无论是什么样的解x案，一定要牢记我们要解决的问题是什么，切不能将解决Ҏ当做问题本n。具体到q程改进Q不是何种方式的改q，它们所要解决的问题永远只有一个：~短从品想法到可用软g之间的时间周?/strong>。自动化发布正是如此Q如果Y件发布只做一ơ，我们说根本不需要自动化Q但如果三次以上Q那么Y件开发的黄金法则DRY必遵守，让时间真正用到开发当中去?/span>
二?/strong>与发布相关的问题
所谓自动化只不q是原先手工做的工作谦让给机器做，所以自动化之前一定要先清楚与发布相关的问题有哪些Q即使不自动化，q些工作也一个也不能：
- 应用E序如何打包Q?/span> 发布包能否追t到SVN版本P
- 对目标机器环境有什么样的要求？
- 配置信息是否需要根据目标机器信息做_
- 应用E序如何安装和启动？
- 应用E序启动后如何切量Q?/span>
- 应用E序如何升Q?/span> 旧版本程序数据如何迁U？
- 升q程中和l束后如何切量Q?/span>
- 应用E序如何卸蝲Q?/span>
三?/strong>我们的方?/span>
李安的少q?/span>Pi正在狂刷房Q我们的自动化发布方案也要跟上潮：Puppet+CI我们的少q?/span>Pi?/span>
Ø 使用CI自动化打包，q踪每个发布包的SVN版本Q?/span>
Ø 使用Puppet理发布包、目标机器环境、应用程序配|信息以及应用程序线上生命周期；
Ø 使用伽利略系l提供应用程序的命名服务和进行流量切换?/span>
现在应用E序的发布需要两步：CI一键打包?/span>puppet指定应用E序版本SVN提交?/span>
四?/strong>具体Ҏ
具体Ҏ也就是如何解决与发布相关八个问题的过E?/span>
1. 如何安装、升U和卸蝲应用E序
我们使用操作pȝ原生包管理系l来安装、升U和卸蝲应用E序Q我们的应用E序打出RPM二进制包。免安装Q所有机器自带，l色的，有机的?/span>
打包Q?/span>rpm -ba ./team_member-1.spec
安装Q?/span>rpm –ivh team_ member-2.0.1-48.x86_64.rpm
升Q?/span>rpm –U team_ member-2.0.1-49.x86_64.rpm
卸蝲Q?/span>rpm –e team_ member-2.0.1-48.x86_64.rpm
E序升前要停旧版本服务怎么办？旧版本数据要做处理怎么办？RPM已经帮我们料理好q一切，只要写出spec文gQ剩下的交给我们。尽情的插入吧：
2. 如何理目标机器环境和应用配|信?/span>
应用E序已经打好rpm包了Q但q还不够Q应用程序发布到哪台机器上？应用E序对目标机器有什么要求？发布旉要修改哪些配|和参数Q实际发布如何执行，N需要登陆到每台目标机器q行rpm命o吗？
我们使用Puppet来搞定这一切，Puppet是现在应用第一?/span>devops工具Q它通过master/agent的工作模式管理机器。我们通过声明来控制我们的机器辑ֈ目标状态。同Ӟ所?/span>puppet文g全部?/span>SVN里，所有对机器的修改全?/span>codereview和可审计?/span>
如何理应用E序发布到哪台机器上Q在回答q个问题前我们必d应用E序在线上的生命周期再进行一ơ封装?/span>
应用E序TeamMember被我们封装成一?/span>puppet moduleQ配|文件和参数被封装在对应templates?/span>files里，每次发布前都要修攚w|文件和传递不同的参数Q?/span>out了吧Q?/span>puppet帮你传参搞定:
Teammember.conf文g内容Q?/span>
装完成后的效果是这LQ?/span>
最后在理部v?/span>site.pp文g里声明一下，应用E序TeamMember?/span>2146版本p自动部v?/span>10.128.34.141.test.back.shequq台机器上了Q我们后l的工作也就是维护这?/span>site.pp文g了，所有应用程序的部v信息都在SVN被集中管理v来：
登陆到每台目标机器运?/span>rpm命oQ?/span>No!现在TeamMember已经被封装，我们修改完毕site.ppq提交后Q?/span>puppetp动执行命令了Q要不怎么说是自动化呢。（现在puppet默认?/span>agent每半时同步一ơ，但同时支持马上触发执行）?/span>
3. 如何q踪每次发布?/span>SVN版本?/span>
我们使用CIq行应用E序的打包，?/span>build号作为包命名的一部分Q?/span>
4. 如何在发布过E中切换量
q是另外一个很大的话题Q参见伽利略计划?/span>
五?/strong>下一步工?/span>
使用CI环境的自动化部|与自动化测试串联v来，搭徏h个研发流E自动化q_Q?/span>
六?/strong>结
没有银弹Q自动化所做的只是之前手工工作交l计机完成Q需要做的工作一个都不能，此外Q我们还要多做一些封装或脚本工作Q但是，当我们需要重复做q些事情的时候，价值就出现了。我们的目标永远是羃短从产品x到可用Y件之间的旉周期。让旉真正用到开发当中去?/span>
]]>
最开始我们需要验证在用户名和密码都正的情况下，能够正常dpȝQ我们这L写测试代码（以下都是伪代码，使用TestNG和SeleniumQ：
@Test
def should_login_success_with_exist_username_and_correct_password(){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",['user1','1234'],on(page))
assert page.successLogin
}
def should_login_success_with_exist_username_and_correct_password(){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",['user1','1234'],on(page))
assert page.successLogin
}
恩，很不错，q行一下，出现U条。ؓ什么呢Q原来测试数据库里没有用户名为user1的用P好吧Q写个数据库数据初始化脚本。再q行QOKQ绿条！
那么Q接下来我们再增加一个测试，需要覆盖密码错误时不能dpȝ的情况，很快试完成了Q?br />
@Test
def should_login_success_with_exist_username_and_incorrect_password(){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",['user1','4321'],on(page))
assert page.successLogin,false
}
再运行一下测试，l条。好啦，现在可以看下q段代码Q恩Q有些重复，重构一下：
@Test
def should_login_success_with_exist_username_and_correct_password(){
assert login('user1','1234')
}
@Test
