Bug #11766227: InnoDB purge lag much worse for 5.5.8 versus 5.1

Bug #11766501: Multiple RBS break the get rseg with mininum trx_t::no code during purge
      
Bug# 59291 changes:
      
Main problem is that truncating the UNDO log at the completion of every
trx_purge() call is expensive as the number of rollback segments is increased.
We truncate after a configurable amount of pages. The innodb_purge_batch_size
parameter is used to control when InnoDB does the actual truncate. The truncate
is done once after 128 (or TRX_SYS_N_RSEGS iterations). In other words we
truncate after purge 128 * innodb_purge_batch_size. The smaller the batch
size the quicker we truncate.
      
Introduce a new parameter that allows how many rollback segments to use for
storing REDO information. This is really step 1 in allowing complete control
to the user over rollback space management.
      
New parameters:
    i) innodb_rollback_segments = number of rollback_segments to use
       (default is now 128) dynamic parameter, can be changed anytime.
       Currently there is little benefit in changing it from the default.
      
Optimisations in the patch.
      
    i. Change the O(n) behaviour of trx_rseg_get_on_id() to O(log n)
       Backported from 5.6. Refactor some of the binary heap code.
       Create a new include/ut0bh.ic file.
      
    ii. Avoid truncating the rollback segments after every purge.
      
Related changes that were moved to a separate patch:
      
    i. Purge should not do any flushing, only wait for space to be free so that
       it only does purging of records unless it is held up by a long running
       transaction that is preventing it from progressing.
      
   ii. Give the purge thread preference over transactions when acquiring the
       rseg->mutex during commit. This to avoid purge blocking unnecessarily
       when getting the next rollback segment to purge.
      
Bug #11766501 changes:
      
Add the rseg to the min binary heap under the cover of the kernel mutex and
the binary heap mutex. This ensures the ordering of the min binary heap.
      
The two changes have to be committed together because they share the same
that fixes both issues.
      
rb://567 Approved by: Inaam Rana.

storage/innobase/CMakeLists.txt

-# Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2006, 2011, Oracle and/or its affiliates. All rights reserved.
 # 
 # This program is free software; you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -246,7 +246,7 @@ SET(INNOBASE_SOURCES	btr/btr0btr.c btr/btr0cur.c btr/btr0pcur.c btr/btr0sea.c
 			trx/trx0sys.c trx/trx0trx.c trx/trx0undo.c
 			usr/usr0sess.c
 			ut/ut0byte.c ut/ut0dbg.c ut/ut0list.c ut/ut0mem.c ut/ut0rbt.c ut/ut0rnd.c
-			ut/ut0ut.c ut/ut0vec.c ut/ut0wqueue.c)
+			ut/ut0ut.c ut/ut0vec.c ut/ut0wqueue.c ut/ut0bh.c)
 
 IF(WITH_INNODB)
   # Legacy option

storage/innobase/handler/ha_innodb.cc

 /*****************************************************************************
 
-Copyright (c) 2000, 2010, MySQL AB & Innobase Oy. All Rights Reserved.
+Copyright (c) 2000, 2011, MySQL AB & Innobase Oy. All Rights Reserved.
 Copyright (c) 2008, 2009 Google Inc.
 Copyright (c) 2009, Percona Inc.
 
@@ -262,7 +262,7 @@ static PSI_mutex_info all_innodb_mutexes[] = {
 #  endif /* UNIV_MEM_DEBUG */
 	{&mem_pool_mutex_key, "mem_pool_mutex", 0},
 	{&mutex_list_mutex_key, "mutex_list_mutex", 0},
-	{&purge_sys_mutex_key, "purge_sys_mutex", 0},
+	{&purge_sys_bh_mutex_key, "purge_sys_bh_mutex", 0},
 	{&recv_sys_mutex_key, "recv_sys_mutex", 0},
 	{&rseg_mutex_key, "rseg_mutex", 0},
 #  ifdef UNIV_SYNC_DEBUG
@@ -10971,16 +10971,23 @@ static MYSQL_SYSVAR_ULONG(io_capacity, srv_io_capacity,
 
 static MYSQL_SYSVAR_ULONG(purge_batch_size, srv_purge_batch_size,
   PLUGIN_VAR_OPCMDARG,
-  "Number of UNDO logs to purge in one batch from the history list. "
-  "Default is 20",
+  "Number of UNDO log pages to purge in one batch from the history list.",
   NULL, NULL,
   20,			/* Default setting */
   1,			/* Minimum value */
-  5000, 0);		/* Maximum value */
+  10000, 0);		/* Maximum value */
+
+static MYSQL_SYSVAR_ULONG(rollback_segments, srv_rollback_segments,
+  PLUGIN_VAR_OPCMDARG,
+  "Number of UNDO logs to use.",
+  NULL, NULL,
+  128,			/* Default setting */
+  1,			/* Minimum value */
+  TRX_SYS_N_RSEGS, 0);	/* Maximum value */
 
