Merge the patches into MariaDB 10.0 main.
With this patch, parallel replication will now automatically retry a
transaction that fails due to deadlock or other temporary error, same as
single-threaded replication.
We catch deadlocks with InnoDB transactions due to enforced commit order. If
T1 must commit before T2 in parallel replication and T1 ends up waiting for T2
inside InnoDB, we kill T2 and retry it later to resolve the deadlock
automatically.
When a MyISAM query is killed midway, the query is logged to the binlog marked
with the error.
The slave does not attempt to run the query, but aborts with a suitable error
message in the error log for the DBA to act on.
In this case, the parallel replication code would check the sql_errno() code,
even no my_error() had been set. In debug builds, this causes an assertion.
Fixed the code to check that we actually have an error set before querying for
an error code.
After-review changes.
For this patch in 10.0, we do not introduce a new public storage engine API,
we just fix the InnoDB/XtraDB issues. In 10.1, we will make a better public
API that can be used for all storage engines (MDEV-6429).
Eliminate the background thread that did deadlock kills asynchroneously.
Instead, we ensure that the InnoDB/XtraDB code can handle doing the kill from
inside the deadlock detection code (when thd_report_wait_for() needs to kill a
later thread to resolve a deadlock).
(We preserve the part of the original patch that introduces dedicated mutex
and condition for the slave init thread, to remove the abuse of
LOCK_thread_count for start/stop synchronisation of the slave init thread).
The sql_slave_skip_counter is important to be able to recover replication from
certain errors. Often, an appropriate solution is to set
sql_slave_skip_counter to skip over a problem event. But setting
sql_slave_skip_counter produced an error in GTID mode, with a suggestion to
instead set @@gtid_slave_pos to point past the problem event. This however is
not always possible; for example, in case of an INCIDENT event, that event
does not have any GTID to assign to @@gtid_slave_pos.
With this patch, sql_slave_skip_counter now works in GTID mode the same was as
in non-GTID mode. When set, that many initial events are skipped when the SQL
thread starts, plus as many extra events are needed to completely skip any
partially skipped event group. The GTID position is updated to point past the
skipped event(s).
Analysis: For some reason table stats for a table pointed from a index
is not initialized. Added additional warning output on this situation
and table stats initialization. This is better than asserting.
These tests use search_pattern_in_file.inc to search the error log for
expected output. However, search_pattern_in_file.inc by default searched only
the first 50000 bytes, so if the error log grew too big the tests would fail.
This patch extends search_pattern_in_file.inc with an option to specify how
much of the file to search, and whether to search from the start of the file
or from the end. Then the rpl.rpl_checksum and rpl.rpl_gtid_errorlog test
cases are fixed to search the last 50000 bytes of the error log, which will
work no matter how large prior tests have made it.
The direct cause of the assertion was missing error handling in
record_gtid(). If ha_commit_trans() fails for the statement commit, there was
missing code to catch the error and do ha_rollback_trans() in this case; this
caused close_thread_tables() to assert.
Normally, this error case is not hit, but in this case it was triggered due to
another bug: When a transaction T1 fails during parallel replication, the code
would signal following transactions that they could start to run without
properly marking the error condition. This caused subsequent transactions to
incorrectly start replicating, only to get an error later during their own
commit step. This was particularly serious if the subsequent transactions were
DDL or MyISAM updates, which cannot be rolled back and would leave replication
in an inconsistent state.
Fixed by 1) in case of error, only signal following transactions to continue
once the error has been properly marked and those transactions will know not
to start; and 2) implement proper error handling in record_gtid() in the case
that statement commit fails.
If replication breaks in GTID mode, it is not trivial to determine the GTID of
the failing event group. This is a problem, as such GTID is needed eg. to
explicitly set @@gtid_slave_pos to skip to after that event group, or to
compare errors on different servers, etc.
Fix by ensuring that relevant slave errors logged to the error log include the
GTID of the event group containing the problem event.
This is MySQL Bug#59123. The message string stored in an INCIDENT event was
not zero-terminated. This caused any following checksum bytes (if enabled on
the master) to be output to the error log as trailing garbage when the message
was printed to the error log.
Backport the patch from MySQL 5.6:
revno: 2876.228.200
revision-id: zhenxing.he@sun.com-20110111051323-w2xnzvcjn46x6h6u
committer: He Zhenxing <zhenxing.he@sun.com>
timestamp: Tue 2011-01-11 13:13:23 +0800
message:
BUG#59123 rpl_stm_binlog_max_cache_size fails sporadically with found warnings
Also add a test case.
