Problem:- Gtid are not transferred in Galera Cluster.
Solution:- We need to transfer gtid in the case on either when cluster is
slave/master in async replication. In normal Gtid replication gtid are generated on
recieving node itself and it is always on sync with other nodes. Because galera keeps
node in sync , So all nodes get same no of event groups. So the issue arises when
say galera is slave in async replication.
A
| (Async replication)
D <-> E <-> F {Galera replication}
So what should happen is that all node should apply the master gtid but this does
node happen, becuase node E, F does not recieve gtid from D in write set , So what E(or F)
does is that it applies wsrep_gtid_domain_id, D server-id , E gtid next seq no. This
generated gtid does not always work when say A has different domain id.
So In this commit, on galera node when we see that this event is recieved from master
we simply write Gtid_Log_Event in write_set and send it to other nodes.
As reported in MDEV-11969 "there's no way to ditch knowledge" about some
domain that is no longer updated on a server. Besides being of annoyance to
clutter output in DBA console stale domains can prevent the slave
to connect the master as MDEV-12012 witnesses.
What domain is obsolete must be evaluated by the user (DBA) according
to whether the domain info is still relevant and will the domain ever
receive any update.
This patch introduces a method to discard obsolete gtid domains from
the server binlog state. The removal requires no event group from such
domain present in existing binlog files though. If there are any the
containing logs must be first PURGEd in order for
FLUSH BINARY LOGS DELETE_DOMAIN_ID=(list-of-domains)
succeed. Otherwise the command returns an error.
The list of obsolete domains can be computed through
intersecting two sets - the earliest (first) binlog's Gtid_list
and the current value of @@global.gtid_binlog_state - and extracting
the domain id components from the intersection list items.
The new DELETE_DOMAIN_ID featured FLUSH continues to rotate binlog
omitting the deleted domains from the active binlog file's Gtid_list.
Notice though when the command is ineffective - that none of requested to delete
domain exists in the binlog state - rotation does not occur.
Obsolete domain deletion is not harmful for connected slaves as long
as master side binlog files *purge* is synchronized with FLUSH-DELETE_DOMAIN_ID.
The slaves must have the last event from purged files processed as usual,
in order not to bump later into requesting a gtid from a file which
was already gone.
While the command is not replicated (as ordinary FLUSH BINLOG LOGS is)
slaves, even though having extra domains, won't suffer from reconnection errors
thanks to master-slave gtid connection protocol allowing the master
to be ignorant about a gtid domain.
Should at failover such slave to be promoted into master role it may run
the ex-master's
FLUSH BINARY LOGS DELETE_DOMAIN_ID=(list-of-domains)
to clean its own binlog state.
NOTES.
suite/perfschema/r/start_server_low_digest.result
is re-recorded as consequence of internal parser codes changes.
.. file '/var/log/mysql/mariadb-bin.000001' not found in binlog
index, needed for recovery. Aborting.
In Galera cluster, while preparing for rsync/xtrabackup based
SST, the donor node takes an FTWRL followed by (REFRESH_ENGINE_LOG
in rsync based state transfer and) REFRESH_BINARY_LOG. The latter
rotates the binary log and logs Binlog_checkpoint_log_event
corresponding to the penultimate binary log file into the new file.
The checkpoint event for the current file is later logged
synchronously by binlog_background_thread.
Now, since in rsync/xtrabackup based snapshot state transfer methods,
only the last binary log file is transferred to the joiner node; the
file could get transferred even before the checkpoint event for the
same file gets written to it. As a result, the joiner node would fail
to start complaining about the missing binlog file needed for recovery.
In order to fix this, a mechanism has been put in place to make
REFRESH_BINARY_LOG operation wait for Binlog_checkpoint_log_event
to be logged for the current binary log file if the node is part of
a Galera cluster. As further safety, during rsync based state transfer
the donor node now acquires and owns LOCK_log for the duration of file
transfer during SST.
Revert following bug fix:
Bug#20685029: SLAVE IO THREAD SHOULD STOP WHEN DISK IS
FULL
Bug#21753696: MAKE SHOW SLAVE STATUS NON BLOCKING IF IO
THREAD WAITS FOR DISK SPACE
This fix results in a deadlock between slave IO thread
and SQL thread.
