| 1 |
############################################################################### |
| 2 |
# Description : |
| 3 |
# Test that using server-id-bits and composite AnyValues works in that. |
| 4 |
# 1) Composite AnyValues are correctly split into ServerId and UserData |
| 5 |
# by MySQLDs listening to NdbApi events. |
| 6 |
# 2) Composite AnyValues are placed into the ServerId of events in the |
| 7 |
# Binary logs of attached binlogging MySQLDs |
| 8 |
# 3) Composite ServerIds are correctly split into ServerId by the replication |
| 9 |
# framework on the slave-side for the purposes of : |
| 10 |
# a) Loop detection |
| 11 |
# b) ndb_apply_status updates |
| 12 |
# 4) Composite ServerIds are placed into the composite AnyValues of row |
| 13 |
# operations by the slave |
| 14 |
# 5) Binlogging MySQLDs on the slave correctly identify replicated composite |
| 15 |
# AnyValues as being replicated. |
| 16 |
# |
| 17 |
# This test is a re-run of rpl_ndb_slave_lsu with different configuration : |
| 18 |
# a) server-id-bits is set to 7, implying that : |
| 19 |
# - Only the bottom 7 bits of event ServerIds written to the Binlog are |
| 20 |
# set to the ServerId. The other bits come from the event AnyValue |
| 21 |
# received by the Binlog injector. |
| 22 |
# - Only the bottom 7 bits of ServerIds received in Binlogs are used as |
| 23 |
# the ServerId in the Slave, the other bits are ignored, but are placed |
| 24 |
# in the AnyValue of resulting NdbApi operations. |
| 25 |
# b) NDB_TEST_ANYVALUE_USERDATA is set to Y |
| 26 |
# - This causes MySQLD to set the UserData part of the AnyValue (in this |
| 27 |
# case bits 8-30) to '1'. |
| 28 |
# This results in AnyValues and replicated ServerIds having an 'offset' |
| 29 |
# of 2^24 << 7 == 2147483520 |
| 30 |
# This offset can be seen when comparing the SHOW BINLOG EVENTS ServerId |
| 31 |
# values (which use the server-id-bits to only show the server-id, to |
| 32 |
# output from the mysqlbinlog tool when it passed no server-id-bits |
| 33 |
# option, and shows the full serverid. |
| 34 |
# It can be seen that events originating on the master are given a |
| 35 |
# composite server id which is available at the slave as it is written |
| 36 |
# into the binlog of the slave running with --log-replica-updates on. |
| 37 |
# |
| 38 |
# NOTE! The main difference between this test and ndb_rpl_slave_lsu is that |
| 39 |
# the output "SHOW BINLOG EVENTS" will show the actual server_id while the |
| 40 |
# the output from "mysqlbinlog" will show the full anyvalue (which contains |
| 41 |
# both the server_id as well as other flags transported between servers |
| 42 |
# in the anyvalue). |
| 43 |
# |
| 44 |
############################################################################### |
| 45 |
--source include/have_debug.inc |
| 46 |
--source suite/ndb_rpl/t/ndb_rpl_slave_lsu.test |