既能滿足現有代碼的修改量，又想能提高系統的併發註文量，
經過幾周的摸索，設計以下的結構：
http://blog.chinaunix.net/photo/33594_081104223443.jpg
screen.width*0.7) {this.resized=true; this.width=screen.width*0.7; this.alt='Click here to open new window/nCTRL+Mouse wheel to zoom in/out';}" οnmοuseοver="if(this.width>screen.width*0.7) {this.resized=true; this.width=screen.width*0.7; this.style.cursor='hand'; this.alt='Click here to open new window/nCTRL+Mouse wheel to zoom in/out';}" οnclick="if(!this.resized) {return true;} else {window.open(this.src);}" onmousewheel="return imgzoom(this);" alt="" border="0">http://blog.chinaunix.net/photo/33594_081104223443.jpg
爲了應對我們的交易系統中非常高的併發註文的能力，
特設計了一下的結構：
特點介紹：
1，採用 MySQL Cluster To MySQL Server 的Replication 結構。
將MySQL Cluster 的併發插入能力 在 Cluster上發揮，將許許多多的
Join 查詢扔到Slave 上進行查詢，也是揚長避短吧，雖然有些延遲，但對此要求不高的
就可以直接在Slave上查詢，
http://bugs.mysql.com/39622
在搭建之初遇到很鬱悶的問題，所幸通過在Cluster的SQL節點上參數：
ndb-log-update-as-write=0
得以解決。否則會出現更新數據從服務器上出現主鍵重複的錯誤。
從服務器上的 mysql.ndb_apply_status 表要重新創建，
修改成 MyISAM 表，並將主鍵去掉，並建立一個索引。
爲了在切換的時候查詢的速度能快些。
2，採用 NDB表 和 INNODB表 放一起的方式，圖中的SQL-4 SQL-5，
兩臺機器屬於SQL節點中的兩個普通節點，但是在SQL-4上同時又
創建了很多的Innodb的表，對Innodb的操作就直接在此上操作，
通過設定SQL-5 上的Replication 信息來將這些 Innodb 的表
內容全部複製到 SQl-5上，作爲SQL-4 的貼身Slave，可以作爲
SQL-5 的真正的替換，而不出現系統中丟失數據的問題。
3，SQL-4,SQL-5 採用MM 的結構，他們間雖然是互爲複製，但
同時在寫的，對應用程序來說只有一臺，這樣不會產生一些不必要
的混亂。
對SQL-5 爲什麼一定要是Cluster中的節點而不是 單獨拿出來
作爲一個普通的Slave 的原因是：
1）系統的設計目標是 SQL-4 宕機後，系統還能正常交易，因爲主要的
交易表都在Cluster中，全部的配置相關表都在應用的緩存中， SQL-4
宕機2-3 分鐘不會影響應用程序。
2）在SQL-4宕機的過程切換過程中，系統的交易數據的寫入是不停止的，
如何保證將這部分數據續寫到衆多的Slave 中去，這就是SQL-5作爲
Cluster中節點的目的。
作爲SQL節點，打開了binlog 日誌並設置爲row複製後，其他 SQL節點
上的寫入的信息就會反映在SQL-5 上的日誌中，並記錄到 mysql.ndb_binlog_index
表中，通過 從服務器上 停頓的時間點的查詢,看 mysql.ndb_apply_status 表
中最大的 epoch 值，到SQL-5 中的 mysql.ndb_binlog_index來找下一個binglog
日誌 和 postion ，注意這點和以前的 Master----->Master-Slave ----> Slave
的切換方式不一樣，和 ( Master --- Master ) -----> Slave 的切換方式也完全
不一樣。
4，雖然MySQL5.1 還沒有 GA , 但是 mysql_5.1.23_ndb_6.2.15 Cluster
已經是一個穩定的版本了，有很多成功的案例，在試用階段系統也會穩定，
因此 各方面最終都同意了這個版本。
5，附上
SQL1 :10.4.6.220,
SQL5 :10.4.6.226,
SQL4 :10.4.6.227,
Slave01 : 10.4.6.228
Manager : 10.4.6.224 這樣結構的所有 my.cnf ,ndbconfig.ini 的配置文件信息：
SQL1 :10.4.6.220,
[MYSQLD]
ndbcluster
ndb-connectstring=10.4.6.224
default-storage-engine=ndbcluster
max_connections=1024
default-character-set=cp932
ndb-log-update-as-write=0
memlock
#ndb-cluster-connection-pool=4
#added 2008 08 20 for ...