(cherry picked from commit e3fea6c6dbb36c6ab21c4ab777224560e9608b53)
FULL
Bug#21753696: MAKE SHOW SLAVE STATUS NON BLOCKING IF IO
THREAD WAITS FOR DISK SPACE
Problem:
========
Currently SHOW SLAVE STATUS blocks if IO thread waits for
disk space. This makes automation tools verifying
server health block on taking relevant action. Finally this
will create SHOW SLAVE STATUS piles.
Analysis:
=========
SHOW SLAVE STATUS hangs on mi->data_lock if relay log write
is waiting for free disk space while holding mi->data_lock.
mi->data_lock is needed to protect the format description
event (mi->format_description_event) which is accessed by
the clients running FLUSH LOGS and slave IO thread. Note
relay log writes don't need to be protected by
mi->data_lock, LOCK_log is used to protect relay log between
IO and SQL thread (see MYSQL_BIN_LOG::append_event). The
code takes mi->data_lock to protect
mi->format_description_event during relay log rotate which
might get triggered right after relay log write.
Fix:
====
Release the data_lock just for the duration of writing into
relay log.
Made change to ensure the following lock order is maintained
to avoid deadlocks.
data_lock, LOCK_log
data_lock is held during relay log rotations to protect
the description event.
Introduce Log_event_writer() that encapsulates
writing data to an IO_CACHE with automatic checksum calculation.
Now all events properly checksum themselves as needed.
Use Log_event_writer in MYSQL_BIN_LOG::write_cache() instead
of copy-pasting its logic all over.
Later Log_event_writer will also do encryption.
Other things:
- Avoid calling init_and_set_log_file_name() when opening binary log.
- Remove newlines early when reading from index file.
- Ensure that reset_logs() will work even if thd is 0 (Can happen on startup)
- Added thd to sart_slave_threads() for better error handling.
Add log_bin_index, log_bin_basename and relay_log_basename system
variables. Also, convert relay_log_index system variable to
NO_CMD_LINE and implement --relay-log-index as a command line
option.
Do not use format function attribute for sql_print_xxx() family of
functions as they use a MariaDB-specific extension of printf instead
of one provided by the system.
remove group_commit_reason_immediate
rename group_commit_reason_transaction to group_commit_trigger_lock_wait
rename group_commit_reason_usec to group_commit_trigger_timeout
rename group_commit_reason_count to group_commit_triggger_count
The following global status variables where added:
* binlog_group_commit_reason_count
* binlog_group_commit_reason_usec
* binlog_group_commit_reason_transaction
* binlog_group_commit_reason_immediate
binlog_group_commit_reason_count corresponds to group commits made by
virtue of the binlog_commit_wait_count variable.
binlog_group_commit_reason_usec corresponds to the binlog_commit_wait_usec
variable.
binlog_group_commit_reason_transaction is a result of ordered
transaction that need to occur in the same order on the slave and can't
be parallelised.
binlog_group_commit_reason_immediate is caused to prevent stalls with
row locks as described in log.cc:binlog_report_wait_for. This immediate
count is also counted a second time in binlog_group_commit_reason_transaction.
Overall binlog_group_commits = binlog_group_commit_reason_count +
binlog_group_commit_reason_usec + binlog_group_commit_reason_transaction
This work was funded thanks to Open Source Developers Club Australia.
Parallel replication (in 10.0 / "conservative" mode) relies on binlog group
commits to group transactions that can be safely run in parallel on the
slave. The --binlog-commit-wait-count and --binlog-commit-wait-usec options
exist to increase the number of commits per group. But in case of conflicts
between transactions, this can cause unnecessary delay and reduced througput,
especially on a slave where commit order is fixed.
This patch adds a heuristics to reduce this problem. When transaction T1 goes
to commit, it will first wait for N transactions to queue up for a group
commit. However, if we detect that another transaction T2 is waiting for a row
lock held by T1, then we will skip the wait and let T1 commit immediately,
releasing locks and let T2 continue.
On a slave, this avoids the unfortunate situation where T1 is waiting for T2
to join the group commit, but T2 is waiting for T1 to release locks, causing
no work to be done for the duration of the --binlog-commit-wait-usec timeout.
(The heuristic seems reasonable on the master as well, so it is enabled for
all transactions, not just replication transactions).
Using a boolean flag for 'there is a RESET MASTER in progress' doesn't
work very well for multiple concurrent RESET MASTER statements.