def should_login_success_with_exist_username_and_incorrect_password(){
assert login('user1','4321'),false
}
def login(username,password){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",[username,password],on(page))
return page.successLogin
}
def should_login_success_with_exist_username_and_correct_password(){
assert login('user1','1234')
}
@Test
def should_login_success_with_exist_username_and_incorrect_password(){
assert login('user1','4321'),false
}
def login(username,password){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",[username,password],on(page))
return page.successLogin
}
重构完成Q可以看刎ͼ我们的测试方法里现在没有了Q何行为，仅仅是数据！q样让我感觉有点怪，不管了，先用TestNG提供的@dataProvider整理一下：
@Test(dataProvider="testdata")
def testLogin(username,password,expected){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",[username,password],on(page))
assert page.successLogin,expected
}
@DataProvider(name="testdata")
def Object[][] dataForLogin(){
def data=new Object[2][]
data[0]=['user1','1234',true] as Object[]
data[1]=['user1','4321',false] as Object[]
}
def testLogin(username,password,expected){
LoginPage page = user.open(LoginPage,"/login.html")
user.perform("login",[username,password],on(page))
assert page.successLogin,expected
}
@DataProvider(name="testdata")
def Object[][] dataForLogin(){
def data=new Object[2][]
data[0]=['user1','1234',true] as Object[]
data[1]=['user1','4321',false] as Object[]
}
试Ҏ只剩下了一个！如果要测试不存在的用户不能登录系l呢Q很单，增加数据卛_Q?br />
@DataProvider(name="testdata")
def Object[][] dataForLogin(){
def data=new Object[2][]
data[0]=['user1','1234',true] as Object[]
data[1]=['user1','4321',false] as Object[]
data[1]=['inexistuser','1234',false] as Object[]
}
def Object[][] dataForLogin(){
def data=new Object[2][]
data[0]=['user1','1234',true] as Object[]
data[1]=['user1','4321',false] as Object[]
data[1]=['inexistuser','1234',false] as Object[]
}
在我们的试Ҏ里，试数据和测试的行ؓq行了完全的分离。从pȝ的功能来_功能一旦实玎ͼ那么是一个黑盒，我们只要提供数据卛_q行试Q这个数据包括两部分Q输入和期待的输出。我开始暗自嘀咕：N我以前那么多的洋z得意测试方法很多都是不需要的吗？q些方式Z么会存在呢？恩，惌v来了Q这些方法是Z覆盖功能的各个\径的Q是提高试覆盖率的。那么ؓ什么会产生q么多的试Ҏ呢？哦，在这些测试方法里Q测试数据和试行ؓ是耦合在一LQ?br />
我了个懒腰Q突然想Q这下好了，我已l封装好了功能行为，QA惛_加测试用例只需要自己增加数据就可以了，嘿嘿Q爽啊。说ҎQQA到。QA mm_你的某个功能实现有问题。我瞅了一|_不可能啊Q这是dev面对bug的第一反应Q，俺有试的，持箋集成一直是l条的。QA mm_在你的开发环境下试是没有问题的Q但是在QA环境Q因为数据库变了Q数据变了，应用服务器变了，所以会有些问题。我极不情愿的登录到QA环境Q一试Q还真是Q郁闗?br />
怎么办？修复完BUGQ第一反应是自动化测试能不能跑在QA环境呢？一般情况下Q这些测试需要干净的测试环境，我们会制造很多的试数据Q可是在QA环境下，QA有她自己的测试数据，q些数据都不存在了哈。恩Q看看刚才的试代码Q哈Q就用QA的数据也可以啊，心情愉悦的改下：
@DataProvider(name="testdata")
def Object[][] dataForLogin(){
def data=new Object[2][]
data[0]=['hrong','1234',true] as Object[]
data[1]=['hrong','4321',false] as Object[]
data[1]=['rhao','1234',false] as Object[]
}
def Object[][] dataForLogin(){
def data=new Object[2][]
data[0]=['hrong','1234',true] as Object[]
data[1]=['hrong','4321',false] as Object[]
data[1]=['rhao','1234',false] as Object[]
}
OK,完成Qؓ了该试既能在开发测试环境运行又能在QA环境下运行，我们可以引入一个环境变量，测试数据扔到文仉Q通过环境变量来加载不同的试数据Q测试文Ӟ?br />
好吧Q喝点东西（甲流很厉宻I喝板蓝根好了Q，ȝ一下：
数据驱动试Q?strong>试数据与测试行为分，通过数据来驱动测试?/strong>
好处Q?strong>在对试行ؓ装好的情况下，QA mm能够自己通过数据修改自动化测试；
自动化测试能够运行在多个环境下（开发环境、QA环境、品环境）
试的可L?br /> 试Ҏ大量压羃
适用范围Q?strong>功能试Qselenium试Q?br /> 通过环境准备试数据Q非试用例自己准备数据Q?/strong>
可能存在的问题：比一般的试~写困难Q特别是在静态语a?/strong>
最后：该文章的思考来自于徐昊在团队内部的相应Session.