MySQL 5.6 implemented WL#344, which is about a MASTER_DELAY option to CHANGE
MASTER. But as part of this worklog, the format of the realy-log.info file was
changed. The new format is not understood by earlier versions, and nor by
MariaDB 10.0, so changing server to those versions would cause the slave to
abort with an error due to reading incorrect data out of relay-log.info.
Fix this by backporting from the WL#344 patch just the code that understands
the new relay-log.info format. We still write out the old format, and none of
the MASTER_DELAY feature is backported with this commit.
The INCIDENT_EVENT always caused slave error and abort, without checking
--slave-skip-errors.
Now, if error 1590, ER_SLAVE_INCIDENT is included in the --slave-skip-errors
list, incident events will be ignored.
This is a merge of this MySQL 5.6 patch:
revision-id: frazer@mysql.com-20110314170916-ypgin17otj3ucx95
committer: Frazer Clement <frazer@mysql.com>
timestamp: Mon 2011-03-14 17:09:16 +0000
message:
Bug#11799671 NOT POSSIBLE TO SKIP INCIDENT ERRORS
replication causing replication to fail.
Remove the temporary fix for MDEV-5914, which used READ COMMITTED for parallel
replication worker threads. Replace it with a better, more selective solution.
The issue is with certain edge cases of InnoDB gap locks, for example between
INSERT and ranged DELETE. It is possible for the gap lock set by the DELETE to
block the INSERT, if the DELETE runs first, while the record lock set by
INSERT does not block the DELETE, if the INSERT runs first. This can cause a
conflict between the two in parallel replication on the slave even though they
ran without conflicts on the master.
With this patch, InnoDB will ask the server layer about the two involved
transactions before blocking on a gap lock. If the server layer tells InnoDB
that the transactions are already fixed wrt. commit order, as they are in
parallel replication, InnoDB will ignore the gap lock and allow the two
transactions to proceed in parallel, avoiding the conflict.
Improve the fix for MDEV-6020. When InnoDB itself detects a deadlock, it now
asks the server layer for any preferences about which transaction to roll
back. In case of parallel replication with two transactions T1 and T2 fixed to
commit T1 before T2, the server layer will ask InnoDB to roll back T2 as the
deadlock victim, not T1. This helps in some cases to avoid excessive deadlock
rollback, as T2 will in any case need to wait for T1 to complete before it can
itself commit.
Also some misc. fixes found during development and testing:
- Remove thd_rpl_is_parallel(), it is not used or needed.
- Use KILL_CONNECTION instead of KILL_QUERY when a parallel replication
worker thread is killed to resolve a deadlock with fixed commit
ordering. There are some cases, eg. in sql/sql_parse.cc, where a KILL_QUERY
can be ignored if the query otherwise completed successfully, and this
could cause the deadlock kill to be lost, so that the deadlock was not
correctly resolved.
- Fix random test failure due to missing wait_for_binlog_checkpoint.inc.
- Make sure that deadlock or other temporary errors during parallel
replication are not printed to the the error log; there were some places
around the replication code with extra error logging. These conditions can
occur occasionally and are handled automatically without breaking
replication, so they should not pollute the error log.
- Fix handling of rgi->gtid_sub_id. We need to be able to access this also at
the end of a transaction, to be able to detect and resolve deadlocks due to
commit ordering. But this value was also used as a flag to mark whether
record_gtid() had been called, by being set to zero, losing the value. Now,
introduce a separate flag rgi->gtid_pending, so rgi->gtid_sub_id remains
valid for the entire duration of the transaction.
- Fix one place where the code to handle ignored errors called reset_killed()
unconditionally, even if no error was caught that should be ignored. This
could cause loss of a deadlock kill signal, breaking deadlock detection and
resolution.
- Fix a couple of missing mysql_reset_thd_for_next_command(). This could
cause a prior error condition to remain for the next event executed,
causing assertions about errors already being set and possibly giving
incorrect error handling for following event executions.
- Fix code that cleared thd->rgi_slave in the parallel replication worker
threads after each event execution; this caused the deadlock detection and
handling code to not be able to correctly process the associated
transactions as belonging to replication worker threads.
- Remove useless error code in slave_background_kill_request().
- Fix bug where wfc->wakeup_error was not cleared at
wait_for_commit::unregister_wait_for_prior_commit(). This could cause the
error condition to wrongly propagate to a later wait_for_prior_commit(),
causing spurious ER_PRIOR_COMMIT_FAILED errors.
- Do not put the binlog background thread into the processlist. It causes
too many result differences in mtr, but also it probably is not useful
for users to pollute the process list with a system thread that does not
really perform any user-visible tasks...
replication causing replication to fail.