#BASE config.....
skip-name-resolve
back_log =350
log-slow-queries = slow.log
long_query_time = 1
table_cache =350
sort_buffer_size = 1M
query_cache_size=512M
query_cache_limit=2M
ndb_force_send=1
engine_condition_pushdown=1
ndb_use_exact_count=0
#1 -- means enable, 0 -- means disable
slave-allow-batching=1
#transaction-isolation = READ-COMMITTED
#add by wangxl ,for "converting HEAP to MyISAM "
max_heap_table_size=64M
tmp_table_size=64M
thread_cache_size=350
#add by liuyf for primary key repeat
auto_increment_increment=4
auto_increment_offset=4
delayed_queue_size=500
bulk_insert_buffer_size=16M
#MySQL slave config ...........
log-bin=mfx220-bin
#binlog_format="STATEMENT"
binlog_format="ROW"
#binlog_format="MIXED"
server-id = 20
SQL5 :10.4.6.226,
[root@mfx226 ~]# cat /etc/my.cnf
[MYSQL_CLUSTER]
ndb-connectstring=10.4.6.224
[MYSQLD]
ndbcluster
ndb-connectstring=10.4.6.224
default-storage-engine=ndbcluster
max_connections=1024
default-character-set=cp932
ndb-log-update-as-write=0
#ndb-cluster-connection-pool=4
memlock
#added 2008 08 20 for ...
#BASE config.....
skip-name-resolve
back_log =350
log-slow-queries = slow.log
long_query_time = 1
table_cache =350
sort_buffer_size = 1M
query_cache_size=512M
query_cache_limit=2M
ndb_force_send=1
engine_condition_pushdown=1
ndb_use_exact_count=0
#1 -- means enable, 0 -- means disable
slave-allow-batching=1
transaction-isolation = READ-COMMITTED
#add by wangxl ,for "converting HEAP to MyISAM "
max_heap_table_size=64M
tmp_table_size=64M
thread_cache_size=350
#add by liuyf for primary key repeat
auto_increment_increment=4
auto_increment_offset=3
#MySQL slave config ...........
log-bin=mfx226-bin
#binlog_format="STATEMENT"
#binlog_format="ROW"
binlog_format="MIXED"
server-id = 26
replicate-do-db = MAIN
master-host = 10.4.6.227
master-user = root
master-password = 123456
replicate-ignore-table = MAIN.JHF_SYS_POSITION_INSERT
replicate-ignore-table = mysql.%
replicate-ignore-table = test.%
relay-log=mfx226-relay-bin
skip-slave-start
SQL4 :10.4.6.227,
[MYSQLD]
ndbcluster
ndb-connectstring=10.4.6.224
default-storage-engine=ndbcluster
max_connections=1024
default-character-set=cp932
ndb-log-update-as-write=0
memlock
#ndb-cluster-connection-pool=4
#added 2008 08 20 for ...
#BASE config.....
skip-name-resolve
back_log =350
log-slow-queries = slow.log
long_query_time = 1
table_cache =350
sort_buffer_size = 1M
query_cache_size=512M
query_cache_limit=2M
ndb_force_send=1
engine_condition_pushdown=1
ndb_use_exact_count=0
#1 -- means enable, 0 -- means disable
slave-allow-batching=1
transaction-isolation = READ-COMMITTED
#add by wangxl ,for "converting HEAP to MyISAM "
max_heap_table_size=64M
tmp_table_size=64M
thread_cache_size=350
#add by liuyf for primary key repeat
auto_increment_increment=4
auto_increment_offset=4
delayed_queue_size=500
bulk_insert_buffer_size=16M
#MySQL slave config ...........