Changed to a counter.
log-tc-size is 24K by default. Page size is 64K on PPC64. But log-tc-size
must be at least 3 x page size. This is enforced by TC_LOG_MMAP::open()
with a comment: to guarantee non-empty pool.
This all makes server not startable in default configuration on PPC64.
Autosize log-tc-size, so that it's min value= page size * 3, default
value= page size * 6, block size= page size.
Implement --semi-sync-master-wait-point=AFTER_SYNC|AFTER_COMMIT.
When AFTER_SYNC, the semi-sync wait will be done earlier, before the storage
engine commit rather than after. This means that a transaction will not be
visible on the master until at least one slave has received it.
Make the binlog dump threads not need to take LOCK_log while sending
binlog events to slave. Instead, a new LOCK_binlog_end_pos is used
just to coordinate tracking the current end-of-log.
This is a pre-requisite for MDEV-162, "Enhanced semisync
replication". It should also help reduce the contention on LOCK_log on
a busy master.
Also does some much-needed refactoring/cleanup of the related code in
the binlog dump thread.
Merged lp:maria/maria-10.0-galera up to revision 3879.
Added a new functions to handler API to forcefully abort_transaction,
producing fake_trx_id, get_checkpoint and set_checkpoint for XA. These
were added for future possiblity to add more storage engines that
could use galera replication.
We don't support changing tc_log implementation at run time.
If the first XA-capable engine is loaded with INSTALL PLUGIN - disable its
XA capabilities with a warning
Now if CREATE OR REPLACE fails but we have deleted a table already, we will generate a DROP TABLE in the binary log.
This fixes this issue.
In addition, for a failing CREATE OR REPLACE TABLE ... SELECT we don't generate a log of all the inserted rows, only the DROP TABLE.
I added code for not logging DROP TEMPORARY TABLE for tables where the CREATE TABLE was not logged. This code will be activated in 10.1
by removing the code protected by DONT_LOG_DROP_OF_TEMPORARY_TABLES.
mysql-test/suite/rpl/r/create_or_replace_mix.result:
More test cases
mysql-test/suite/rpl/r/create_or_replace_row.result:
More test cases
mysql-test/suite/rpl/r/create_or_replace_statement.result:
More test cases
mysql-test/suite/rpl/t/create_or_replace.inc:
More test cases
sql/log.cc:
Added binlog_reset_cache() to clear the binary log.
sql/log.h:
Added prototype
sql/sql_insert.cc:
If CREATE OR REPLACE TABLE ... SELECT fails:
- Don't log anything if nothing changed
- If table was deleted, log a DROP TABLE.
Remember if we table creation of temporary tables was logged.
sql/sql_table.cc:
Added log_drop_table()
Remember if we table creation of temporary tables was logged.
If CREATE OR REPLACE TABLE ... SELECT fails and a table was deleted, log a DROP TABLE.
sql/sql_table.h:
Added prototype
sql/sql_truncate.cc:
Remember if we table creation of temporary tables was logged.
sql/table.h:
Added table_creation_was_logged
The problem is a deadlock between MYSQL_BIN_LOG::reset_logs() and
MYSQL_BIN_LOG::mark_xid_done(). The former takes LOCK_log and waits for the
latter to complete. But the latter also tries to take LOCK_log; this can lead
to a deadlock.
There was already code that tries to deal with this, with the flag
reset_master_pending. However, there was still a small opportunity for
deadlock, when an previous mark_xid_done() is still running when reset_logs()
is called and is at the precise point where it first releases LOCK_xid_list
and then re-aquires both LOCK_log and LOCK_xid_list.
Solve by setting reset_master_pending in reset_logs() before taking
LOCK_log. And also count how many invocations of LOCK_xid_list are in the
progress of releasing and re-aquiring locks, and in reset_logs() wait for that
number to drop to zero after setting reset_master_pending and before taking
LOCK_log.
Using CREATE OR REPLACE TABLE is be identical to
DROP TABLE IF EXISTS table_name;
CREATE TABLE ...;
Except that:
* CREATE OR REPLACE is be atomic (now one can create the same table between drop and create).
* Temporary tables will not shadow the table name for the DROP as the CREATE TABLE tells us already if we are using a temporary table or not.
* If the table was locked with LOCK TABLES, the new table will be locked with the same lock after it's created.
