MDEV-32050: Look up tables in the purge coordinator

dr-m · dr-m · commit 88733282fb15 · 2023-10-25T10:08:20.000+03:00
The InnoDB table lookup in purge worker threads is a bottleneck that can
degrade a slow shutdown to utilize less than 2 threads. Let us fix that
bottleneck by constructing a local lookup table that does not require any
synchronization while the undo log records of the current batch
are being processed.

TRX_PURGE_TABLE_BUCKETS: The initial number of std::unordered_map
hash buckets used during a purge batch. This could avoid some
resizing and rehashing in trx_purge_attach_undo_recs().

purge_node_t::tables: A lookup table from table ID to an already
looked up and locked table. Replaces many fields.

trx_purge_attach_undo_recs(): Look up each table in the purge batch
only once.

trx_purge(): Close all tables and release MDL at the end of the batch.

trx_purge_table_open(), trx_purge_table_acquire(): Open a table in purge
and acquire a metadata lock on it. This replaces
dict_table_open_on_id&lt;true&gt;() and dict_acquire_mdl_shared().

purge_sys_t::close_and_reopen(): In case of an MDL conflict, close and
reopen all tables that are covered by the current purge batch.
It may be that some of the tables have been dropped meanwhile and can
be ignored. This replaces wait_SYS() and wait_FTS().

row_purge_parse_undo_rec(): Make purge_coordinator_task issue a
MDL warrant to any purge_worker_task which might need it
when innodb_purge_threads&gt;1.

purge_node_t::end(): Clear the MDL warrant.

Reviewed by: Vladislav Lesin and Vladislav Vaintroub
diff --git a/mysql-test/suite/innodb/r/innodb-index-debug.result b/mysql-test/suite/innodb/r/innodb-index-debug.result
@@ -1,3 +1,7 @@
+SET GLOBAL innodb_max_purge_lag_wait=0;
+connect  stop_purge,localhost,root;
+START TRANSACTION WITH CONSISTENT SNAPSHOT;
+connection default;
 CREATE TABLE t1(c1 INT NOT NULL, c2 INT, PRIMARY KEY(c1)) Engine=InnoDB;
 SHOW CREATE TABLE t1;
 Table	Create Table
@@ -143,6 +147,7 @@ ERROR 23000: Duplicate entry '0' for key 'i'
 SET DEBUG_SYNC='now SIGNAL log2';
 connection con1;
 disconnect con1;
+disconnect stop_purge;
 connection default;
 SET DEBUG_SYNC='RESET';
 DROP TABLE t1;
diff --git a/mysql-test/suite/innodb/r/innodb-index-online.result b/mysql-test/suite/innodb/r/innodb-index-online.result
@@ -437,6 +437,7 @@ COUNT(c22f)
 CHECK TABLE t1;
 Table	Op	Msg_type	Msg_text
 test.t1	check	status	OK
+SET GLOBAL innodb_max_purge_lag_wait=0;
 ALTER TABLE t1 ADD UNIQUE INDEX c3p5(c3(5));
 ERROR 23000: Duplicate entry 'NULL' for key 'c3p5'
 UPDATE t1 SET c3 = NULL WHERE c3 = '';
diff --git a/mysql-test/suite/innodb/r/innodb-table-online.result b/mysql-test/suite/innodb/r/innodb-table-online.result
@@ -274,6 +274,7 @@ WHERE variable_name = 'innodb_encryption_n_rowlog_blocks_encrypted');
 SET @rowlog_decrypt_1=
 (SELECT variable_value FROM information_schema.global_status
 WHERE variable_name = 'innodb_encryption_n_rowlog_blocks_decrypted');
+SET GLOBAL innodb_max_purge_lag_wait=0;
 SET DEBUG_SYNC = 'row_log_table_apply1_before SIGNAL rebuilt3 WAIT_FOR dml3_done';
 ALTER TABLE t1 ADD PRIMARY KEY(c22f), CHANGE c2 c22f INT;
 ERROR 42000: Multiple primary key defined
diff --git a/mysql-test/suite/innodb/r/row_format_redundant.result b/mysql-test/suite/innodb/r/row_format_redundant.result
@@ -42,14 +42,14 @@ TRUNCATE TABLE t2;
 ERROR HY000: Table 't2' is read only
 TRUNCATE TABLE t3;
 ERROR HY000: Table 't3' is read only
-# restart: --innodb-data-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-log-group-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-data-file-path=ibdata1:1M:autoextend --innodb-undo-tablespaces=0 --innodb-stats-persistent=0
+# restart: --innodb-data-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-log-group-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-data-file-path=ibdata1:1M:autoextend --innodb-undo-tablespaces=0 --innodb-stats-persistent=0 --skip-innodb-fast-shutdown
 TRUNCATE TABLE t1;
 TRUNCATE TABLE t2;
 TRUNCATE TABLE t3;
 corrupted SYS_TABLES.MIX_LEN for test/t1
 corrupted SYS_TABLES.MIX_LEN for test/t2
 corrupted SYS_TABLES.MIX_LEN for test/t3
-# restart: --innodb-data-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-log-group-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-data-file-path=ibdata1:1M:autoextend --innodb-undo-tablespaces=0 --innodb-stats-persistent=0
+# restart: --innodb-data-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-log-group-home-dir=MYSQLTEST_VARDIR/tmp/row_format_redundant --innodb-data-file-path=ibdata1:1M:autoextend --innodb-undo-tablespaces=0 --innodb-stats-persistent=0 --skip-innodb-fast-shutdown
 TRUNCATE TABLE t1;
 ERROR 42S02: Table 'test.t1' doesn't exist in engine
 TRUNCATE TABLE t2;
diff --git a/mysql-test/suite/innodb/t/innodb-index-debug.test b/mysql-test/suite/innodb/t/innodb-index-debug.test
@@ -6,6 +6,10 @@
 --source include/big_test.inc
 
