1
0
mirror of https://github.com/MariaDB/server.git synced 2025-07-27 18:02:13 +03:00

MDEV-9501: rpl.rpl_binlog_index, rpl.rpl_gtid_crash, rpl.rpl_stm_multi_query fail sporadically in buildbot with Master command COM_REGISTER_SLAVE failed

Analysis:
========
Slave server will send COM_REGISTER_SLAVE command at the time of establishing
a connection to master. If master is down, then the command will fail and
COM_REGISTER_SLAVE failed warning is reported.

'rpl_binlog_index.test' shutsdown the master and it relocates binary logs to a
new location and attempts to start master by pointing 'log-bin' to new
location. During this process the slave threads are active. IO thread actively
checks for the presence of master when it finds that the connection is lost it
attempts a reconnect, as master is down COM_REGISTER_SLAVE command fails.

As part of fix, stop the slave threads and then shutdown the master and do the
binlog relocation. Once master is restarted start the slave threads and sync
them with the master. In test binary logs and index files on master are
relocated to /tmpdir but during master restart only --log-bin option is
provided, this is incorrect. Even --log-bin-index also should be pointed to
/tmpdir otherwise upon master server restart two index files will be created.
One master-bin.index in /tmpdir and a new master-bin.index as per log_basename
in datadir. Due to this slave will fail to connect to master.

'rpl_gtid_crash.test' tests following scenario "crashing master, causing slave
IO thread to reconnect while SQL thread is running". When IO thread tries to
connect to crashed master on slow platforms COM_REGISTER_SLAVE command fails.
This is expected hence the warning should be added to suppression list.
This commit is contained in:
Sujatha
2020-06-30 16:42:44 +05:30
parent 420c4dcc7e
commit a8f6bbb7a8
4 changed files with 23 additions and 5 deletions

View File

@ -3,18 +3,22 @@ include/master-slave.inc
CREATE TABLE t1 (a INT);
FLUSH BINARY LOGS;
INSERT INTO t1 VALUES (1);
include/stop_slave.inc
# Shutdown master
include/rpl_stop_server.inc [server_number=1]
# Move the master binlog files and the index file to a new place
# Restart master with log-bin option set to the new path
# Master has restarted successfully
include/start_slave.inc
# Create the master-bin.index file with the old format
include/stop_slave.inc
# Shutdown master
include/rpl_stop_server.inc [server_number=1]
# Move back the master binlog files
# Remove the unneeded master-bin.index file
# Restart master with log-bin option set to default
# Master has restarted successfully
include/start_slave.inc
# stop slave
include/stop_slave.inc
include/rpl_stop_server.inc [server_number=2]