]]>
]]>
目前我们要进行持l集成的对象是一个有着100人左右的开发团队，他们开发着一套很庞大的系l。整个开发团队划分ؓ多个开发小l进行协同开发，每个开发小l负?-3个模块的开发，实际q里的模块已l相当于一个中型pȝ。各模块所有的c都通过eclipse整体~译在一P直接攄在WEB-INF/classes下。本地是无法启动整个pȝ的，需要耗费大量的资源?br />
二、碰到的问题
在了解具体情况之前，我们最初的x是ؓ整个产品做一个持l集成，但是很快发现这一x存在很多的问题：
1、整个品每ơ构建的旉会很长，q个旉包括代码的编译、启动WeblogicQ完成自动化试Q同时对服务器的g要求非常?br /> 2、因为构建时间长Q所以如果本地构建通过后再提交会严重媄响开发效率，况且本地的硬件条件很可能启动不了
3、如果本C构徏提交Q则׃开发hC多，构徏会非怸E_Q会l常处于p|状态。而构建失败会D后箋提交的阻塞?br /> 4、作Z?00人的开发团队，代码提交会引发频J的服务器构建，服务器无法负担?br />
同时作ؓ客户Q他们有q样一U想法：敏捷开发是好的Q但是不适合于大的项目和大的团队?br />
最重要的问题集中在两个斚wQ?br /> 1、启动整个品过于重量Q不包括自动化测试的情况下已l如此）
2、如何不影响开发h员的频繁提交
三、我们的x
我们现在的想法是做多阶段的持l集成（multi-stage CIQ?br />
可以参考这?a target="_blank">http://www.ddj.com/development-tools/212201506
具体而言Q?br /> 1、各个开发小l内做小l内的持l集?br /> 2、开发小l间集成做整个品的持箋集成
大概Q?br /> 1、每个开发小l一个分支，整个产品一条主U?br /> 2、在组分支上搭建持l集成环境，组内的开发向该分支上提交Q各个小l可以ƈ发开发，互不影响
3、小l完成一个完整的功能后，从主U更新合q代码，本地构徏通过Q提交，触发整个产品的持l集?br />
Zɞ组内持l集成构建加快，组内尽量划分清楚对其他模块的依赖，不必要的模块Q这里的模块包括基础模块Q例如工作流模块Q不必加载?br /> 同时推荐轻量U的web服务器例如Tomcat来完成小l内的测试环境。需要启动weblogic的情冉|功能依赖q多的情况下Q徏议在产品持箋集成时进行测试?br /> 同时保留原有的启动单独测试服务器q行手工试的习惯?br />
]]>
~存要用统一的缓存服务器Q集中式~存?br /> 原先的实现采用ehcache?br /> 在spring里的配置Q以资源~存ZQ?/p>
<!-- EhCache Manager -->
<bean id="cacheManager" class="org.springframework.cache.ehcache.EhCacheManagerFactoryBean">
<property name="configLocation">
<value>classpath:ehcache.xml</value>
</property>
</bean>
<bean id="resourceCacheBackend"
class="org.springframework.cache.ehcache.EhCacheFactoryBean">
<property name="cacheManager" ref="cacheManager"/>
<property name="cacheName" value="resourceCache"/>
</bean>
<bean id="resourceCache"
class="com.framework.extcomponent.security.authentication.services.acegi.cache.EhCacheBasedResourceCache"
autowire="byName">
<property name="cache" ref="resourceCacheBackend"/>
</bean>
<bean id="cacheManager" class="org.springframework.cache.ehcache.EhCacheManagerFactoryBean">
<property name="configLocation">
<value>classpath:ehcache.xml</value>
</property>
</bean>
<bean id="resourceCacheBackend"
class="org.springframework.cache.ehcache.EhCacheFactoryBean">
<property name="cacheManager" ref="cacheManager"/>
<property name="cacheName" value="resourceCache"/>
</bean>
<bean id="resourceCache"
class="com.framework.extcomponent.security.authentication.services.acegi.cache.EhCacheBasedResourceCache"
autowire="byName">
<property name="cache" ref="resourceCacheBackend"/>
</bean>
cacheManager负责对ehcacheq行理Q初始化、启动、停止?br />
resourceCacheBackend负责实际执行~存操作Qput 、get、remove?br />
resourceCache实现h业务语义的业务应用层面的~存操作Q内部调用resourceCacheBackend操作?br />
现在采用memcached?br />
关于客户端，采用文初装的客LQ地址?a mce_>http://code.google.com/p/memcache-client-forjava/?br />
使用spring的FactoryBeanq行二次装。同理：
memcachedManager负责对memcachedq行理Q初始化、启动、停止?br />
代码Q?/p>
/**
* User: ronghao
* Date: 2008-10-14
* Time: 10:36:30
* 理Memcached 的CacheManager
*/
public class MemcachedCacheManagerFactoryBean implements FactoryBean, InitializingBean, DisposableBean {
protected final Log logger = LogFactory.getLog(getClass());
private ICacheManager<IMemcachedCache> cacheManager;
public Object getObject() throws Exception {
return cacheManager;
}
public Class getObjectType() {
return this.cacheManager.getClass();
}
public boolean isSingleton() {
return true;
}
public void afterPropertiesSet() throws Exception {
logger.info("Initializing Memcached CacheManager");
cacheManager = CacheUtil.getCacheManager(IMemcachedCache.class,
MemcachedCacheManager.class.getName());
cacheManager.start();
}
public void destroy() throws Exception {
logger.info("Shutting down Memcached CacheManager");
cacheManager.stop();
}
}
* User: ronghao
* Date: 2008-10-14
* Time: 10:36:30
* 理Memcached 的CacheManager
*/
public class MemcachedCacheManagerFactoryBean implements FactoryBean, InitializingBean, DisposableBean {
protected final Log logger = LogFactory.getLog(getClass());
private ICacheManager<IMemcachedCache> cacheManager;
public Object getObject() throws Exception {
return cacheManager;
}
public Class getObjectType() {
return this.cacheManager.getClass();
}
public boolean isSingleton() {
return true;
}
public void afterPropertiesSet() throws Exception {
logger.info("Initializing Memcached CacheManager");
cacheManager = CacheUtil.getCacheManager(IMemcachedCache.class,
MemcachedCacheManager.class.getName());
cacheManager.start();
}
public void destroy() throws Exception {