 let $MYSQLD_DATADIR= `select @@datadir`;
+SET GLOBAL innodb_max_purge_lag_wait=0;
+connect (stop_purge,localhost,root);
+START TRANSACTION WITH CONSISTENT SNAPSHOT;
+connection default;
 
 #
 # Test for BUG# 12739098, check whether trx->error_status is reset on error.
@@ -155,6 +159,7 @@ SET DEBUG_SYNC='now SIGNAL log2';
 --connection con1
 reap;
 --disconnect con1
+--disconnect stop_purge
 --connection default
 SET DEBUG_SYNC='RESET';
 DROP TABLE t1;
diff --git a/mysql-test/suite/innodb/t/innodb-index-online.test b/mysql-test/suite/innodb/t/innodb-index-online.test
@@ -369,6 +369,8 @@ SELECT
 connection con1;
 SELECT COUNT(c22f) FROM t1;
 CHECK TABLE t1;
+# Avoid a strange DEBUG_SYNC timeout on c3p5_created.
+SET GLOBAL innodb_max_purge_lag_wait=0;
 
 # Create a column prefix index.
 --error ER_DUP_ENTRY
diff --git a/mysql-test/suite/innodb/t/innodb-table-online.test b/mysql-test/suite/innodb/t/innodb-table-online.test
@@ -255,6 +255,7 @@ SET @rowlog_decrypt_1=
 (SELECT variable_value FROM information_schema.global_status
 WHERE variable_name = 'innodb_encryption_n_rowlog_blocks_decrypted');
 
+SET GLOBAL innodb_max_purge_lag_wait=0;
 # Accumulate and apply some modification log.
 SET DEBUG_SYNC = 'row_log_table_apply1_before SIGNAL rebuilt3 WAIT_FOR dml3_done';
 --error ER_MULTIPLE_PRI_KEY
diff --git a/mysql-test/suite/innodb/t/row_format_redundant.test b/mysql-test/suite/innodb/t/row_format_redundant.test
@@ -78,7 +78,7 @@ TRUNCATE TABLE t2;
 --error ER_OPEN_AS_READONLY
 TRUNCATE TABLE t3;
 
---let $restart_parameters= $d
+--let $restart_parameters= $d --skip-innodb-fast-shutdown
 --source include/restart_mysqld.inc
 
 TRUNCATE TABLE t1;
diff --git a/mysql-test/suite/parts/r/partition_special_innodb.result b/mysql-test/suite/parts/r/partition_special_innodb.result
@@ -269,6 +269,7 @@ Table	Op	Msg_type	Msg_text
 test.t1	check	status	OK
 connection default;
 UNLOCK TABLES;
+SET GLOBAL innodb_max_purge_lag_wait=0;
 connection con2;
 DROP TABLE t1;
 connection default;
@@ -348,6 +349,7 @@ SELECT * FROM t1;
 ERROR HY000: Lock wait timeout exceeded; try restarting transaction
 connection default;
 UNLOCK TABLES;
+SET GLOBAL innodb_max_purge_lag_wait=0;
 connection con2;
 DROP TABLE t1;
 disconnect con2;
diff --git a/mysql-test/suite/parts/t/partition_special_innodb.test b/mysql-test/suite/parts/t/partition_special_innodb.test
@@ -114,6 +114,8 @@ CHECK TABLE t1;
 