 static MYSQL_SYSVAR_ULONG(purge_threads, srv_n_purge_threads,
   PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY,
-  "Purge threads can be either 0 or 1. Default is 0.",
+  "Purge threads can be either 0 or 1.",
   NULL, NULL,
   0,			/* Default setting */
   0,			/* Minimum value */
@@ -11331,6 +11338,7 @@ static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(io_capacity),
   MYSQL_SYSVAR(purge_threads),
   MYSQL_SYSVAR(purge_batch_size),
+  MYSQL_SYSVAR(rollback_segments),
   NULL
 };
 

storage/innobase/include/buf0flu.h

@@ -138,7 +138,18 @@ UNIV_INTERN
 void
 buf_flush_wait_batch_end(
 /*=====================*/
-	buf_pool_t*	buf_pool,	/*!< buffer pool instance */
+	buf_pool_t*	buf_pool,	/*!< in: buffer pool instance */
+	enum buf_flush	type);		/*!< in: BUF_FLUSH_LRU
+					or BUF_FLUSH_LIST */
+/******************************************************************//**
+Waits until a flush batch of the given type ends. This is called by
+a thread that only wants to wait for a flush to end but doesn't do
+any flushing itself. */
+UNIV_INTERN
+void
+buf_flush_wait_batch_end_wait_only(
+/*===============================*/
+	buf_pool_t*	buf_pool,	/*!< in: buffer pool instance */
 	enum buf_flush	type);		/*!< in: BUF_FLUSH_LRU
 					or BUF_FLUSH_LIST */
 /********************************************************************//**

storage/innobase/include/srv0srv.h

 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Innobase Oy. All Rights Reserved.
 Copyright (c) 2008, 2009, Google Inc.
 Copyright (c) 2009, Percona Inc.
 
@@ -294,9 +294,12 @@ extern ulint srv_log_waits;
 /* the number of purge threads to use from the worker pool (currently 0 or 1) */
 extern ulong srv_n_purge_threads;
 
-/* the number of records to purge in one batch */
+/* the number of pages to purge in one batch */
 extern ulong srv_purge_batch_size;
 
+/* the number of rollback segments to use */
+extern ulong srv_rollback_segments;
+
 /* variable that counts amount of data read in total (in bytes) */
 extern ulint srv_data_read;
 

storage/innobase/include/sync0sync.h

 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Innobase Oy. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -93,7 +93,7 @@ extern mysql_pfs_key_t	mem_hash_mutex_key;
 # endif /* UNIV_MEM_DEBUG */
 extern mysql_pfs_key_t	mem_pool_mutex_key;
 extern mysql_pfs_key_t	mutex_list_mutex_key;
-extern mysql_pfs_key_t	purge_sys_mutex_key;
+extern mysql_pfs_key_t	purge_sys_bh_mutex_key;
 extern mysql_pfs_key_t	recv_sys_mutex_key;
 extern mysql_pfs_key_t	rseg_mutex_key;
 # ifdef UNIV_SYNC_DEBUG
@@ -637,7 +637,6 @@ or row lock! */
 #define SYNC_TREE_NODE_NEW	892
 #define SYNC_TREE_NODE_FROM_HASH 891
 #define SYNC_TREE_NODE		890
-#define	SYNC_PURGE_SYS		810
 #define	SYNC_PURGE_LATCH	800
 #define	SYNC_TRX_UNDO		700
 #define SYNC_RSEG		600
@@ -659,6 +658,7 @@ or row lock! */
 #define SYNC_REC_LOCK		299
 #define	SYNC_TRX_LOCK_HEAP	298
 #define SYNC_TRX_SYS_HEADER	290
+#define	SYNC_PURGE_QUEUE	200
 #define SYNC_LOG		170
 #define SYNC_LOG_FLUSH_ORDER	147
 #define SYNC_RECV		168