Implementation details:
- We don't anymore open the to-be-created table during CREATE TABLE, which the original code did.
- There is no need to open a table we are planning to create. It's enough to check if the table exists or not.
- Removed some of duplicated code for CREATE IF NOT EXISTS.
- Give an error when using CREATE OR REPLACE with IF NOT EXISTS (conflicting options).
- As a side effect of the code changes, we don't anymore have to internally re-prepare prepared statements with CREATE TABLE if the table exists.
- Made one code path for all testing if log table are in use.
- Better error message if one tries to create/drop/alter a log table in use
- Added back disabled rpl_row_create_table test as it now seams to work and includes a lot of interesting tests.
- Added HA_LEX_CREATE_REPLACE to mark if we are using CREATE OR REPLACE
- Aligned CREATE OR REPLACE parsing code in sql_yacc.yy for TABLE and VIEW
- Changed interface for drop_temporary_table() to make it more reusable
- Changed Locked_tables_list::init_locked_tables() to work on the table object instead of the table list object. Before this it used a mix of both, which was not good.
- Locked_tables_list::unlock_locked_tables(THD *thd) now requires a valid thd argument. Old usage of calling this with 0 i changed to instead call Locked_tables_list::reset()
- Added functions Locked_tables_list:restore_lock() and Locked_tables_list::add_back_last_deleted_lock() to be able to easily add back a locked table to the lock list.
- Added restart_trans_for_tables() to be able to restart a transaction.
- DROP_ACL is required if one uses CREATE TABLE OR REPLACE.
- Added drop of normal and temporary tables in create_table_imp() if CREATE OR REPLACE was used.
- Added reacquiring of table locks in mysql_create_table() and mysql_create_like_table()
mysql-test/include/commit.inc:
With new code we get fewer status increments
mysql-test/r/commit_1innodb.result:
With new code we get fewer status increments
mysql-test/r/create.result:
Added testing of create or replace with timeout
mysql-test/r/create_or_replace.result:
Basic testing of CREATE OR REPLACE TABLE
mysql-test/r/partition_exchange.result:
New error message
mysql-test/r/ps_ddl.result:
Fewer reprepares with new code
mysql-test/suite/archive/discover.result:
Don't rediscover archive tables if the .frm file exists
(Sergei will look at this if there is a better way...)
mysql-test/suite/archive/discover.test:
Don't rediscover archive tables if the .frm file exists
(Sergei will look at this if there is a better way...)
mysql-test/suite/funcs_1/r/innodb_views.result:
New error message
mysql-test/suite/funcs_1/r/memory_views.result:
New error message
mysql-test/suite/rpl/disabled.def:
rpl_row_create_table should now be safe to use
mysql-test/suite/rpl/r/rpl_row_create_table.result:
Updated results after adding back disabled test
mysql-test/suite/rpl/t/rpl_create_if_not_exists.test:
Added comment
mysql-test/suite/rpl/t/rpl_row_create_table.test:
Added CREATE OR REPLACE TABLE test
mysql-test/t/create.test:
Added CREATE OR REPLACE TABLE test
mysql-test/t/create_or_replace-master.opt:
Create logs
mysql-test/t/create_or_replace.test:
Basic testing of CREATE OR REPLACE TABLE
mysql-test/t/partition_exchange.test:
Error number changed as we are now using same code for all log table change issues
mysql-test/t/ps_ddl.test:
Fewer reprepares with new code
sql/handler.h:
Moved things around a bit in a structure to get better alignment.
Added HA_LEX_CREATE_REPLACE to mark if we are using CREATE OR REPLACE
Added 3 elements to end of HA_CREATE_INFO to be able to store state to add backs locks in case of LOCK TABLES.
sql/log.cc:
Reimplemented check_if_log_table():
- Simpler and faster usage
- Can give error messages
This gives us one code path for allmost all error messages if log tables are in use
sql/log.h:
New interface for check_if_log_table()
sql/slave.cc:
More logging
sql/sql_alter.cc:
New interface for check_if_log_table()
sql/sql_base.cc:
More documentation
Changed interface for drop_temporary_table() to make it more reusable
Changed Locked_tables_list::init_locked_tables() to work on the table object instead of the table list object. Before this it used a mix of both, which was not good.
