TimesTen 应用层数据库缓存学习：17. 全局数据缓存(cache grid)的高可用性

Posted 2020-07-25 dingdingfish

概述

本文有两个目的：
1. 介绍TimesTen Global Cache Grid的高可用性
2. 给出了一个简单的建立和清理Global Cache Grid的过程，前面已经有一篇文章: TimesTen 应用层数据库缓存学习：13. 全局数据缓存(cache grid)，但那个Cache Group太复杂

建立一个简单的Global Cache Grid

首先建立两个TimesTen instance，一个为tt1122, 一个为ttnew。
之所以建立两个实例，是为了停instance方便。现在还没有找到一个方法来停DB。

详细的过程如下：
在tt1122实例上，建立cachedb1的DSN

[cachedb1]
Driver=/home/oracle/TimesTen/tt1122/lib/libtten.so
DataStore=/home/oracle/TimesTen/tt1122/info/DemoDataStore/cachedb1
PermSize=32
TempSize=64
LogFileSize=32
LogBufMB=32
DatabaseCharacterSet=AL32UTF8
OracleNetServiceName=ttorcl

在ttnew实例上，建立cachedb2的DSN

[cachedb2]
Driver=/home/oracle/TimesTen/tt1122/lib/libtten.so
DataStore=/home/oracle/TimesTen/tt1122/info/DemoDataStore/cachedb2
PermSize=32
TempSize=64
LogFileSize=32
LogBufMB=32
DatabaseCharacterSet=AL32UTF8
OracleNetServiceName=ttorcl

建立cachedb1中的schema用户

$ ttisql -v1 cachedb1
Command> set prompt 'cachedb1> '
cachedb1> create user tthr identified by tthr;
User created.
cachedb1> grant admin to tthr;

在cachedb1和cachedb2中设置cacheadm的用户名和口令

$ ttadmin -connstr "dsn=cachedb1;uid=tthr;pwd=tthr;oraclepwd=tthr" -cacheUidPwdSet -cacheUid cacheadm -cachePwd oracle

cachedb1上创建cache grid

$ ttisql -v1 -e "set prompt 'cachedb1> '" "dsn=cachedb1;uid=tthr;pwd=tthr;oraclepwd=oracle" 
cachedb1> call ttGridCreate ('samplegrid');
cachedb1> call ttGridInfo;
< SAMPLEGRID, CACHEADM, Linux x86-64, 64-bit, 11, 2, 2 >
cachedb1> call ttGridNameSet ('samplegrid');

启动cachedb1上的cache agent

ttadmin -cacheStart cachedb1

创建Oracle中的Schema:

$ sqlplus tthr/oracle@ttorcl
create table a(id int, name varchar(32), primary key(id));

SQL> grant select, insert, delete, update on a to cacheadm;

SQL> insert into a values(1, 'beijing');
SQL> insert into a values(2, 'shanghai');
SQL> insert into a values(3, 'guangzhou');

cachedb1上创建Global Dynamic Asynchronous Writethrough Cache Group

CREATE DYNAMIC ASYNCHRONOUS WRITETHROUGH GLOBAL CACHE GROUP "G_AWT" 
 FROM
  "TTHR"."A" (
    "ID"   NUMBER(38)        NOT NULL,
    "NAME" VARCHAR2(32 BYTE),
    PRIMARY KEY("ID")
  )

cachedb1上启动replication agent:

ttadmin -repStart cachedb1

将cachedb1 attach到cache grid

$ ttisql -v1 -e "set prompt 'cachedb1> '" "dsn=cachedb1;uid=tthr;pwd=tthr" 
cachedb1> call ttGridNodeStatus;
cachedb1> call ttGridAttach (1,'cachedb1','localhost',9991);
cachedb1> call ttGridNodeStatus;
< SAMPLEGRID, 1, 1, T, localhost, SAMPLEGRID_cachedb1_1, 127.0.0.1, 9991, <NULL>, <NULL>, <NULL>, <NULL>, <NULL> >

