky failure. I've opened CASSANDRA-12465 and assigned > myself. > > =================================================== > dtest: All passed! > > =================================================== > novnode: All passed! > > =================================================== > upgrade: 3 failures > upgrade_tests.cql_tests > .TestCQLNodes2RF1_Upgrade_current_3_0_x_To_indev_3_x > .map_keys_indexing_test > CASSANDRA-12192. Tyler Hobbs as assignee. They have identified > the cause and proposed a test fix. They are also investigating a C* > change here to improve robustness. > upgrade_tests.cql_tests > .TestCQLNodes3RF3_Upgrade_current_3_x_To_indev_3_x > .map_keys_indexing_test > Same as above. > upgrade_tests.paging_test > .TestPagingDataNodes2RF1_Upgrade_current_2_2_x_To_indev_3_x > .static_columns_paging_test > Potentially CASSANDRA-11195, which is open with no clear > progress. I'll follow up with those on that issue tomorrow and see if > they agree that this is the same problem. > > =================================================== > Overall, the testing situation continues to look better. The massive > upgrade failures seem to have subsided, so we can continue to target > individual failures. > > Since the 3.9 tests are getting to a manageable level, we should focus > on managing test failures on trunk as well. I will soon start tracking > these failures, as well as failures on the large dtest runs, which consist > of tests that have been segmented off due to increased cluster size. > > On months that we're maintaining a 3.x bugfix branch as well as trunk, > is there any preference toward a single email or a separate email for > each branch? Any other feedback is welcome, as always. >