Locked_tables_list::unlock_locked_tables(THD *thd) now requires a valid thd argument. Old usage of calling this with 0 i changed to instead call Locked_tables_list::reset()
Added functions Locked_tables_list:restore_lock() and Locked_tables_list::add_back_last_deleted_lock() to be able to easily add back a locked table to the lock list.
Check for command number instead of open_strategy of CREATE TABLE was used.
Added restart_trans_for_tables() to be able to restart a transaction. This was needed in "create or replace ... select" between the drop table and the select.
sql/sql_base.h:
Added and updated function prototypes
sql/sql_class.h:
Added new prototypes to Locked_tables_list class
Added extra argument to select_create to avoid double call to eof() or send_error()
- I needed this in some edge case where the table was not created against expections.
sql/sql_db.cc:
New interface for check_if_log_table()
sql/sql_insert.cc:
Remember position to lock information so that we can reaquire table lock for LOCK TABLES + CREATE OR REPLACE TABLE SELECT. Later add back the lock by calling restore_lock().
Removed one not needed indentation level in create_table_from_items()
Ensure we don't call send_eof() or abort_result_set() twice.
sql/sql_lex.h:
Removed variable that I temporarly added in an earlier changeset
sql/sql_parse.cc:
Removed old test code (marked with QQ)
Ensure that we have open_strategy set as TABLE_LIST::OPEN_STUB in CREATE TABLE
Removed some IF NOT EXISTS code as this is now handled in create_table_table_impl().
Set OPTION_KEEP_LOGS later. This code had to be moved as the test for IF EXISTS has changed place.
DROP_ACL is required if one uses CREATE TABLE OR REPLACE.
sql/sql_partition_admin.cc:
New interface for check_if_log_table()
sql/sql_rename.cc:
New interface for check_if_log_table()
sql/sql_table.cc:
New interface for check_if_log_table()
Moved some code in mysql_rm_table() under a common test.
- Safe as temporary tables doesn't have statistics.
- !is_temporary_table(table) test was moved out from drop_temporary_table() and merged with upper level code.
- Added drop of normal and temporary tables in create_table_imp() if CREATE OR REPLACE was used.
- Added reacquiring of table locks in mysql_create_table() and mysql_create_like_table()
- In mysql_create_like_table(), restore table->open_strategy() if it was changed.
- Re-test if table was a view after opening it.
sql/sql_table.h:
New prototype for mysql_create_table_no_lock()
sql/sql_yacc.yy:
Added syntax for CREATE OR REPLACE TABLE
Reuse new code for CREATE OR REPLACE VIEW
sql/table.h:
Added name for enum type
sql/table_cache.cc:
More DBUG
MDEV-4725: Incorrect binlog state recovery if crash while writing event group
The binlog state was not recovered correctly if XA is not used (eg. InnoDB
disabled), or if server crashed in the middle of writing an event group to the
binlog.
With this patch, we ensure that recovery of binlog state is done even if we do
not do the full XA binlog recovery, and we ensure that we only recover fully
written event groups into the binlog state.
Implement @@gtid_binlog_state. This is the internal state of the binlog
(most recent GTID logged for every domain_id and server_id). This allows
to save the state before RESET MASTER and restore it afterwards.
Fix a bunch of issues found with locking, ordering, and non-thread-safe stuff
in Relay_log_info.
Now able to do a simple benchmark, showing 4.5 times speedup for applying a
binlog with 10000 REPLACE statements.
Impement options --binlog-commit-wait-count and
--binlog-commit-wait-usec.
These options permit the DBA to deliberately increase latency
of an individual commit to get more transactions in each
binlog group commit. This increases the opportunity for
parallel replication on the slave, and can also decrease I/O
load on the master.
The options also make it easier to test the parallel
replication with mysql-test-run.
Implement facility for the commit in one thread to wait for the commit of
another to complete first. The wait is done in a way that does not hinder
that a waiter and a waitee can group commit together with a single fsync()
in both binlog and InnoDB. The wait is done efficiently with respect to
locking.
The patch was originally made to support TaoBao parallel replication with
in-order commit; now it will be adapted to also be used for parallel
replication of group-committed transactions.
A waiter THD registers itself with a prior waitee THD. The waiter will then
complete its commit at the earliest in the same group commit of the waitee
(when using binlog). The wait can also be done explicitly by the waitee.
