I have a simpler test case that may help. I ran pt-table-checksum against 5 empty tables that are persistently empty (i.e., no DML occurs against these tables). I repeatedly ran pt-table-checksum until a difference was identified by the tool. I will upload the PTDEBUG output shortly (note line 953 identifies a difference in table AO_563AEE_ACTIVITY_ENTITY, though this table is empty on both the master and slave.
I have a simpler test case that may help. I ran pt-table-checksum against 5 empty tables that are persistently empty (i.e., no DML occurs against these tables). I repeatedly ran pt-table-checksum until a difference was identified by the tool. I will upload the PTDEBUG output shortly (note line 953 identifies a difference in table AO_563AEE_ ACTIVITY_ ENTITY, though this table is empty on both the master and slave.
The following is relevant to this new test:
Tool version = pt-table-checksum 2.1.7 percona/ bin/pt- table-checksum --user=root --password=hidden --chunk- size-limit= 50.0 --replicate-check --recursion- method= DSN=D=percona, t=dsns_ ssodb03 --databases= vsp_jira_ current --tables= AO_563AEE_ ACTIVITY_ ENTITY, AO_563AEE_ ACTOR_ENTITY, AO_563AEE_ MEDIA_LINK_ ENTITY, AO_563AEE_ OBJECT_ ENTITY, AO_563AEE_ TARGET_ ENTITY --no-check- replication- filters --no-check- binlog- format --replicate= percona. checksums_ ssodb03 > bug2.txt 2>&1
Command line = /sso/sfw/