logger.info("Shutting down Memcached CacheManager");
cacheManager.stop();
}
}
配置Q?/p>
<bean id="memcachedManager"
class="com.framework.extcomponent.cache.MemcachedCacheManagerFactoryBean"/>
class="com.framework.extcomponent.cache.MemcachedCacheManagerFactoryBean"/>
resourceCacheBackend负责实际执行~存操作Qput 、get、remove?br /> 代码Q?/p>
/**
* User: ronghao
* Date: 2008-10-14
* Time: 10:37:16
* q回 MemcachedCache
*/
public class MemcachedCacheFactoryBean implements FactoryBean, BeanNameAware, InitializingBean {
protected final Log logger = LogFactory.getLog(getClass());
private ICacheManager<IMemcachedCache> cacheManager;
private String cacheName;
private String beanName;
private IMemcachedCache cache;
public void setCacheManager(ICacheManager<IMemcachedCache> cacheManager) {
this.cacheManager = cacheManager;
}
public void setCacheName(String cacheName) {
this.cacheName = cacheName;
}
public Object getObject() throws Exception {
return cache;
}
public Class getObjectType() {
return this.cache.getClass();
}
public boolean isSingleton() {
return true;
}
public void setBeanName(String name) {
this.beanName=name;
}
public void afterPropertiesSet() throws Exception {
// If no cache name given, use bean name as cache name.
if (this.cacheName == null) {
this.cacheName = this.beanName;
}
cache = cacheManager.getCache(cacheName);
}
}
* User: ronghao
* Date: 2008-10-14
* Time: 10:37:16
* q回 MemcachedCache
*/
public class MemcachedCacheFactoryBean implements FactoryBean, BeanNameAware, InitializingBean {
protected final Log logger = LogFactory.getLog(getClass());
private ICacheManager<IMemcachedCache> cacheManager;
private String cacheName;
private String beanName;
private IMemcachedCache cache;
public void setCacheManager(ICacheManager<IMemcachedCache> cacheManager) {
this.cacheManager = cacheManager;
}
public void setCacheName(String cacheName) {
this.cacheName = cacheName;
}
public Object getObject() throws Exception {
return cache;
}
public Class getObjectType() {
return this.cache.getClass();
}
public boolean isSingleton() {
return true;
}
public void setBeanName(String name) {
this.beanName=name;
}
public void afterPropertiesSet() throws Exception {
// If no cache name given, use bean name as cache name.
if (this.cacheName == null) {
this.cacheName = this.beanName;
}
cache = cacheManager.getCache(cacheName);
}
}
配置Q?/p>
<bean id="resourceCacheBackend"
class="com.framework.extcomponent.cache.MemcachedCacheFactoryBean">
<property name="cacheManager" ref="memcachedManager"/>
<property name="cacheName" value="memcache"/>
</bean>
resourceCache同上Q替换新的实现类MemcachedBasedResourceCache卛_?br />
class="com.framework.extcomponent.cache.MemcachedCacheFactoryBean">
<property name="cacheManager" ref="memcachedManager"/>
<property name="cacheName" value="memcache"/>
</bean>
二?Session失效的处?/strong>
采用memcached作ؓhttpsession的存储，q不直接保存httpsession对象Q自定义SessionMapQSessionMap直接l承HashMapQ保存SessionMap?br />
会话胶粘Q未p|转发的情况下没必要在memcached保存的SessionMap和httpsession之间复制来复制去Q眉来眼厅R?br />
利用memcached计数器保存在Uh数?br />
pȝ权限采用了acegiQ在acegi的拦截器N配置snaFilter
<bean id="filterChainProxy"
class="org.acegisecurity.util.FilterChainProxy">
<property name="filterInvocationDefinitionSource">
<value>
CONVERT_URL_TO_LOWERCASE_BEFORE_COMPARISON
PATTERN_TYPE_APACHE_ANT
/**=snaFilter,httpSessionContextIntegrationFilter,logoutFilter,authenticationProcessingFilter,basicProcessingFilter,securityContextHolderAwareRequestFilter,exceptionTranslationFilter,filterInvocationInterceptor
</value>
</property>
</bean>
class="org.acegisecurity.util.FilterChainProxy">
<property name="filterInvocationDefinitionSource">
<value>
CONVERT_URL_TO_LOWERCASE_BEFORE_COMPARISON
PATTERN_TYPE_APACHE_ANT
/**=snaFilter,httpSessionContextIntegrationFilter,logoutFilter,authenticationProcessingFilter,basicProcessingFilter,securityContextHolderAwareRequestFilter,exceptionTranslationFilter,filterInvocationInterceptor
</value>
</property>
</bean>
注意需要配|在W一个?br />
snaFilter的职责：
1?没有HttpSessionӞ创徏HttpSession;
2?创徏Cookie保存HttpSession id;
3?如果Cookie保存的HttpSession id与当前HttpSession id一_说明是正常请求；
4?如果Cookie保存的HttpSession id与当前HttpSession id不一_说明是失败{发；p|转发的处理：
4.1、根据Cookie保存的HttpSession id从memcached获取SessionMapQ?br />
4.2、SessionMap属性复制到当前HttpSessionQ?br />
4.3、memcached删除SessionMap?br />
5?判断当前hurl是否是登出urlQ是则删除SessionMapQ在Uh数减1.
