This file contains the outcome of 3 mysqlslap tests that are supposed to create 369 rows in the table t1. I have patched the file sql/rpl_gtid.cc to print some extra info. The test #1 is without Replication Filter B2C, all 369 records are eventually applied. Test #2 and #3 are with the same Replication filter aimed to isntruct C to not apply any transactions coming from B (using the non-existing schema name dummy), and a variable number of records are not applied. TEST#1: bin/mysqlslap --host=127.0.0.1 --port=30034 --auto-generate-sql --auto-generate-sql-execute-number=50 --concurrency=10 --number-char-cols=5 --number-int-cols=8 -vv --no-drop --create-schema=slap_530 [root@tossanc data]# egrep "A2C|B2C" error.20034.log | awk '{print $5" "$6" "$7" "$8" "$9" "$10" "$11" "$12}' | sort | uniq -c 134 'A2C': RES_0: This sequence number is already applied, 237 'A2C': RES_1.A: The domain became free, grab it 134 'A2C': Someone else is currently processing this GTID 237 'B2C': RES_0: This sequence number is already applied, 134 'B2C': RES_1.A: The domain became free, grab it 237 'B2C': Someone else is currently processing this GTID MariaDB [(none)]> select count(*) from slap_530.t1; +----------+ | count(*) | +----------+ | 369 | +----------+ 1 row in set (0.01 sec) TEST#2: MariaDB [(none)]> stop slave 'B2C'; MariaDB [(none)]> set global B2C.Replicate_Wild_Do_Table='dummy%.%'; ariaDB [(none)]> start slave 'B2C'; [root@tossanc data]# cat /dev/null > error.20034.log MariaDB [(none)]> flush logs; bin/mysqlslap --host=127.0.0.1 --port=30034 --auto-generate-sql --auto-generate-sql-execute-number=50 --concurrency=10 --number-char-cols=5 --number-int-cols=8 -vv --no-drop --create-schema=slap_540 [root@tossanc data]# egrep "A2C|B2C" error.20034.log | awk '{print $5" "$6" "$7" "$8" "$9" "$10" "$11" "$12}' | sort | uniq -c 3 'A2C': RES_0: This sequence number is already applied, 368 'A2C': RES_1.A: The domain became free, grab it 3 'A2C': Someone else is currently processing this GTID 368 'B2C': RES_0: This sequence number is already applied, 3 'B2C': RES_1.A: The domain became free, grab it 229 'B2C': Someone else is currently processing this GTID MariaDB [(none)]> select count(*) from slap_540.t1; +----------+ | count(*) | +----------+ | 366 | +----------+ 1 row in set (0.00 sec) TEST#3: bin/mysqlslap --host=127.0.0.1 --port=30034 --auto-generate-sql --auto-generate-sql-execute-number=50 --concurrency=10 --number-char-cols=5 --number-int-cols=8 -vv --no-drop --create-schema=slap_550 [root@tossanc data]# egrep "A2C|B2C" error.20034.log | awk '{print $5" "$6" "$7" "$8" "$9" "$10" "$11" "$12}' | sort | uniq -c 20 'A2C': RES_0: This sequence number is already applied, 351 'A2C': RES_1.A: The domain became free, grab it 20 'A2C': Someone else is currently processing this GTID 351 'B2C': RES_0: This sequence number is already applied, 20 'B2C': RES_1.A: The domain became free, grab it 187 'B2C': Someone else is currently processing this GTID MariaDB [(none)]> select count(*) from slap_550.t1; +----------+ | count(*) | +----------+ | 349 | +----------+ From my understanding the GTID_IGNORE_DUPLICATES ignores the replication filters, so it takes for granted that all GTID coming from any replication channel are applied so that the ones that arrive and are elaborated first via the 'filtered' channel are considered duplicates, but in reality they won't be applied at all. In the ouput above the number NN in: NN 'A2C': RES_0: This sequence number is already applied, Is the number of skipped records.