storage/innobase/include/trx0purge.h

 /*****************************************************************************
 
-Copyright (c) 1996, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Innobase Oy. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -68,8 +68,9 @@ Creates the global purge system control structure and inits the history
 mutex. */
 UNIV_INTERN
 void
-trx_purge_sys_create(void);
-/*======================*/
+trx_purge_sys_create(
+/*=================*/
+	ib_bh_t*	ib_bh);	/*!< in/own: UNDO log min binary heap*/
 /********************************************************************//**
 Frees the global purge system control structure. */
 UNIV_INTERN
@@ -128,20 +129,20 @@ struct trx_purge_struct{
 	ulint		state;		/*!< Purge system state */
 	sess_t*		sess;		/*!< System session running the purge
 					query */
-	trx_t*		trx;		/*!< System transaction running the purge
+	trx_t*		trx;		/*!< System transaction running the
+					purge
 					query: this trx is not in the trx list
 					of the trx system and it never ends */
 	que_t*		query;		/*!< The query graph which will do the
 					parallelized purge operation */
-	rw_lock_t	latch;		/*!< The latch protecting the purge view.
-					A purge operation must acquire an
-					x-latch here for the instant at which
+	rw_lock_t	latch;		/*!< The latch protecting the purge
+					view.  A purge operation must acquire
+					an x-latch here for the instant at which
 					it changes the purge view: an undo
 					log operation can prevent this by
 					obtaining an s-latch here. */
 	read_view_t*	view;		/*!< The purge will not remove undo logs
 					which are >= this view (purge view) */
-	mutex_t		mutex;		/*!< Mutex protecting the fields below */
 	ulint		n_pages_handled;/*!< Approximate number of undo log
 					pages processed in purge */
 	ulint		handle_limit;	/*!< Target of how many pages to get
@@ -179,6 +180,11 @@ struct trx_purge_struct{
 	mem_heap_t*	heap;		/*!< Temporary storage used during a
 					purge: can be emptied after purge
 					completes */
+	/*-----------------------------*/
+	ib_bh_t*	ib_bh;		/*!< Binary min-heap, ordered on
+					rseg_queue_t::trx_no. It is protected
+					by the bh_mutex */
+	mutex_t		bh_mutex;	/*!< Mutex protecting ib_bh */
 };
 
 #define TRX_PURGE_ON		1	/* purge operation is running */

storage/innobase/include/trx0rseg.h

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Innobase Oy. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -113,7 +113,9 @@ void
 trx_rseg_list_and_array_init(
 /*=========================*/
 	trx_sysf_t*	sys_header,	/*!< in: trx system header */
+	ib_bh_t*	ib_bh,		/*!< in: rseg queue */
 	mtr_t*		mtr);		/*!< in: mtr */
+
 /***************************************************************************
 Free's an instance of the rollback segment in memory. */
 UNIV_INTERN
@@ -180,6 +182,14 @@ struct trx_rseg_struct{
 					memory objects */
 };
 
+/** For prioritising the rollback segments for purge. */
+struct rseg_queue_struct {
+	trx_id_t	trx_no;		/*!< trx_rseg_t::last_trx_no */
+	trx_rseg_t*	rseg;		/*!< Rollback segment */
+};
+
+typedef struct rseg_queue_struct rseg_queue_t;
+
 /* Undo log segment slot in a rollback segment header */
 /*-------------------------------------------------------------*/
 #define	TRX_RSEG_SLOT_PAGE_NO	0	/* Page number of the header page of

storage/innobase/include/trx0sys.h

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Innobase Oy. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -38,6 +38,7 @@ Created 3/26/1996 Heikki Tuuri
 #include "mem0mem.h"
 #include "sync0sync.h"
 #include "ut0lst.h"
+#include "ut0bh.h"
 #include "read0types.h"
 #include "page0types.h"
 
@@ -221,13 +222,6 @@ UNIV_INLINE
 trx_id_t
 trx_sys_get_new_trx_id(void);
 /*========================*/
-/*****************************************************************//**
-Allocates a new transaction number.
-@return	new, allocated trx number */
-UNIV_INLINE
-trx_id_t
-trx_sys_get_new_trx_no(void);
-/*========================*/
 #endif /* !UNIV_HOTBACKUP */
 /*****************************************************************//**
 Writes a trx id to an index page. In case that the id size changes in

storage/innobase/include/trx0sys.ic

@@ -369,16 +369,4 @@ trx_sys_get_new_trx_id(void)
 	return(id);
 }
 
-/*****************************************************************//**
-Allocates a new transaction number.
-@return	new, allocated trx number */
-UNIV_INLINE
-trx_id_t
-trx_sys_get_new_trx_no(void)
-/*========================*/
-{
-	ut_ad(mutex_own(&kernel_mutex));
-
-	return(trx_sys_get_new_trx_id());
-}
 #endif /* !UNIV_HOTBACKUP */