代码Q?/p>
public void doFilter(ServletRequest servletRequest, ServletResponse servletResponse,
FilterChain filterChain) throws IOException, ServletException {
final HttpServletRequest hrequest = (HttpServletRequest) servletRequest;
final HttpServletResponse hresponse = (HttpServletResponse) servletResponse;
String uri = hrequest.getRequestURI();
logger.debug("开始SNA拦截-----------------" + uri);
HttpSession httpSession = hrequest.getSession();
String sessionId = httpSession.getId();
//如果是登?则直接干掉sessionMap
if (uri.equals(logoutUrl)) {
logger.debug("remove sessionmap:" + sessionId);
//在线人数?
getCache().addOrDecr("userCount",1);
getCache().remove(sessionId);
} else {
String cookiesessionid = getSessionIdFromCookie(hrequest, hresponse);
if (!sessionId.equals(cookiesessionid)) {
createCookie(sessionId, hresponse);
SessionMap sessionMap = getSessionMap(cookiesessionid);
if (sessionMap != null) {
logger.debug("fail over--------sessionid:" + sessionId + "cookiesessionid:" + cookiesessionid);
initialHttpSession(sessionMap, httpSession);
cache.remove(cookiesessionid);
}
}
}
filterChain.doFilter(hrequest, hresponse);
}
FilterChain filterChain) throws IOException, ServletException {
final HttpServletRequest hrequest = (HttpServletRequest) servletRequest;
final HttpServletResponse hresponse = (HttpServletResponse) servletResponse;
String uri = hrequest.getRequestURI();
logger.debug("开始SNA拦截-----------------" + uri);
HttpSession httpSession = hrequest.getSession();
String sessionId = httpSession.getId();
//如果是登?则直接干掉sessionMap
if (uri.equals(logoutUrl)) {
logger.debug("remove sessionmap:" + sessionId);
//在线人数?
getCache().addOrDecr("userCount",1);
getCache().remove(sessionId);
} else {
String cookiesessionid = getSessionIdFromCookie(hrequest, hresponse);
if (!sessionId.equals(cookiesessionid)) {
createCookie(sessionId, hresponse);
SessionMap sessionMap = getSessionMap(cookiesessionid);
if (sessionMap != null) {
logger.debug("fail over--------sessionid:" + sessionId + "cookiesessionid:" + cookiesessionid);
initialHttpSession(sessionMap, httpSession);
cache.remove(cookiesessionid);
}
}
}
filterChain.doFilter(hrequest, hresponse);
}
利用HttpSessionAttributeListener监听httpsession的属性变?同步到memecached中的sessionmap?
public void attributeAdded(HttpSessionBindingEvent event) {
HttpSession httpSession = event.getSession();
String attrName = event.getName();
Object attrValue = event.getValue();
String sessionId = httpSession.getId();
logger.debug("attributeAdded sessionId:" + sessionId + "name:" + attrName + ",value:" + attrValue);
SessionMap sessionMap = getSessionMap(sessionId);
if (sessionMap == null){
//在线人数?
getCache().addOrIncr("userCount",1);
sessionMap = new SessionMap();
}
logger.debug("name:" + attrName + ",value:" + attrValue);
sessionMap.put(attrName, attrValue);
getCache().put(sessionId, sessionMap);
}
public void attributeRemoved(HttpSessionBindingEvent event) {
HttpSession httpSession = event.getSession();
String attrName = event.getName();
String sessionId = httpSession.getId();
logger.debug("attributeRemoved sessionId:" + sessionId + "name:" + attrName);
SessionMap sessionMap = getSessionMap(sessionId);
if (sessionMap != null) {
logger.debug("remove:" + attrName);
sessionMap.remove(attrName);
getCache().put(sessionId, sessionMap);
}
}
public void attributeReplaced(HttpSessionBindingEvent event) {
attributeAdded(event);
}
利用HttpSessionListenerQsessionDestroyed事g时根据sessionid删除memcached里的sessionMapQ如果存在）。不再担心httpsession的过期问题?
HttpSession httpSession = event.getSession();
String attrName = event.getName();
Object attrValue = event.getValue();
String sessionId = httpSession.getId();
logger.debug("attributeAdded sessionId:" + sessionId + "name:" + attrName + ",value:" + attrValue);
SessionMap sessionMap = getSessionMap(sessionId);
if (sessionMap == null){
//在线人数?
getCache().addOrIncr("userCount",1);
sessionMap = new SessionMap();
}
logger.debug("name:" + attrName + ",value:" + attrValue);
sessionMap.put(attrName, attrValue);
getCache().put(sessionId, sessionMap);
}
public void attributeRemoved(HttpSessionBindingEvent event) {
HttpSession httpSession = event.getSession();
String attrName = event.getName();
String sessionId = httpSession.getId();
logger.debug("attributeRemoved sessionId:" + sessionId + "name:" + attrName);
SessionMap sessionMap = getSessionMap(sessionId);
if (sessionMap != null) {
logger.debug("remove:" + attrName);
sessionMap.remove(attrName);
getCache().put(sessionId, sessionMap);
}
}
public void attributeReplaced(HttpSessionBindingEvent event) {
attributeAdded(event);
}
public void sessionDestroyed(HttpSessionEvent event) {
HttpSession httpSession = event.getSession();
String sessionId = httpSession.getId();
logger.debug("session Removed sessionId:" + sessionId);
SessionMap sessionMap = getSessionMap(sessionId);
if (sessionMap != null) {
logger.debug("remove sessionmap:" + sessionId);
//在线人数?
