mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-01-28 01:34:17 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/mysql-test/suite/binlog/t/binlog_xa_recover.test

276 lines
9.5 KiB
Text
Raw Normal View History

MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/have_innodb.inc
--source include/have_debug.inc
--source include/have_debug_sync.inc
--source include/have_binlog_format_row.inc
# Valgrind does not work well with test that crashes the server
--source include/not_valgrind.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
# (We do not need to restore these settings, as we crash the server).
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
SET GLOBAL max_binlog_size= 4096;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET GLOBAL innodb_flush_log_at_trx_commit= 1;
RESET MASTER;
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
CREATE TABLE t1 (a INT PRIMARY KEY, b MEDIUMTEXT) ENGINE=Innodb;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
# Insert some data to force a couple binlog rotations (3), so we get some
# normal binlog checkpoints before starting the test.
INSERT INTO t1 VALUES (100, REPEAT("x", 4100));
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
# Wait for the master-bin.000002 binlog checkpoint to appear.
--let $wait_for_all= 0
--let $show_statement= SHOW BINLOG EVENTS IN "master-bin.000002"
--let $field= Info
--let $condition= = "master-bin.000002"
--source include/wait_show_condition.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
INSERT INTO t1 VALUES (101, REPEAT("x", 4100));
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
--let $wait_for_all= 0
--let $show_statement= SHOW BINLOG EVENTS IN "master-bin.000003"
--let $field= Info
--let $condition= = "master-bin.000003"
--source include/wait_show_condition.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
INSERT INTO t1 VALUES (102, REPEAT("x", 4100));
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
--let $wait_for_all= 0
--let $show_statement= SHOW BINLOG EVENTS IN "master-bin.000004"
--let $field= Info
--let $condition= = "master-bin.000004"
--source include/wait_show_condition.inc
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
# Now start a bunch of transactions that span multiple binlog
# files. Leave then in the state prepared-but-not-committed in the engine
# and crash the server. Check that crash recovery is able to recover all
# of them.
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
#
# We use debug_sync to get all the transactions into the prepared state before
# we commit any of them. This is because the prepare step flushes the InnoDB
# redo log - including any commits made before, so recovery would become
# unnecessary, decreasing the value of this test.
#
# We arrange to have con1 with a prepared transaction in master-bin.000004,
# con2 and con3 with a prepared transaction in master-bin.000005, and a new
# empty master-bin.000006. So the latest binlog checkpoint should be
# master-bin.000006.
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connect(con1,localhost,root,,);
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
# First wait after prepare and before write to binlog.
SET DEBUG_SYNC= "ha_commit_trans_before_log_and_order SIGNAL con1_wait WAIT_FOR con1_cont";
# Then complete InnoDB commit in memory (but not commit checkpoint / write to
# disk), and hang until crash, leaving a transaction to be XA recovered.
SET DEBUG_SYNC= "commit_after_group_release_commit_ordered SIGNAL con1_ready WAIT_FOR _ever";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
send INSERT INTO t1 VALUES (1, REPEAT("x", 4100));
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now WAIT_FOR con1_wait";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connect(con2,localhost,root,,);
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "ha_commit_trans_before_log_and_order SIGNAL con2_wait WAIT_FOR con2_cont";
SET DEBUG_SYNC= "commit_after_group_release_commit_ordered SIGNAL con2_ready WAIT_FOR _ever";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
send INSERT INTO t1 VALUES (2, NULL);
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now WAIT_FOR con2_wait";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connect(con3,localhost,root,,);
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "ha_commit_trans_before_log_and_order SIGNAL con3_wait WAIT_FOR con3_cont";
SET DEBUG_SYNC= "commit_after_group_release_commit_ordered SIGNAL con3_ready WAIT_FOR _ever";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
send INSERT INTO t1 VALUES (3, REPEAT("x", 4100));
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now WAIT_FOR con3_wait";
connect(con4,localhost,root,,);
SET DEBUG_SYNC= "ha_commit_trans_before_log_and_order SIGNAL con4_wait WAIT_FOR con4_cont";
SET SESSION debug_dbug="+d,crash_commit_after_log";
send INSERT INTO t1 VALUES (4, NULL);
connection default;
SET DEBUG_SYNC= "now WAIT_FOR con4_wait";
SET DEBUG_SYNC= "now SIGNAL con1_cont";
SET DEBUG_SYNC= "now WAIT_FOR con1_ready";
SET DEBUG_SYNC= "now SIGNAL con2_cont";
SET DEBUG_SYNC= "now WAIT_FOR con2_ready";
SET DEBUG_SYNC= "now SIGNAL con3_cont";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
SET DEBUG_SYNC= "now WAIT_FOR con3_ready";
# Check that everything is committed in binary log.
--source include/show_binary_logs.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--let $binlog_file= master-bin.000003
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--let $binlog_start= 4
--source include/show_binlog_events.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--let $binlog_file= master-bin.000004
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binlog_events.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--let $binlog_file= master-bin.000005
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binlog_events.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--let $binlog_file= master-bin.000006
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binlog_events.inc
# Check that server will not purge too much.
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
PURGE BINARY LOGS TO "master-bin.000006";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binary_logs.inc
# Now crash the server with one more transaction in prepared state.
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--write_file $MYSQLTEST_VARDIR/tmp/mysqld.1.expect
wait-binlog_xa_recover.test
EOF
MDEV-26, Global transaction ID. Fix three more test failures seen in buildbot (one of them ported from main tree).
