################################################################################ # # Bug#25674926 VIEW CHANGE LOG EVENTS GENERATE NEW GTIDS # # Problem: When there is replication (async) between two groups, # group nodes are confused when it receives view_change_log_event # that is generated on the other group and the event is making # infinite round trips between the groups in case if the async # replication is circular replication. # # Scenario: Verify Group Replication behavior when we establish a # circular replication connection from a server in a group # to another server in the other group. # Steps to reproduce: # 0. The test requires six servers: M1, M2, M3, M4, M5 and M6. # 1. Bootstrap start a group, GR1, on M1. Start GR on M2 and add it GR1. # 2. Bootstrap start a group, GR2, on M3. Start GR on M4 and add it GR2. # 3. Setup a asynchronous replication connection from M3(GR2) into M1(GR1) # i.e. M3 -> M1. # 4. Setup a asynchronous replication connection from M1(GR1) into M3(GR2) # i.e. M1 -> M3. # 5. Test view change event issue by adding/dropping nodes from the groups. # Add node M5 to group GR1, and add node M6 to group GR2 and immediately # drop the node from group (GR2). (This step is just to generate more # VIEW_EVENTs) # 6. Execute some transactions on M3. # 7. Wait until transactions executed on M3 are applied on all nodes of GR1 # and GR2. # 8. Execute some transactions on M1. # 9. Wait until transactions executed on M1 are applied on all nodes of GR1 # and GR2. # 10. Validate data on all nodes (M1, M2, M3, M4 and M5). # 11. Check that GTID_EXECUTED on all servers are as expected. # 12. Clean data. # 13. Stop circular replication between groups. # 14. Stop Group Replication on all nodes (M1, M2, M3, M4 and M5). ################################################################################ --source ../include/gr_replicate_group_to_group.inc