getCache().addOrDecr("userCount",1);
getCache().remove(sessionId);
}
}
HttpSession httpSession = event.getSession();
String sessionId = httpSession.getId();
logger.debug("session Removed sessionId:" + sessionId);
SessionMap sessionMap = getSessionMap(sessionId);
if (sessionMap != null) {
logger.debug("remove sessionmap:" + sessionId);
//在线人数?
getCache().addOrDecr("userCount",1);
getCache().remove(sessionId);
}
}
三?文g保存的处?/strong>
和缓存类|采用集中式的文g服务。对于linuxQ采用nfs。参考文?a mce_>http://linux.vbird.org/linux_server/0330nfs.php#What_NFS_perm。关键在于对权限的分配?br /> 应用E序本n不用修改?/p>
]]>
最开始的做法Q?/strong>
一个实体类叫做WorkitemQ指的是一个工作项或者称ZQ务项
一个DAOcd做WorkitemDao
一个业务逻辑cd做WorkitemManager(或者叫做WorkitemService)
主要看看WorkitemManagerQ因Z要逻辑集中在这?br />
public class WorkitemManager {
private WorkItemDAO workItemDAO;
public void setWorkItemDAO(WorkItemDAO workItemDAO) {
this.workItemDAO = workItemDAO;
}
/**
* 提交工作?br /> * @param workitemId 工作ID
*/
public void commitWorkitem(String workitemId){
WorkItem workitem = workItemDAO.getWorkItem(workitemId);
//当前工作结?/span>
workitem.complete();
int sID = workitem.getSequenceId();
//扑ֈ所对应的节?/span>
InstActivity instActivity=workitem.getInstActivity();
//查找是否存在下一工作?/span>
WorkItem sequenceWorkitem = workItemDAO.findSequenceWorkItem(instActivity.getId(), sID + 1);
//如果不存在则触发节点{
if (sequenceWorkitem == null) {
instActivity.signal();
}
//否则把下一工作Ҏz?/span>
else {
sequenceWorkitem.setExecutive();
}
}
}
private WorkItemDAO workItemDAO;
public void setWorkItemDAO(WorkItemDAO workItemDAO) {
this.workItemDAO = workItemDAO;
}
/**
* 提交工作?br /> * @param workitemId 工作ID
*/
public void commitWorkitem(String workitemId){
WorkItem workitem = workItemDAO.getWorkItem(workitemId);
//当前工作结?/span>
workitem.complete();
int sID = workitem.getSequenceId();
//扑ֈ所对应的节?/span>
InstActivity instActivity=workitem.getInstActivity();
//查找是否存在下一工作?/span>
WorkItem sequenceWorkitem = workItemDAO.findSequenceWorkItem(instActivity.getId(), sID + 1);
//如果不存在则触发节点{
if (sequenceWorkitem == null) {
instActivity.signal();
}
//否则把下一工作Ҏz?/span>
else {
sequenceWorkitem.setExecutive();
}
}
}
Workitemc里有一些状态{换的逻辑Q这样避免直接调用get/set属性方?br />
public class Workitem{
private int state = WorkitemInfo.PREPARE;
/**
* 委派工作?br /> */
public void commission() {
if (state != WorkitemInfo.EXECUTE && state != WorkitemInfo.SIGNINED
&& state != WorkitemInfo.TOREAD&& state != WorkitemInfo.SUSPEND)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMMISSIONED);
setCommitted(new Timestamp(System.currentTimeMillis()));
}
/**
* 完成工作?br /> */
public void complete() {
if (state != WorkitemInfo.SIGNINED)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMPLETE);
setCompleted(new Timestamp(System.currentTimeMillis()));
}
}
private int state = WorkitemInfo.PREPARE;
/**
* 委派工作?br /> */
public void commission() {
if (state != WorkitemInfo.EXECUTE && state != WorkitemInfo.SIGNINED
&& state != WorkitemInfo.TOREAD&& state != WorkitemInfo.SUSPEND)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMMISSIONED);
setCommitted(new Timestamp(System.currentTimeMillis()));
}
/**
* 完成工作?br /> */
public void complete() {
if (state != WorkitemInfo.SIGNINED)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMPLETE);
setCompleted(new Timestamp(System.currentTimeMillis()));
}
}
接下来的做法Q?/strong>
三个cM变，WorkitemManager打^Q将逻辑Ud到Workitem
public class WorkitemManager {
private WorkItemDAO workItemDAO;
public void setWorkItemDAO(WorkItemDAO workItemDAO) {
this.workItemDAO = workItemDAO;
}
/**
* 提交工作?br /> * @param workitemId 工作ID
*/
public void commitWorkitem(String workitemId){
WorkItem workitem = workItemDAO.getWorkItem(workitemId);
//当前工作Ҏ?/span>
workitem.commit();
}
}
private WorkItemDAO workItemDAO;
public void setWorkItemDAO(WorkItemDAO workItemDAO) {
this.workItemDAO = workItemDAO;
}
/**
* 提交工作?br /> * @param workitemId 工作ID
*/
public void commitWorkitem(String workitemId){
WorkItem workitem = workItemDAO.getWorkItem(workitemId);
//当前工作Ҏ?/span>
workitem.commit();
}
}
实际上此时WorkitemManager的功能非常有限，仅仅是事务边界和获取workitem对象Q甚臛_一些情况下可以省略?br />