2013-04-09 11:46:38 +02:00
--error 0,2006,2013
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now SIGNAL con4_cont";
connection con4;
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--error 2006,2013
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
reap;
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
Add missing check for thd->killed in mysql_binlog_send(). The slave dump thread running on the master only checked thd->killed whenever it reached the end of a binlog file, not between events. This could unnecessarily delay server shutdown. This was found by code inspection while tracking down some occasional "forcing close of thread..." errors in Buildbot. Hopefully this will fix the failures, but the fix is correct in any case. Also increase the wait during server shutdown, 2 seconds is a bit tight in case of heavy I/O stall, and it seems better to delay shutdown a bit than force-kill threads unnecessarily. Also fix some races in test cases that restart the mysqld server. The .expect file should be changed with --append_file, --remove_file + --write_file creates a short window where mysqld can error out due to .expect file missing.
2013-04-24 13:05:40 +02:00
--append_file $MYSQLTEST_VARDIR/tmp/mysqld.1.expect
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
restart-group_commit_binlog_pos.test
EOF
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection default;
--enable_reconnect
--source include/wait_until_connected_again.inc
# Check that all transactions are recovered.
SELECT a FROM t1 ORDER BY a;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--echo Test that with multiple binlog checkpoints, recovery starts from the last one.
SET GLOBAL max_binlog_size= 4096;
SET GLOBAL innodb_flush_log_at_trx_commit= 1;
RESET MASTER;
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
# Rotate to binlog master-bin.000003 while delaying binlog checkpoints.
# So we get multiple binlog checkpoints in master-bin.000003.
# Then complete the checkpoints, crash, and check that we only scan
# the necessary binlog file (ie. that we use the _last_ checkpoint).
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connect(con10,localhost,root,,);
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "commit_after_group_release_commit_ordered SIGNAL con10_ready WAIT_FOR con10_cont";
send INSERT INTO t1 VALUES (10, REPEAT("x", 4100));
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection default;
SET DEBUG_SYNC= "now WAIT_FOR con10_ready";
connect(con11,localhost,root,,);
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "commit_after_group_release_commit_ordered SIGNAL con11_ready WAIT_FOR con11_cont";
send INSERT INTO t1 VALUES (11, REPEAT("x", 4100));
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now WAIT_FOR con11_ready";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
connect(con12,localhost,root,,);
SET DEBUG_SYNC= "commit_after_group_release_commit_ordered SIGNAL con12_ready WAIT_FOR con12_cont";
send INSERT INTO t1 VALUES (12, REPEAT("x", 4100));
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now WAIT_FOR con12_ready";
INSERT INTO t1 VALUES (13, NULL);
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binary_logs.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--let $binlog_file= master-bin.000004
--let $binlog_start= 4
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binlog_events.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now SIGNAL con10_cont";
connection con10;
reap;
connection default;
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
# We need to sync the test case with the background processing of the
# commit checkpoint, otherwise we get nondeterministic results.
SET @old_dbug= @@global.DEBUG_DBUG;
SET GLOBAL debug_dbug="+d,binlog_background_checkpoint_processed";
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now SIGNAL con12_cont";
connection con12;
reap;
connection default;
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
SET DEBUG_SYNC= "now WAIT_FOR binlog_background_checkpoint_processed";
SET GLOBAL debug_dbug= @old_dbug;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
SET DEBUG_SYNC= "now SIGNAL con11_cont";
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection con11;
reap;
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
# Wait for the last (master-bin.000004) binlog checkpoint to appear.
--let $wait_for_all= 0
--let $show_statement= SHOW BINLOG EVENTS IN "master-bin.000004"
--let $field= Info
--let $condition= = "master-bin.000004"
--source include/wait_show_condition.inc
--echo Checking that master-bin.000004 is the last binlog checkpoint
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
--source include/show_binlog_events.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--echo Now crash the server
# It is not too easy to test XA recovery, as it runs early during server
# startup, before any connections can be made.
# What we do is set a DBUG error insert which will crash if XA recovery
# starts from any other binlog than master-bin.000004 (check the file
# binlog_xa_recover-master.opt). Then we will fail here if XA recovery
# would start from the wrong place.
--write_file $MYSQLTEST_VARDIR/tmp/mysqld.1.expect
wait-binlog_xa_recover.test
EOF
SET SESSION debug_dbug="+d,crash_commit_after_log";
--error 2006,2013
INSERT INTO t1 VALUES (14, NULL);
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
Add missing check for thd->killed in mysql_binlog_send(). The slave dump thread running on the master only checked thd->killed whenever it reached the end of a binlog file, not between events. This could unnecessarily delay server shutdown. This was found by code inspection while tracking down some occasional "forcing close of thread..." errors in Buildbot. Hopefully this will fix the failures, but the fix is correct in any case. Also increase the wait during server shutdown, 2 seconds is a bit tight in case of heavy I/O stall, and it seems better to delay shutdown a bit than force-kill threads unnecessarily. Also fix some races in test cases that restart the mysqld server. The .expect file should be changed with --append_file, --remove_file + --write_file creates a short window where mysqld can error out due to .expect file missing.