 --connection default
 UNLOCK TABLES;
+# Avoid a MDL conflict between purge and the DROP TABLE below
+SET GLOBAL innodb_max_purge_lag_wait=0;
 
 --connection con2
 DROP TABLE t1;
@@ -222,6 +224,8 @@ SELECT * FROM t1;
 
 --connection default
 UNLOCK TABLES;
+# Avoid a MDL conflict between purge and the DROP TABLE below
+SET GLOBAL innodb_max_purge_lag_wait=0;
 
 --connection con2
 DROP TABLE t1;
diff --git a/storage/innobase/dict/dict0dict.cc b/storage/innobase/dict/dict0dict.cc
@@ -662,7 +662,7 @@ dict_table_t::parse_name<>(char(&)[NAME_LEN + 1], char(&)[NAME_LEN + 1],
 @param[in]      table_op        operation to perform when opening
 @return table object after locking MDL shared
 @retval nullptr if the table is not readable, or if trylock && MDL blocked */
-template<bool trylock, bool purge_thd>
+template<bool trylock>
 dict_table_t*
 dict_acquire_mdl_shared(dict_table_t *table,
                         THD *thd,
@@ -678,7 +678,6 @@ dict_acquire_mdl_shared(dict_table_t *table,
 
   if (trylock)
   {
-    static_assert(!trylock || !purge_thd, "usage");
     dict_sys.freeze(SRW_LOCK_CALL);
     db_len= dict_get_db_name_len(table->name.m_name);
     dict_sys.unfreeze();
@@ -749,25 +748,14 @@ dict_acquire_mdl_shared(dict_table_t *table,
     }
   }
 
-retry_table_open:
   dict_sys.freeze(SRW_LOCK_CALL);
-  if (purge_thd && purge_sys.must_wait_FTS())
-  {
-    not_found= reinterpret_cast<dict_table_t*>(-1);
-    goto return_without_mdl;
-  }
   table= dict_sys.find_table(table_id);
   if (table)
     table->acquire();
   if (!table && table_op != DICT_TABLE_OP_OPEN_ONLY_IF_CACHED)
   {
     dict_sys.unfreeze();
     dict_sys.lock(SRW_LOCK_CALL);
-    if (purge_thd && purge_sys.must_wait_FTS())
-    {
-      dict_sys.unlock();
-      goto retry_table_open;
-    }
     table= dict_load_table_on_id(table_id,
                                  table_op == DICT_TABLE_OP_LOAD_TABLESPACE
                                  ? DICT_ERR_IGNORE_RECOVER_LOCK
@@ -826,24 +814,21 @@ dict_acquire_mdl_shared(dict_table_t *table,
   goto retry;
 }
 
-template dict_table_t* dict_acquire_mdl_shared<false, false>
+template dict_table_t* dict_acquire_mdl_shared<false>
 (dict_table_t*,THD*,MDL_ticket**,dict_table_op_t);
-template dict_table_t* dict_acquire_mdl_shared<true, false>
+template dict_table_t* dict_acquire_mdl_shared<true>
 (dict_table_t*,THD*,MDL_ticket**,dict_table_op_t);
 
 /** Look up a table by numeric identifier.
-@tparam purge_thd Whether the function is called by purge thread
 @param[in]      table_id        table identifier
 @param[in]      dict_locked     data dictionary locked
 @param[in]      table_op        operation to perform when opening
 @param[in,out]  thd             background thread, or NULL to not acquire MDL
 @param[out]     mdl             mdl ticket, or NULL
 @return table, NULL if does not exist */
-template <bool purge_thd>
-dict_table_t*
-dict_table_open_on_id(table_id_t table_id, bool dict_locked,
-                      dict_table_op_t table_op, THD *thd,
-                      MDL_ticket **mdl)
+dict_table_t *dict_table_open_on_id(table_id_t table_id, bool dict_locked,
+                                    dict_table_op_t table_op, THD *thd,
+                                    MDL_ticket **mdl)
 {
   if (!dict_locked)
     dict_sys.freeze(SRW_LOCK_CALL);
@@ -852,16 +837,9 @@ dict_table_open_on_id(table_id_t table_id, bool dict_locked,
 