storage/innobase/include/ut0bh.ic

+/***************************************************************************//**
+Copyright (c) 2011, Oracle Corpn. All Rights Reserved.
+
+This program is free software; you can redistribute it and/or modify it under
+the terms of the GNU General Public License as published by the Free Software
+Foundation; version 2 of the License.
+
+This program is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along with
+this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+Place, Suite 330, Boston, MA 02111-1307 USA
+
+*****************************************************************************/
+
+/******************************************************************//**
+@file include/ut0bh.ic
+Binary min-heap implementation.
+
+Created 2011-01-15 by Sunny Bains
+*******************************************************/
+
+#include "ut0bh.h"
+
+/**********************************************************************//**
+Get the number of elements in the binary heap.
+@return number of elements */
+UNIV_INLINE
+ulint
+ib_bh_size(
+/*=======*/
+	const ib_bh_t*	ib_bh)			/*!< in: instance */
+{
+	return(ib_bh->n_elems);
+}
+
+/**********************************************************************//**
+Test if binary heap is empty.
+@return TRUE if empty. */
+UNIV_INLINE
+ibool
+ib_bh_is_empty(
+/*===========*/
+	const ib_bh_t*	ib_bh)			/*!< in: instance */
+{
+	return(ib_bh_size(ib_bh) == 0);
+}
+
+/**********************************************************************//**
+Test if binary heap is full.
+@return TRUE if full. */
+UNIV_INLINE
+ibool
+ib_bh_is_full(
+/*===========*/
+	const ib_bh_t*	ib_bh)			/*!< in: instance */
+{
+	return(ib_bh_size(ib_bh) >= ib_bh->max_elems);
+}
+
+/**********************************************************************//**
+Get a pointer to the element.
+@return pointer to element */
+UNIV_INLINE
+void*
+ib_bh_get(
+/*=======*/
+	ib_bh_t*	ib_bh,			/*!< in: instance */
+	ulint		i)			/*!< in: index */
+{
+	byte*		ptr = (byte*) (ib_bh + 1);
+
+	ut_a(i < ib_bh_size(ib_bh));
+
+	return(ptr + (ib_bh->sizeof_elem * i));
+}
+
+/**********************************************************************//**
+Copy an element to the binary heap.
+@return pointer to copied element */
+UNIV_INLINE
+void*
+ib_bh_set(
+/*======*/
+	ib_bh_t*	ib_bh,			/*!< in/out: instance */
+	ulint		i,			/*!< in: index */
+	const void*	elem)			/*!< in: element to add */
+{
+	void*		ptr = ib_bh_get(ib_bh, i);
+
+	ut_memcpy(ptr, elem, ib_bh->sizeof_elem);
+
+	return(ptr);
+}
+
+/**********************************************************************//**
+Return the first element from the binary heap.
+@return pointer to first element or NULL if empty. */
+UNIV_INLINE
+void*
+ib_bh_first(
+/*========*/
+	ib_bh_t*	ib_bh)			/*!< in: instance */
+{
+	return(ib_bh_is_empty(ib_bh) ? NULL : ib_bh_get(ib_bh, 0));
+}
+
+/**********************************************************************//**
+Return the last element from the binary heap.
+@return pointer to last element or NULL if empty. */
+UNIV_INLINE
+void*
+ib_bh_last(
+/*========*/
+	ib_bh_t*	ib_bh)			/*!< in/out: instance */
+{
+	return(ib_bh_is_empty(ib_bh)
+		? NULL
+		: ib_bh_get(ib_bh, ib_bh_size(ib_bh) - 1));
+}
+

storage/innobase/srv/srv0srv.c

 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, 2009 Google Inc.
 Copyright (c) 2009, Percona Inc.
 
@@ -269,9 +269,12 @@ UNIV_INTERN ulong	srv_max_buf_pool_modified_pct	= 75;
 /* the number of purge threads to use from the worker pool (currently 0 or 1).*/
 UNIV_INTERN ulong srv_n_purge_threads = 0;
 
-/* the number of records to purge in one batch */
+/* the number of pages to purge in one batch */
 UNIV_INTERN ulong srv_purge_batch_size = 20;
 
+/* the number of rollback segments to use */
+UNIV_INTERN ulong srv_rollback_segments = TRX_SYS_N_RSEGS;
+
 /* variable counts amount of data read in total (in bytes) */
 UNIV_INTERN ulint srv_data_read = 0;
 
@@ -3096,6 +3099,7 @@ srv_purge_thread(
 						required by os_thread_create */
 {
 	srv_slot_t*	slot;
+	ulint		retries = 0;
 	ulint		slot_no = ULINT_UNDEFINED;
 	ulint		n_total_purged = ULINT_UNDEFINED;
 
@@ -3122,7 +3126,7 @@ srv_purge_thread(
 
 	while (srv_shutdown_state != SRV_SHUTDOWN_EXIT_THREADS) {
 
-		ulint	n_pages_purged;
+		ulint	n_pages_purged = 0;
 
 		/* If there are very few records to purge or the last
 		purge didn't purge any records then wait for activity.
@@ -3130,7 +3134,8 @@ srv_purge_thread(
 		because in the worst case we will end up waiting for
 		the next purge event. */
 		if (trx_sys->rseg_history_len < srv_purge_batch_size
-		    || n_total_purged == 0) {
+		    || (n_total_purged == 0
+			&& retries >= TRX_SYS_N_RSEGS)) {
 
 			os_event_t	event;
 
