database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-09-02 09:41:40 +03:00
Code Activity

MDEV-7888, MDEV-7929: Parallel replication hangs sometimes on ANALYZE TABLE or DDL

Browse Source 
The hangs occur when the group_commit_orderer object is freed before the last
mark_start_commit() call on it - this loses the wakeup to other waiting worker
threads, causing them to hang until killed manually.

The object was freed because wakeup_subsequent_commits() was called two early
in two places. For MDEV-7888, during ANALYZE TABLE, and for MDEV-7929 during
record_gtid() after processing a DDL event. The group_commit_orderer object
can be freed when its last transaction has called wait_for_prior_commit().

Fix by implementing a suspend/resume mechanism for wakeup_subsequent_commits()
that can be used in places where a transaction is committed without this being
the commit of the actual replication event group.

Also add a protection mechanism (that asserts in debug builds) which can
prevent the too-early free and hang if other similar bugs should remain in
other parts of the code.
This commit is contained in:
Kristian Nielsen
2015-04-08 11:01:18 +02:00
parent 880f2273fd
commit 3b961347db
7 changed files with 231 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

58
mysql-test/suite/rpl/r/rpl_parallel.result
View File

@@ -1421,6 +1421,64 @@ a b
99 99
include/stop_slave.inc
SET GLOBAL slave_transaction_retries= @old_retries;
SET GLOBAL slave_parallel_threads=10;
include/start_slave.inc
*** MDEV-7888: ANALYZE TABLE does wakeup_subsequent_commits(), causing wrong binlog order and parallel replication hang ***
include/stop_slave.inc
SET @old_dbug= @@GLOBAL.debug_dbug;
SET GLOBAL debug_dbug= '+d,inject_analyze_table_sleep';
SET @old_dbug= @@SESSION.debug_dbug;
SET SESSION debug_dbug="+d,binlog_force_commit_id";
SET @commit_id= 10000;
ANALYZE TABLE t2;
Table Op Msg_type Msg_text
test.t2 analyze status OK
INSERT INTO t3 VALUES (120, 0);
SET @commit_id= 10001;
INSERT INTO t3 VALUES (121, 0);
SET SESSION debug_dbug=@old_dbug;
SELECT * FROM t3 WHERE a >= 120 ORDER BY a;
a b
120 0
121 0
include/save_master_gtid.inc
include/start_slave.inc
include/sync_with_master_gtid.inc
SELECT * FROM t3 WHERE a >= 120 ORDER BY a;
a b
120 0
121 0
include/stop_slave.inc
SET GLOBAL debug_dbug= @old_debug;
include/start_slave.inc
*** MDEV-7929: record_gtid() for non-transactional event group calls wakeup_subsequent_commits() too early, causing slave hang. ***
include/stop_slave.inc
SET @old_dbug= @@GLOBAL.debug_dbug;
SET GLOBAL debug_dbug= '+d,inject_record_gtid_serverid_100_sleep';
SET @old_dbug= @@SESSION.debug_dbug;
SET SESSION debug_dbug="+d,binlog_force_commit_id";
SET @old_server_id= @@SESSION.server_id;
SET SESSION server_id= 100;
SET @commit_id= 10010;
ALTER TABLE t1 COMMENT "Hulubulu!";
SET SESSION server_id= @old_server_id;
INSERT INTO t3 VALUES (130, 0);
SET @commit_id= 10011;
INSERT INTO t3 VALUES (131, 0);
SET SESSION debug_dbug=@old_dbug;
SELECT * FROM t3 WHERE a >= 130 ORDER BY a;
a b
130 0
131 0
include/save_master_gtid.inc
include/start_slave.inc
include/sync_with_master_gtid.inc
SELECT * FROM t3 WHERE a >= 130 ORDER BY a;
a b
130 0
131 0
include/stop_slave.inc
SET GLOBAL debug_dbug= @old_debug;
include/start_slave.inc
include/stop_slave.inc
SET GLOBAL slave_parallel_threads=@old_parallel_threads;