Fix reparent tests which fail with tracker on#6711
Fix reparent tests which fail with tracker on#6711rohit-nayak-ps wants to merge 6 commits intovitessio:masterfrom
Conversation
Signed-off-by: Rohit Nayak <rohit@planetscale.com>
…er on Signed-off-by: Rohit Nayak <rohit@planetscale.com>
…sing the issue Signed-off-by: Rohit Nayak <rohit@planetscale.com>
…hers with schema tracker on Signed-off-by: Rohit Nayak <rohit@planetscale.com>
Signed-off-by: Rohit Nayak <rohit@planetscale.com>
Signed-off-by: Rohit Nayak <rohit@planetscale.com>
|
@deepthi after skipping a couple of tests the test which used to fail earlier with the tracker on is working. The changing of dbuser to allprivs from dba doesn't really seem to matter though I have left it in since we want to do it anyway. There seems to be some leakage from the two tests I skipped. If I enable TestMasterToSpareStateChangeImpossible then TestReparentNoChoiceDownMaster fails with the tracker on. If TestReparentGracefulRangeBased is enabled then TestReparentNoChoiceDownMaster fails earlier. Is this dependence of tests on others by design? |
Any dependency is not by design. Not sure why enabling one test fails the next one because each test cleans up the tablets when it is done. cluster teardown should bring down mysql as well. |
|
Closed. Real fix is in #6726 |
Previously vreplication (and hence the schema tracker) was using dba privileges. This causes an issue during reparenting when the tracker continues to write to the schema_version table of the (readonly) replica due to the dba privileges causing errant gtids.
Signed-off-by: Rohit Nayak rohit@planetscale.com