@@ -3141,6 +3146,8 @@ srv_purge_thread(
 			mutex_exit(&kernel_mutex);
 
 			os_event_wait(event);
+
+			retries = 0;
 		}
 
 		/* Check for shutdown and whether we should do purge at all. */
@@ -3151,7 +3158,12 @@ srv_purge_thread(
 			break;
 		}
 
+		if (n_total_purged == 0 && retries <= TRX_SYS_N_RSEGS) {
+			++retries;
+		} else if (n_total_purged > 0) {
+			retries = 0;
 			n_total_purged = 0;
+		}
 
 		/* Purge until there are no more records to purge and there is
 		no change in configuration or server state. */

storage/innobase/sync/sync0sync.c

 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Innobase Oy. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -1172,7 +1172,7 @@ sync_thread_add_level(
 	case SYNC_RSEG:
 	case SYNC_TRX_UNDO:
 	case SYNC_PURGE_LATCH:
-	case SYNC_PURGE_SYS:
+	case SYNC_PURGE_QUEUE:
 	case SYNC_DICT_AUTOINC_MUTEX:
 	case SYNC_DICT_OPERATION:
 	case SYNC_DICT_HEADER:
@@ -1239,10 +1239,16 @@ sync_thread_add_level(
 		     || sync_thread_levels_g(array, SYNC_FSP, TRUE));
 		break;
 	case SYNC_TRX_UNDO_PAGE:
+		/* Purge is allowed to read in as many UNDO pages as it likes,
+		there was a bogus rule here earlier that forced the caller to
+		acquire the purge_sys_t::mutex. The purge mutex did not really
+		protect anything because it was only ever acquired by the
+		single purge thread. The purge thread can read the UNDO pages
+		without any covering mutex. */
+
 		ut_a(sync_thread_levels_contain(array, SYNC_TRX_UNDO)
 		     || sync_thread_levels_contain(array, SYNC_RSEG)
-		     || sync_thread_levels_contain(array, SYNC_PURGE_SYS)
-		     || sync_thread_levels_g(array, SYNC_TRX_UNDO_PAGE, TRUE));
+		     || sync_thread_levels_g(array, level - 1, TRUE));
 		break;
 	case SYNC_RSEG_HEADER:
 		ut_a(sync_thread_levels_contain(array, SYNC_RSEG));

storage/innobase/trx/trx0rseg.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Oracle Corpn. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -50,11 +50,11 @@ trx_rseg_get_on_id(
 {
 	trx_rseg_t*	rseg;
 
-	rseg = UT_LIST_GET_FIRST(trx_sys->rseg_list);
+	ut_a(id < TRX_SYS_N_RSEGS);
 
-	while (rseg && rseg->id != id) {
-		rseg = UT_LIST_GET_NEXT(rseg_list, rseg);
-	}
+	rseg = trx_sys->rseg_array[id];
+
+	ut_a(rseg == NULL || id == rseg->id);
 
 	return(rseg);
 }
@@ -182,10 +182,13 @@ trx_rseg_t*
 trx_rseg_mem_create(
 /*================*/
 	ulint		id,		/*!< in: rollback segment id */
-	ulint	space,		/*!< in: space where the segment placed */
+	ulint		space,		/*!< in: space where the segment
+					placed */
 	ulint		zip_size,	/*!< in: compressed page size in bytes
 					or 0 for uncompressed pages */
-	ulint	page_no,	/*!< in: page number of the segment header */
+	ulint		page_no,	/*!< in: page number of the segment
+					header */
+	ib_bh_t*	ib_bh,		/*!< in/out: rseg queue */
 	mtr_t*		mtr)		/*!< in: mtr */
 {
 	ulint		len;
@@ -225,6 +228,9 @@ trx_rseg_mem_create(
 
 	len = flst_get_len(rseg_header + TRX_RSEG_HISTORY, mtr);
 	if (len > 0) {
+		const void*	ptr;
+		rseg_queue_t	rseg_queue;
+
 		trx_sys->rseg_history_len += len;
 
 		node_addr = trx_purge_get_log_from_hist(
@@ -240,6 +246,17 @@ trx_rseg_mem_create(
 			undo_log_hdr + TRX_UNDO_TRX_NO);
 		rseg->last_del_marks = mtr_read_ulint(
 			undo_log_hdr + TRX_UNDO_DEL_MARKS, MLOG_2BYTES, mtr);
+
+		rseg_queue.rseg = rseg;
+		rseg_queue.trx_no = rseg->last_trx_no;
+
+		if (rseg->last_page_no != FIL_NULL) {
+			/* There is no need to cover this operation by the purge
+			mutex because we are still bootstrapping. */
+
+			ptr = ib_bh_push(ib_bh, &rseg_queue);
+			ut_a(ptr != NULL);
+		}
 	} else {
 		rseg->last_page_no = FIL_NULL;
 	}
@@ -255,6 +272,7 @@ void
 trx_rseg_create_instance(
 /*=====================*/
 	trx_sysf_t*	sys_header,	/*!< in: trx system header */
+	ib_bh_t*	ib_bh,		/*!< in/out: rseg queue */
 	mtr_t*		mtr)		/*!< in: mtr */
 {
 	ulint		i;
@@ -278,7 +296,7 @@ trx_rseg_create_instance(
 			zip_size = space ? fil_space_get_zip_size(space) : 0;
 
