Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.6
Description
MDEV-515 will replace row-level locking with table-level locking in the special case that an INSERT is performed into an empty table or partition.
This causes a few replication testMajors to fail. We have adjusted some tests by inserting an extra record into the table, so that the normal row-level locking and undo logging will be used. But, some tests will require deeper studying to assess whether such a work-around is acceptable or whether some improvements to the replication logic will be needed.
Attachments
Issue Links
- is caused by
-
MDEV-515 innodb bulk insert
-
- Closed
-
- relates to
-
MDEV-30420 Assertion `mode_ == m_local' failed in int wsrep::client_state::append_key(const wsrep::key&)
-
- Closed
-
Here is the list of tests that failed for me locally (with non-debug galera-4 library):
galera.galera_bf_kill
galera.galera_ist_mariabackup
galera.galera_ist_mariabackup_innodb_flush_logs
galera.galera_ist_mysqldump
galera.galera_ist_rsync
galera.galera_mdev_15611
galera.galera_partition
galera.galera_pc_recovery
galera.galera_sst_mariabackup
galera.galera_sst_mariabackup_data_dir
galera.galera_sst_mysqldump_with_key
galera.galera_sst_rsync
galera.galera_sst_rsync2
galera.galera_sst_rsync_data_dir
galera.galera_var_innodb_disallow_writes
galera.MDEV-6860
galera.mysql-wsrep#33
galera_sr.galera_sr_dupkey_error
galera_sr.galera_sr_parallel_apply
galera_sr.GCF-1018B
galera_sr.GCF-572
galera_sr.GCF-623
galera_sr.mysql-wsrep-features#32
galera_sr.GCF-1008
I killed the test while it was executing galera_sr.GCF-1008 (I think it was hung).