通过一个Containercdspring的applicationContextq行装Q然后通过getBean()的静态方法即可访问被spring所理的bean。实际是workItemDAO隐式注入了Workitem?br />
public class Workitem{
/**
* 提交工作?br /> */
public void commit() {
if (state != WorkitemInfo.EXECUTE && state != WorkitemInfo.SIGNINED
&& state != WorkitemInfo.TOREAD&& state != WorkitemInfo.SUSPEND)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMMISSIONED);
setCommitted(new Timestamp(System.currentTimeMillis()));
int sID = workitem.getSequenceId();
WorkItemDAO workItemDAO=(WorkItemDAO)Container.getBean("workItemDAO");
//查找是否存在下一工作?/span>
WorkItem sequenceWorkitem = workItemDAO.findSequenceWorkItem(instActivity.getId(), sID + 1);
//如果不存在则触发节点{
if (sequenceWorkitem == null) {
instActivity.signal();
}
//否则把下一工作Ҏz?/span>
else {
sequenceWorkitem.setExecutive();
}
}
}
/**
* 提交工作?br /> */
public void commit() {
if (state != WorkitemInfo.EXECUTE && state != WorkitemInfo.SIGNINED
&& state != WorkitemInfo.TOREAD&& state != WorkitemInfo.SUSPEND)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMMISSIONED);
setCommitted(new Timestamp(System.currentTimeMillis()));
int sID = workitem.getSequenceId();
WorkItemDAO workItemDAO=(WorkItemDAO)Container.getBean("workItemDAO");
//查找是否存在下一工作?/span>
WorkItem sequenceWorkitem = workItemDAO.findSequenceWorkItem(instActivity.getId(), sID + 1);
//如果不存在则触发节点{
if (sequenceWorkitem == null) {
instActivity.signal();
}
//否则把下一工作Ҏz?/span>
else {
sequenceWorkitem.setExecutive();
}
}
}
q样带来的好处是业务逻辑全部被封装到Domain ModelQDomain Model之间的交互变得非常的单，没有频繁的set/getQ直接调用有业务语义的Domain Model的方法即可。问题在于单元测试时q不了spring的容器，workItemDAO需要stub。我觉得q个问题不大Q问题是Domain Model开始变得臃肿，在业务逻辑复杂时代码行急剧膨胀?br />
现在的做?/strong>
以上三个cM持不变，增加一个类WorkitemExecutorQ将业务逻辑UL?br />
public class Workitem{
/**
* 提交工作?br /> */
public void commit() {
if (state != WorkitemInfo.EXECUTE && state != WorkitemInfo.SIGNINED
&& state != WorkitemInfo.TOREAD&& state != WorkitemInfo.SUSPEND)
throw new WorkflowException(Messages.CANNOT_ALTER_WORKITEM_STATE);
setState(WorkitemInfo.COMMISSIONED);
setCommitted(new Timestamp(System.currentTimeMillis()));
WorkitemExecutor workitemExecutor=(WorkitemExecutor)Container.getBean("workitemExecutor");
workitemExecutor.commitWorkitem(this);
}
}
public class WorkitemExecutor {
private WorkItemDAO workItemDAO;
public void setWorkItemDAO(WorkItemDAO workItemDAO) {
this.workItemDAO = workItemDAO;
}
/**
* 提交工作?br /> * @param workitemId 工作ID
*/
public void commitWorkitem(Workitem workitem){
int sID = workitem.getSequenceId();
//扑ֈ所对应的节?/span>
InstActivity instActivity=workitem.getInstActivity();
//查找是否存在下一工作?/span>
WorkItem sequenceWorkitem = workItemDAO.findSequenceWorkItem(instActivity.getId(), sID + 1);
//如果不存在则触发节点{
if (sequenceWorkitem == null) {
instActivity.signal();
}
//否则把下一工作Ҏz?/span>
else {
sequenceWorkitem.setExecutive();
}
}
}
业务逻辑拆分成两部分Q一部分在WorkitemQ另一部分委托lWorkitemExecutor。实际上是Domain Model复杂逻辑的情况重新外包出厅R调用的时候，面向的接口还是Domain Model的方法。注意到WorkitemExecutor和WorkitemManager的API是非常相似的。实际可以这栯为，传统的方?
Client->(Business Facade)->service(Business Logic 部分依赖Domain Model)->Data Access(DAO)?br /> 现在的方?br /> Client->(Business Facade)->Domain Model->service->Data Access(DAO)?br />
另外Q在q回client端的查询的时候还是們于直接调用DAOQ而不是通过Domain Model?br />
改进Q?br /> 注意C码中有这么一?br />
WorkItemDAO workItemDAO=(WorkItemDAO)Container.getBean("workItemDAO");
实是一个bad smell.当代码中大量出现后，q种造型是很恐怖的。所以采取了一U处理方式：l所有Domain Modell承一个父c，在父c里集中理所有Domain Model所依赖的servicesQ在父类里进行造型?br />
]]>
1、对同一张表先insert再update是很快会引v死锁的，不管操作的是否是同一记录
解决ҎQ对于同一记录Q需要调整hibernate的映策略，使得一ơinsert完成操作。对于不同的记录需要在代码中手动flushQ得update先于insert?br />
2、对两张表进行多ơupdate操作Ӟ两张表交替update也会很快引v死锁
解决ҎQ在代码中手动flushQ保证对两张表的update不会出现交替的情c?br />
3、部分大范围扫描的select和update混合也会D死锁
解决ҎQ优化sqlQ尽量减sql语句Q通过lpo增加持久化字D늚方式减少兌查询
l过优化Q大部分情况下数据库死锁的情况得以避免，另外奇怪的是通过事g探查器在死锁时ƈ未发现锁升的事件。但是在一些特D情况下Q例如多个ƈ发汇聚的直接联合Q，死锁依旧发生。最后不得不Ҏ法进行synchronized关键字同步，q个通过synchronized flush完成。业务方法不必同步，最后批量操作数据库时进行同步?br />
换oracleq行试Q在未synchronized的情况下Q未发生死锁情况。由此可见sqlserver与oracle锁实现机制存在很大的差别。对sqlserver鄙视之。另Q同事说Qsqlserver2005后性能和机制发生了很大的变化，未测试?br />
补充一下我的一个最单情况下的测试用例：
POQ?br />
public class TestPO {
String id;
String name;
int num;
.