 			rseg = trx_rseg_mem_create(
-				i, space, zip_size, page_no, mtr);
+				i, space, zip_size, page_no, ib_bh, mtr);
 
 			ut_a(rseg->id == i);
 		}
@@ -327,7 +345,8 @@ trx_rseg_create(void)
 		zip_size = space ? fil_space_get_zip_size(space) : 0;
 
 		rseg = trx_rseg_mem_create(
-			slot_no, space, zip_size, page_no, &mtr);
+			slot_no, space, zip_size, page_no,
+			purge_sys->ib_bh, &mtr);
 	}
 
 	mutex_exit(&kernel_mutex);
@@ -342,13 +361,14 @@ UNIV_INTERN
 void
 trx_rseg_list_and_array_init(
 /*=========================*/
-	trx_sysf_t*	sys_header,	/* in: trx system header */
-	mtr_t*		mtr)		/* in: mtr */
+	trx_sysf_t*	sys_header,	/*!< in: trx system header */
+	ib_bh_t*	ib_bh,		/*!< in: rseg queue */
+	mtr_t*		mtr)		/*!< in: mtr */
 {
 	UT_LIST_INIT(trx_sys->rseg_list);
 
 	trx_sys->rseg_history_len = 0;
 
-	trx_rseg_create_instance(sys_header, mtr);
+	trx_rseg_create_instance(sys_header, ib_bh, mtr);
 }
 

storage/innobase/trx/trx0sys.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Innobase Oy. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -945,6 +945,31 @@ trx_sysf_create(
 	mutex_exit(&kernel_mutex);
 }
 
+/*****************************************************************//**
+Compare two trx_rseg_t instances on last_trx_no. */
+static
+int
+trx_rseg_compare_last_trx_no(
+/*=========================*/
+	const void*	p1,		/*!< in: elem to compare */
+	const void*	p2)		/*!< in: elem to compare */
+{
+	ib_int64_t	cmp;
+
+	const rseg_queue_t*	rseg_q1 = (const rseg_queue_t*) p1;
+	const rseg_queue_t*	rseg_q2 = (const rseg_queue_t*) p2;
+
+	cmp = rseg_q1->trx_no - rseg_q2->trx_no;
+
+	if (cmp < 0) {
+		return(-1);
+	} else if (cmp > 0) {
+		return(1);
+	}
+
+	return(0);
+}
+
 /*****************************************************************//**
 Creates and initializes the central memory structures for the transaction
 system. This is called when the database is started. */
@@ -958,6 +983,7 @@ trx_sys_init_at_db_start(void)
 	const char*	unit		= "";
 	trx_t*		trx;
 	mtr_t		mtr;
+	ib_bh_t*	ib_bh;
 
 	mtr_start(&mtr);
 
@@ -965,11 +991,19 @@ trx_sys_init_at_db_start(void)
 
 	mutex_enter(&kernel_mutex);
 
-	trx_sys = mem_alloc(sizeof(trx_sys_t));
+	/* We create the min binary heap here and pass ownership to
+	purge when we init the purge sub-system. Purge is responsible
+	for freeing the binary heap. */
+
+	ib_bh = ib_bh_create(
+		trx_rseg_compare_last_trx_no,
+		sizeof(rseg_queue_t), TRX_SYS_N_RSEGS);
+
+	trx_sys = mem_zalloc(sizeof(*trx_sys));
 
 	sys_header = trx_sysf_get(&mtr);
 
-	trx_rseg_list_and_array_init(sys_header, &mtr);
+	trx_rseg_list_and_array_init(sys_header, ib_bh, &mtr);
 
 	trx_sys->latest_rseg = UT_LIST_GET_FIRST(trx_sys->rseg_list);
 
@@ -1023,7 +1057,8 @@ trx_sys_init_at_db_start(void)
 
 	UT_LIST_INIT(trx_sys->view_list);
 
-	trx_purge_sys_create();
+	/* Transfer ownership to purge. */
+	trx_purge_sys_create(ib_bh);
 
 	mutex_exit(&kernel_mutex);
 

storage/innobase/trx/trx0trx.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Innobase Oy. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -42,6 +42,7 @@ Created 3/26/1996 Heikki Tuuri
 #include "btr0sea.h"
 #include "os0proc.h"
 #include "trx0xa.h"
+#include "trx0purge.h"
 #include "ha_prototypes.h"
 