In parallel replication, we run transactions from the master in parallel, but
force them to commit in the same order they did on the master. If we force T1
to commit before T2, but T2 holds eg. a row lock that is needed by T1, we get
a deadlock when T2 waits until T1 has committed.
Usually, we do not run T1 and T2 in parallel if there is a chance that they
can have conflicting locks like this, but there are certain edge cases where
it can occasionally happen (eg. MDEV-5914, MDEV-5941, MDEV-6020). The bug was
that this would cause replication to hang, eventually getting a lock timeout
and causing the slave to stop with error.
With this patch, InnoDB will report back to the upper layer whenever a
transactions T1 is about to do a lock wait on T2. If T1 and T2 are parallel
replication transactions, and T2 needs to commit later than T1, we can thus
detect the deadlock; we then kill T2, setting a flag that causes it to catch
the kill and convert it to a deadlock error; this error will then cause T2 to
roll back and release its locks (so that T1 can commit), and later T2 will be
re-tried and eventually also committed.
The kill happens asynchroneously in a slave background thread; this is
necessary, as the reporting from InnoDB about lock waits happen deep inside
the locking code, at a point where it is not possible to directly call
THD::awake() due to mutexes held.
Deadlock is assumed to be (very) rarely occuring, so this patch tries to
minimise the performance impact on the normal case where no deadlocks occur,
rather than optimise the handling of the occasional deadlock.
Also fix transaction retry due to deadlock when it happens after a transaction
already signalled to later transactions that it started to commit. In this
case we need to undo this signalling (and later redo it when we commit again
during retry), so following transactions will not start too early.
Also add a missing thd->send_kill_message() that got triggered during testing
(this corrects an incorrect fix for MySQL Bug#58933).
Handle retry of event groups that span multiple relay log files.
- If retry reaches the end of one relay log file, move on to the next.
- Handle refcounting of relay log files, and avoid purging relay log
files until all event groups have completed that might have needed
them for transaction retry.
Implement that if first retry fails, we can do another attempt.
Add testcases to test multi-retry that succeeds in second attempt, and
multi-retry that eventually fails due to exceeding slave_trans_retries.
on select from I_S.INNODB_CHANGED_PAGES
Analysis: limit_lsn_range_from_condition() incorrectly parses
start_lsn and/or end_lsn conditions.
Fix from SergeyP. Added some test cases.
Start implementing that an event group can be re-tried in parallel replication
if it fails with a temporary error (like deadlock).
Patch is very incomplete, just some very basic retry works.
Stuff still missing (not complete list):
- Handle moving to the next relay log file, if event group to be retried
spans multiple relay log files.
- Handle refcounting of relay log files, to ensure that we do not purge a
relay log file and then later attempt to re-execute events out of it.
- Handle description_event_for_exec - we need to save this somehow for the
possible retry - and use the correct one in case it differs between relay
logs.
- Do another retry attempt in case the first retry also fails.
- Limit the max number of retries.
- Lots of testing will be needed for the various edge cases.
BREAKS RBR
Analysis:
--------
A table created using a query of the format:
CREATE TABLE t1 AS SELECT REPEAT('A',1000) DIV 1 AS a;
breaks the Row Based Replication.
The query above creates a table having a field of datatype
'bigint' with a display width of 3000 which is beyond the
maximum acceptable value of 255.
In the RBR mode, CREATE TABLE SELECT statement is
replicated as a combination of CREATE TABLE statement
equivalent to one the returned by SHOW CREATE TABLE and
row events for rows inserted. When this CREATE TABLE event
is executed on the slave, an error is reported:
Display width out of range for column 'a' (max = 255)
The following is the output of 'SHOW CREATE TABLE t1':
CREATE TABLE t1(`a` bigint(3000) DEFAULT NULL)
ENGINE=InnoDB DEFAULT CHARSET=latin1;
The problem is due to the combination of two facts:
1) The above CREATE TABLE SELECT statement uses the display
width of the result of DIV operation as the display width
of the column created without validating the width for out
of bound condition.
2) The DIV operation incorrectly returns the length of its first
argument as the display width of its result; thus allowing
creation of a table with an incorrect display width of 3000
for the field.
Fix:
----
This fix changes the DIV operation implementation to correctly
evaluate the display width of its result. We check if DIV's
results estimated width crosses maximum width for integer
value (21) and if yes set it to this maximum value.
This patch also fixes fixes maximum display width evaluation
for DIV function when its first argument is in UCS2.