   if (table)
   {
-    if (purge_thd && purge_sys.must_wait_FTS())
-    {
-      table= reinterpret_cast<dict_table_t*>(-1);
-      goto func_exit;
-    }
-
     table->acquire();
     if (thd && !dict_locked)
-      table= dict_acquire_mdl_shared<false, purge_thd>(
-               table, thd, mdl, table_op);
+      table= dict_acquire_mdl_shared<false>(table, thd, mdl, table_op);
   }
   else if (table_op != DICT_TABLE_OP_OPEN_ONLY_IF_CACHED)
   {
@@ -875,44 +853,26 @@ dict_table_open_on_id(table_id_t table_id, bool dict_locked,
                                  ? DICT_ERR_IGNORE_RECOVER_LOCK
                                  : DICT_ERR_IGNORE_FK_NOKEY);
     if (table)
-    {
-      if (purge_thd && purge_sys.must_wait_FTS())
-      {
-        dict_sys.unlock();
-        return reinterpret_cast<dict_table_t*>(-1);
-      }
       table->acquire();
-    }
     if (!dict_locked)
     {
       dict_sys.unlock();
       if (table && thd)
       {
         dict_sys.freeze(SRW_LOCK_CALL);
-        table= dict_acquire_mdl_shared<false, purge_thd>(
-                 table, thd, mdl, table_op);
+        table= dict_acquire_mdl_shared<false>(table, thd, mdl, table_op);
         dict_sys.unfreeze();
       }
       return table;
     }
   }
 
-func_exit:
   if (!dict_locked)
     dict_sys.unfreeze();
 
   return table;
 }
 
-template dict_table_t* dict_table_open_on_id<false>
-(table_id_t table_id, bool dict_locked,
- dict_table_op_t table_op, THD *thd,
- MDL_ticket **mdl);
-template dict_table_t* dict_table_open_on_id<true>
-(table_id_t table_id, bool dict_locked,
- dict_table_op_t table_op, THD *thd,
- MDL_ticket **mdl);
-
 /********************************************************************//**
 Looks for column n position in the clustered index.
 @return position in internal representation of the clustered index */
diff --git a/storage/innobase/fts/fts0fts.cc b/storage/innobase/fts/fts0fts.cc
@@ -1609,10 +1609,11 @@ and common table associated with the fts table.
 			already stopped*/
 void purge_sys_t::stop_FTS(const dict_table_t &table, bool already_stopped)
 {
-  dict_sys.lock(SRW_LOCK_CALL);
   if (!already_stopped)
     purge_sys.stop_FTS();
 
+  dict_sys.lock(SRW_LOCK_CALL);
+
   fts_table_t fts_table;
   char table_name[MAX_FULL_NAME_LEN];
 
diff --git a/storage/innobase/include/dict0dict.h b/storage/innobase/include/dict0dict.h
@@ -132,7 +132,7 @@ enum dict_table_op_t {
 @param[in]      table_op        operation to perform when opening
 @return table object after locking MDL shared
 @retval NULL if the table is not readable, or if trylock && MDL blocked */
-template<bool trylock, bool purge_thd= false>
+template<bool trylock>
 dict_table_t*
 dict_acquire_mdl_shared(dict_table_t *table,
                         THD *thd,
@@ -146,7 +146,6 @@ dict_acquire_mdl_shared(dict_table_t *table,
 @param[in,out]  thd             background thread, or NULL to not acquire MDL
 @param[out]     mdl             mdl ticket, or NULL
 @return table, NULL if does not exist */
-template<bool purge_thd= false>
 dict_table_t*
 dict_table_open_on_id(table_id_t table_id, bool dict_locked,
                       dict_table_op_t table_op, THD *thd= nullptr,
diff --git a/storage/innobase/include/row0purge.h b/storage/innobase/include/row0purge.h
diff --git a/storage/innobase/include/trx0purge.h b/storage/innobase/include/trx0purge.h
diff --git a/storage/innobase/include/trx0types.h b/storage/innobase/include/trx0types.h
diff --git a/storage/innobase/row/row0purge.cc b/storage/innobase/row/row0purge.cc
diff --git a/storage/innobase/srv/srv0srv.cc b/storage/innobase/srv/srv0srv.cc
diff --git a/storage/innobase/trx/trx0purge.cc b/storage/innobase/trx/trx0purge.cc