 /** Dummy session used currently in MySQL interface */
@@ -604,36 +605,26 @@ trx_lists_init_at_db_start(void)
 
 /******************************************************************//**
 Assigns a rollback segment to a transaction in a round-robin fashion.
-Skips the SYSTEM rollback segment if another is available.
-@return	assigned rollback segment id */
+@return	assigned rollback segment instance */
 UNIV_INLINE
-ulint
-trx_assign_rseg(void)
-/*=================*/
+trx_rseg_t*
+trx_assign_rseg(
+/*============*/
+	ulint	max_undo_logs)	/*!< in: maximum number of UNDO logs to use */
 {
 	trx_rseg_t*	rseg = trx_sys->latest_rseg;
 
 	ut_ad(mutex_own(&kernel_mutex));
-loop:
-	/* Get next rseg in a round-robin fashion */
 
 	rseg = UT_LIST_GET_NEXT(rseg_list, rseg);
 
-	if (rseg == NULL) {
+	if (rseg == NULL || rseg->id == max_undo_logs - 1) {
 		rseg = UT_LIST_GET_FIRST(trx_sys->rseg_list);
 	}
 
-	/* If it is the SYSTEM rollback segment, and there exist others, skip
-	it */
-
-	if ((rseg->id == TRX_SYS_SYSTEM_RSEG_ID)
-	    && (UT_LIST_GET_LEN(trx_sys->rseg_list) > 1)) {
-		goto loop;
-	}
-
 	trx_sys->latest_rseg = rseg;
 
-	return(rseg->id);
+	return(rseg);
 }
 
 /****************************************************************//**
@@ -663,12 +654,9 @@ trx_start_low(
 
 	ut_ad(trx->conc_state != TRX_ACTIVE);
 
-	if (rseg_id == ULINT_UNDEFINED) {
+	ut_a(rseg_id == ULINT_UNDEFINED);
 
-		rseg_id = trx_assign_rseg();
-	}
-
-	rseg = trx_sys_get_nth_rseg(trx_sys, rseg_id);
+	rseg = trx_assign_rseg(srv_rollback_segments);
 
 	trx->id = trx_sys_get_new_trx_id();
 
@@ -719,66 +707,111 @@ trx_start(
 }
 
 /****************************************************************//**
-Commits a transaction. */
-UNIV_INTERN
+Set the transaction serialisation number. */
+static
 void
-trx_commit_off_kernel(
-/*==================*/
+trx_serialisation_number_get(
+/*=========================*/
 	trx_t*		trx)	/*!< in: transaction */
 {
-	page_t*		update_hdr_page;
-	ib_uint64_t	lsn		= 0;
 	trx_rseg_t*	rseg;
-	trx_undo_t*	undo;
-	mtr_t		mtr;
 
-	ut_ad(mutex_own(&kernel_mutex));
+	rseg = trx->rseg;
 
-	trx->must_flush_log_later = FALSE;
+	ut_ad(mutex_own(&rseg->mutex));
 
-	rseg = trx->rseg;
+	mutex_enter(&kernel_mutex);
 
-	if (trx->insert_undo != NULL || trx->update_undo != NULL) {
+	trx->no = trx_sys_get_new_trx_id();
+
+	/* If the rollack segment is not empty then the
+	new trx_t::no can't be less than any trx_t::no
+	already in the rollback segment. User threads only
+	produce events when a rollback segment is empty. */
+
+	if (rseg->last_page_no == FIL_NULL) {
+		void*		ptr;
+		rseg_queue_t	rseg_queue;
+
+		rseg_queue.rseg = rseg;
+		rseg_queue.trx_no = trx->no;
+
+		mutex_enter(&purge_sys->bh_mutex);
+
+		/* This is to reduce the pressure on the kernel mutex,
+		though in reality it should make very little (read no)
+		difference because this code path is only taken when the
+		rbs is empty. */
+
+		mutex_exit(&kernel_mutex);
 
+		ptr = ib_bh_push(purge_sys->ib_bh, &rseg_queue);
+		ut_a(ptr);
+
+		mutex_exit(&purge_sys->bh_mutex);
+	} else {
 		mutex_exit(&kernel_mutex);
+	}
+}
+
+/****************************************************************//**
+Assign the transaction its history serialisation number and write the
+update UNDO log record to the assigned rollback segment.
+@return the LSN of the UNDO log write. */
+static
+ib_uint64_t
+trx_write_serialisation_history(
+/*============================*/
+	trx_t*		trx)	/*!< in: transaction */
+{
+	mtr_t		mtr;
+	trx_rseg_t*	rseg;
+
+	ut_ad(!mutex_own(&kernel_mutex));
+
+	rseg = trx->rseg;
 
 	mtr_start(&mtr);
 
