I think this will require more thought, as mentioned here: https://bugs.launchpad.net/percona-xtradb-cluster/+bug/1263991, in 5.6 the abillity to dynamically change the binlog_format at the session level has been removed. pt-table-checksum has previously relied on statement based replication to allow a crc to be calculated separately on the slave for comparison.
By simply commenting out the code and allowing the tool to use row based replication, it won't calculate a separate crc on the slave and therefore the checksum comparison will be redundant.
I think the fix will likely have to be within PXC, maybe by allowing the session binlog_format to be dynamically changed again, or maybe by utilizing the wsrep_consistency_check to force the sessions binlog_format to be statement based.
I think this will require more thought, as mentioned here: https:/ /bugs.launchpad .net/percona- xtradb- cluster/ +bug/1263991, in 5.6 the abillity to dynamically change the binlog_format at the session level has been removed. pt-table-checksum has previously relied on statement based replication to allow a crc to be calculated separately on the slave for comparison.
By simply commenting out the code and allowing the tool to use row based replication, it won't calculate a separate crc on the slave and therefore the checksum comparison will be redundant.
I think the fix will likely have to be within PXC, maybe by allowing the session binlog_format to be dynamically changed again, or maybe by utilizing the wsrep_consisten cy_check to force the sessions binlog_format to be statement based.