确认cachedb1可以访问数据

$  ttisql -v1 -e "set prompt 'cachedb1> '" "dsn=cachedb1;uid=tthr;pwd=tthr;oraclepwd=oracle"
cachedb1> select * from a;
cachedb1> select * from a where id = 1;
< 1, beijing >

建立cachedb2中的schema用户

$ ttisql -v1 -e "set prompt 'cachedb2> '" cachedb2
cachedb2> create user tthr identified by tthr;
User created.
cachedb2> grant admin to tthr;

cachedb2上设置cache admin账户信息，等等…

ttadmin -connstr "dsn=cachedb2;uid=tthr;pwd=tthr;oraclepwd=tthr" -cacheUidPwdSet -cacheUid cacheadm -cachePwd oracle

[oracle@timesten-hol ~]$ ttisql -v1 -e "set prompt 'cachedb2> '" "dsn=cachedb2;uid=tthr;pwd=tthr;oraclepwd=oracle"
cachedb2> call ttGridInfo;
< SAMPLEGRID, CACHEADM, Linux x86-64, 64-bit, 11, 2, 2 >
cachedb2> call ttGridNameSet ('samplegrid');
cachedb2> call ttCacheStart();
cachedb2> CREATE DYNAMIC ASYNCHRONOUS WRITETHROUGH GLOBAL CACHE GROUP "G_AWT" 
        >  FROM
        >   "TTHR"."A" (
        >     "ID"   NUMBER(38)        NOT NULL,
        >     "NAME" VARCHAR2(32 BYTE),
        >     PRIMARY KEY("ID")
        >   );
cachedb2> call ttrepstart();
cachedb2> call ttGridAttach (1,'cachedb2','localhost',9992);
cachedb2> call ttGridNodeStatus;
< SAMPLEGRID, 1, 1, T, localhost, SAMPLEGRID_cachedb1_1, 127.0.0.1, 9991, <NULL>, <NULL>, <NULL>, <NULL>, <NULL> >
< SAMPLEGRID, 2, 1, T, localhost, SAMPLEGRID_cachedb2_2, 127.0.0.1, 9992, <NULL>, <NULL>, <NULL>, <NULL>, <NULL> >

cachedb1> autocommit 0
cachedb1> CALL ttOptSetFlag('GlobalProcessing', 1);
cachedb1> select * from a;
< 2, shanghai >
< 1, beijing >


cachedb1>  SELECT id, TTGRIDUSERASSIGNEDNAME(), TTGRIDMEMBERID() FROM a; 
< 1, cachedb1, 1 >
< 2, cachedb2, 2 >

Cache Grid中某节点异常时的情况

以上即演示，一切正常，下面来看一下异常的情况。

若某一grid member宕机

[oracle@timesten-hol ~]$ ttdaemonadmin -stop
TimesTen Daemon stopped.
[oracle@timesten-hol ~]$ ttstatus cachedb1
ttStatus: Could not connect to the TimesTen daemon.
If the TimesTen daemon is not running, please start it
by running "ttDaemonAdmin -start".
[oracle@timesten-hol ~]$ 

[oracle@timesten-hol ~]$ ttisql -v1 -e "set prompt 'cachedb2> '" "dsn=cachedb2;uid=tthr;pwd=tthr;oraclepwd=oracle"
cachedb2> show autocommit;
autocommit = 1 (ON)
cachedb2> select * from a;
< 2, shanghai >
cachedb2> select * from a where id = 1;
 3333: Time out waiting for a response from member SAMPLEGRID_cachedb2_2 <- 大致一分钟
cachedb2> 

cachedb2> autocommit 0;
cachedb2>  CALL ttOptSetFlag('GlobalProcessing', 1);
cachedb2> select * from a;
好久都没出来！一直都没出来!!!

结论

1. Cache Grid中的每个成员，如果是单节点，此架构是没有弹性(Resilience)的。数据不会重新分布，查询操作会超时或者HANG
2. Cache Grid的每个成员的保护，通常是通过Active Standby Pair来保护的，每一个节点都需要配一个standby，不过这样的架构比较重
3. Cache Grid(Global Cache Group)适用的场景较少，可扩展性方面是有限制的。
4. TimesTen 未来会推出12版，应该可以替代掉现在的Cache Grid功能。