log-bin=mfx227-bin
#binlog_format="STATEMENT"
#binlog_format="ROW"
binlog_format="MIXED"
server-id = 27
Slave01 : 10.4.6.228
[mysqld]
port = 3306
socket = /tmp/mysql.sock
skip-locking
key_buffer = 512M
max_allowed_packet = 1M
table_cache =128
sort_buffer_size =1M
net_buffer_length = 8K
read_buffer_size = 256K
read_rnd_buffer_size = 512K
myisam_sort_buffer_size = 8M
#transaction-isolation = READ-COMMITTED
max_connections=1024
skip-name-resolve
back_log = 300
log-slow-queries = slow.log
long_query_time = 1
server-id = 228
master-host = 10.4.6.227
master-user = root
master-password = 123456
#replicate-do-db = MAIN
#replicate-do-db = mysql
#replicate-ignore-table=mysql.ndb_apply_status
relay-log=mfx222-relay-bin
skip-slave-start
innodb_flush_log_at_trx_commit = 1
innodb_buffer_pool_size =128M
innodb_log_buffer_size = 16M
innodb_log_file_size =8M
innodb_file_per_table
innodb_status_file
innodb_thread_concurrency= 32
query_cache_size=512M
query_cache_limit=2M
[mysqldump]
quick
max_allowed_packet = 16M
[mysql]
no-auto-rehash
# Remove the next comment character if you are not familiar with SQL
#safe-updates
[isamchk]
key_buffer = 20M
sort_buffer_size = 20M
read_buffer = 2M
write_buffer = 2M
#[myisamchk]
#key_buffer = 20M
#sort_buffer_size = 20M
#read_buffer = 2M
#write_buffer = 2M
Manager : 10.4.6.224
[root@mfx224 ~]# cat /etc/ndbconfig.ini
# file "config.ini" - 2 data nodes and 2 SQL nodes
# This file is placed in the startup directory of ndb_mgmd (the management
# server)
# The first MySQL Server can be started from any host. The second can be started
# only on the host mysqld_5.mysql.com
[NDBD DEFAULT]
NoOfReplicas=1
MaxNoOfConcurrentOperations= 100000
MaxNoOfLocalOperations = 110000
#MaxNoOfConcurrentTransactions default 4096,
#MaxNoOfConcurrentTransactions =
# (maximum number of tables accessed in any single transaction + 1)
# * number of cluster SQL nodes
MaxNoOfConcurrentTransactions=40960
#DataMemory= 1280M
#IndexMemory= 256M
DataMemory= 640M
IndexMemory= 128M
StringMemory=25 ##use in ndb62
RedoBuffer=128M
DataDir=/usr/local/mysql/cluster
TimeBetweenWatchDogCheck= 30000
#Default ConcurrentScan.......
MaxNoOfConcurrentScans=256
MaxNoOfLocalScans=256
# Default BatchSize
BatchSizePerLocalScan=64
MaxNoOfOrderedIndexes= 1024
# REDO logs of 16X128M=2G.
NoOfFragmentLogFiles=16
FragmentLogFileSize=128M ## use in ndb62
# Default 1000
MaxNoOfAttributes=80000
# Default 128,limit 1600..
MaxNoOfTables=1500
# Default 64.
MaxNoOfUniqueHashIndexes=3500
#Default is 768 ..
MaxNoOfTriggers=1000
TransactionDeadlockDetectionTimeout=20000
#yunfei test
TransactionInactiveTimeout=16000
#This parameter defines the interval between synchronisation epochs
#for MySQL Cluster Replication. The default value is 100 milliseconds.
#TimeBetweenEpochs is part of the implementation of micro-GCPs,
#which can be used to improve the performance of MySQL Cluster Replication.
TimeBetweenEpochs=20 ## use in ndb62
BackupDataDir=/usr/local/mysql/BACKUP
#yunfei test
LockPagesInMainMemory=1
[ndb_mgmd default]
DataDir=/usr/local/mysql/cluster
[TCP default]
SendBufferMemory=2M
ReceiveBufferMemory=2M
[NDB_MGMD]
Hostname= 10.4.6.224
logDestination=FILE:filename=cluster.log,maxsize=100000,maxfiles=6
[NDBD]
HostName= 10.4.6.223
[MYSQLD default]
BatchByteSize=64k
BatchSize=128
[MYSQLD]
HostName= 10.4.6.220
[MYSQLD]
HostName= 10.4.6.226
[MYSQLD]
HostName= 10.4.6.227
[MYSQLD]
[MYSQLD]
完 。。。
本文來自ChinaUnix博客，如果查看原文請點：http://blog.chinaunix.net/u1/33594/showart_1360118.html