2013-04-24 13:05:40 +02:00
--append_file $MYSQLTEST_VARDIR/tmp/mysqld.1.expect
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
restart-group_commit_binlog_pos.test
EOF
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--enable_reconnect
--source include/wait_until_connected_again.inc
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
# Check that all transactions are recovered.
SELECT a FROM t1 ORDER BY a;
--echo *** Check that recovery works if we crashed early during rotate, before
--echo *** binlog checkpoint event could be written.
SET GLOBAL max_binlog_size= 4096;
SET GLOBAL innodb_flush_log_at_trx_commit= 1;
RESET MASTER;
# We need some initial data to reach binlog master-bin.000004. Otherwise
# crash recovery fails due to the error insert used for previous test.
INSERT INTO t1 VALUES (21, REPEAT("x", 4100));
INSERT INTO t1 VALUES (22, REPEAT("x", 4100));
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
# Wait for the master-bin.000003 binlog checkpoint to appear.
--let $wait_for_all= 0
--let $show_statement= SHOW BINLOG EVENTS IN "master-bin.000003"
--let $field= Info
--let $condition= = "master-bin.000003"
--source include/wait_show_condition.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
INSERT INTO t1 VALUES (23, REPEAT("x", 4100));
MDEV-532: Async InnoDB commit checkpoint. Make the commit checkpoint inside InnoDB be asynchroneous. Implement a background thread in binlog to do the writing and flushing of binlog checkpoint events to disk.
2012-12-14 15:38:07 +01:00
# Wait for the last (master-bin.000004) binlog checkpoint to appear.
--let $wait_for_all= 0
--let $show_statement= SHOW BINLOG EVENTS IN "master-bin.000004"
--let $field= Info
--let $condition= = "master-bin.000004"
--source include/wait_show_condition.inc
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
--write_file $MYSQLTEST_VARDIR/tmp/mysqld.1.expect
wait-binlog_xa_recover.test
EOF
SET SESSION debug_dbug="+d,crash_before_write_checkpoint_event";
--error 2006,2013
INSERT INTO t1 VALUES (24, REPEAT("x", 4100));
Add missing check for thd->killed in mysql_binlog_send(). The slave dump thread running on the master only checked thd->killed whenever it reached the end of a binlog file, not between events. This could unnecessarily delay server shutdown. This was found by code inspection while tracking down some occasional "forcing close of thread..." errors in Buildbot. Hopefully this will fix the failures, but the fix is correct in any case. Also increase the wait during server shutdown, 2 seconds is a bit tight in case of heavy I/O stall, and it seems better to delay shutdown a bit than force-kill threads unnecessarily. Also fix some races in test cases that restart the mysqld server. The .expect file should be changed with --append_file, --remove_file + --write_file creates a short window where mysqld can error out due to .expect file missing.
2013-04-24 13:05:40 +02:00
--append_file $MYSQLTEST_VARDIR/tmp/mysqld.1.expect
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
restart-group_commit_binlog_pos.test
EOF
--enable_reconnect
--source include/wait_until_connected_again.inc
# Check that all transactions are recovered.
SELECT a FROM t1 ORDER BY a;
--source include/show_binary_logs.inc
--let $binlog_file= master-bin.000004
--let $binlog_start= 4
--source include/show_binlog_events.inc
MDEV-181: XID crash recovery across binlog boundaries Keep track of how many pending XIDs (transactions that are prepared in storage engine and written into binlog, but not yet durably committed on disk in the engine) there are in each binlog. When the count of one binlog drops to zero, write a new binlog checkpoint event, telling which is the oldest binlog with pending XIDs. When doing XA recovery after a crash, check the last binlog checkpoint event, and scan all binlog files from that point onwards for XIDs that must be committed if found in prepared state inside engine. Remove the code in binlog rotation that waits for all prepared XIDs to be committed before writing a new binlog file (this is no longer necessary when recovery can scan multiple binlog files).
2012-06-22 11:46:28 +02:00
# Cleanup
connection default;
MDEV-232: Remove one fsync() from commit phase. Introduce a new storage engine API method commit_checkpoint_request(). This is used to replace the fsync() at the end of every storage engine commit with a single fsync() when a binlog is rotated. Binlog rotation is now done during group commit instead of being delayed until unlog(), removing some server stall and avoiding an expensive lock/unlock of LOCK_log inside unlog().
2012-09-13 14:31:29 +02:00
DROP TABLE t1;
Reference in a new issue Copy permalink