 	/* Change the undo log segment states from TRX_UNDO_ACTIVE
 	to some other state: these modifications to the file data
 	structure define the transaction as committed in the file
-		based world, at the serialization point of the log sequence
+	based domain, at the serialization point of the log sequence
 	number lsn obtained below. */
 
-		mutex_enter(&(rseg->mutex));
+	if (trx->update_undo != NULL) {
+		page_t*		undo_hdr_page;
+		trx_undo_t*	undo = trx->update_undo;
 
-		if (trx->insert_undo != NULL) {
-			trx_undo_set_state_at_finish(trx->insert_undo, &mtr);
-		}
+		/* We have to hold the rseg mutex because update
+		log headers have to be put to the history list in the
+		(serialisation) order of the UNDO trx number. This is
+		required for the purge in-memory data structures too. */
 
-		undo = trx->update_undo;
+		mutex_enter(&rseg->mutex);
 
-		if (undo) {
-			mutex_enter(&kernel_mutex);
-			trx->no = trx_sys_get_new_trx_no();
-			mutex_exit(&kernel_mutex);
+		/* Assign the transaction serialisation number and also
+		update the purge min binary heap if this is the first
+		UNDO log being written to the assigned rollback segment. */
+
+		trx_serialisation_number_get(trx);
 
 		/* It is not necessary to obtain trx->undo_mutex here
 		because only a single OS thread is allowed to do the
 		transaction commit for this transaction. */
 
-			update_hdr_page = trx_undo_set_state_at_finish(
-				undo, &mtr);
+		undo_hdr_page = trx_undo_set_state_at_finish(undo, &mtr);
 
-			/* We have to do the cleanup for the update log while
-			holding the rseg mutex because update log headers
-			have to be put to the history list in the order of
-			the trx number. */
+		trx_undo_update_cleanup(trx, undo_hdr_page, &mtr);
+	} else {
+		mutex_enter(&rseg->mutex);
+	}
 
-			trx_undo_update_cleanup(trx, update_hdr_page, &mtr);
+	if (trx->insert_undo != NULL) {
+		trx_undo_set_state_at_finish(trx->insert_undo, &mtr);
 	}
 
-		mutex_exit(&(rseg->mutex));
+	mutex_exit(&rseg->mutex);
 
 	/* Update the latest MySQL binlog name and offset info
 	in trx sys header if MySQL binlogging is on or the database
@@ -786,10 +819,12 @@ trx_commit_off_kernel(
 
 	if (trx->mysql_log_file_name
 	    && trx->mysql_log_file_name[0] != '\0') {
+
 		trx_sys_update_mysql_binlog_offset(
 			trx->mysql_log_file_name,
 			trx->mysql_log_offset,
 			TRX_SYS_MYSQL_LOG_INFO, &mtr);
+
 		trx->mysql_log_file_name = NULL;
 	}
 
@@ -813,13 +848,38 @@ trx_commit_off_kernel(
 	/*--------------*/
 	mtr_commit(&mtr);
 	/*--------------*/
-		lsn = mtr.end_lsn;
+
+	return(mtr.end_lsn);
+}
+
+/****************************************************************//**
+Commits a transaction. */
+UNIV_INTERN
+void
+trx_commit_off_kernel(
+/*==================*/
+	trx_t*	trx)	/*!< in: transaction */
+{
+	ib_uint64_t	lsn;
+
+	ut_ad(mutex_own(&kernel_mutex));
+
+	trx->must_flush_log_later = FALSE;
+
+	/* If the transaction made any updates then we need to write the
+	UNDO logs for the updates to the assigned rollback segment. */
+
+	if (trx->insert_undo != NULL || trx->update_undo != NULL) {
+		mutex_exit(&kernel_mutex);
+
+		lsn = trx_write_serialisation_history(trx);
 
 		mutex_enter(&kernel_mutex);
+	} else {
+		lsn = 0;
 	}
 
-	ut_ad(trx->conc_state == TRX_ACTIVE
-	      || trx->conc_state == TRX_PREPARED);
+	ut_ad(trx->conc_state == TRX_ACTIVE || trx->conc_state == TRX_PREPARED);
 	ut_ad(mutex_own(&kernel_mutex));
 
 	/* The following assignment makes the transaction committed in memory

storage/innobase/ut/ut0ut.c

 /*****************************************************************************
 
-Copyright (c) 1994, 2010, Innobase Oy. All Rights Reserved.
-Copyright (c) 2009, Sun Microsystems, Inc.
-
-Portions of this file contain modifications contributed and copyrighted by
-Sun Microsystems, Inc. Those modifications are gratefully acknowledged and
-are described briefly in the InnoDB documentation. The contributions by
-Sun Microsystems are incorporated with their permission, and subject to the
-conditions contained in the file COPYING.Sun_Microsystems.
+Copyright (c) 2011, Oracle Corpn. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software