}
String id;
String name;
int num;
.}
映射文g hibernate3Q?br />
<hibernate-mapping default-access="field">
<class table="WFMS_TESTPO" name="com.eway.workflow.test.po.TestPO">
<id name="id" column="ID"><generator class="uuid" /></id>
<property name="name" column="NAME" type="string"/>
<property name="num" column="NUM" type="integer"/>
</class>
</hibernate-mapping>
<class table="WFMS_TESTPO" name="com.eway.workflow.test.po.TestPO">
<id name="id" column="ID"><generator class="uuid" /></id>
<property name="name" column="NAME" type="string"/>
<property name="num" column="NUM" type="integer"/>
</class>
</hibernate-mapping>
被测试方法（都配|有事务Q：
public void testSave(int num) {
TestPO po = new TestPO();
po.setName("ronghao");
po.setNum(num);
theadTestDao.save(po);
po.setName("haorong");
}
public void testSaveByJdbc(int num) {
String sql = "insert into WFMS_TESTPO (ID,NAME,NUM) values (?,'RONGHAO',?)";
Object[] params = new Object[]{num,num};
jdbcTemplate.update(sql, params);
sql="update WFMS_TESTPO set name='haorong' where id=?" ;
params = new Object[]{num};
jdbcTemplate.update(sql, params);
}
TestPO po = new TestPO();
po.setName("ronghao");
po.setNum(num);
theadTestDao.save(po);
po.setName("haorong");
}
public void testSaveByJdbc(int num) {
String sql = "insert into WFMS_TESTPO (ID,NAME,NUM) values (?,'RONGHAO',?)";
Object[] params = new Object[]{num,num};
jdbcTemplate.update(sql, params);
sql="update WFMS_TESTPO set name='haorong' where id=?" ;
params = new Object[]{num};
jdbcTemplate.update(sql, params);
}
试用例Q?br />
public void testSave() throws Exception {
TheadtestTemplate template = new TheadtestTemplate();
template.execute(new TheadtestCallback() {
public void doInThead(int suquence) {
// theadTestManager.testSave(suquence);
theadTestManager.testSaveByJdbc(suquence);
}
}, 10);
}
TheadtestTemplate template = new TheadtestTemplate();
template.execute(new TheadtestCallback() {
public void doInThead(int suquence) {
// theadTestManager.testSave(suquence);
theadTestManager.testSaveByJdbc(suquence);
}
}, 10);
}
试l果Q不论是hibernateq是jdbcQƈ发情况下都很快就会引起sqlserver2000的死锁，换用两种数据库驱动jtds和jturbo死锁的情冉|有变化?br />
l论Qsqlserver2000数据库的lock配置{略Q不支持Q或者数据库本nQ就不支持对不同的行做同时操作（或者支持不完善Q，所谓的行锁支持很不完善Q死锁情况非常容易发生?br />
补充Q我Ҏ据库的一些实现机制也q不是很了解Q所以这里也只能列出现象而不能解释死锁的Ҏ原因。另外感谢Alex的讨论?br />
]]>
l过查发现是HSQLDB的问?br /> The bug is in HSQLDB - a well known one (any Google search for HSQLDB and that "failed batch"
message would have told you it).
https://sourceforge.net/tracker/?func=detail&atid=378131&aid=1407528&group_id=23316
解决Ҏ : turn off batching with HSQLDB it doesnt work.
讄<prop key="hibernate.jdbc.batch_size">0</prop>卛_
]]>
]]>
JSR-170 是这样定义内容仓库的Q内容仓库由一l?workspaceQ工作空_l成Q这些workspace通常应该包含怼的内宏V一个内容仓库有一个到多个 workspace。每个workspace都是一个树状结构，都有一个唯一的树根节点（root nodeQ。树上的itemQ元素）或者是个nodeQ节点）或者是个propertyQ属性）。每个node都可以有零个到多个子节点和零个到多个子属性。只有根节点没有父节点，其余所有的节点都有一个父节点。property 也必L一个父节点Q但它没有子节点或是子属性，property 是叶子元素。property是真正存储数据的元素?br />
下图描述了一个blog应用E序的内容仓库模型。每个root nodeQ根节点Q的子节炚w代表了一个blog实体。与q个blog实体有关的数据都存储?bolgEntry 节点的属性里Q其中一?blogAttachment property 存储了一个二q制囄文g?br />
Ҏ内容仓库实现的功能，JSR-170定义了三U别：
Level 1Q定义了一个只ȝ内容仓库。功能包括读取内容，内容导ZؓXML和查扑ֆ宏V?br />Level 2Q定义了可写的内容仓库。Level 2是Level 1的扩展，新增的功能包括往内容仓库里写入内容，和从XML导入数据C库?br />Advanced optionsQ定义实CU附加功能，版本控制、JTA、SQL查询、清晰的内容锁定和监视?br />
什么是Apache JackRabbitQ?/b>
Apache JackRabbit是一个开放源码的JSR-170 实现Q实CLevel 2Q但它还有许多扩展的功能。详l可以去它的官方|站?br />
下面我们军_用Apache JackRabbit来作为我们示例程序的内容仓库?br />
如何配置Apache JackRabbit
JackRabbit需要两个参数来配置一个内容仓库实例?br />1.内容仓库ȝ录：q个文g目录下通常包含了所有的内容Q搜索烦引，内部配置文g和其他持久化信息。它的结构看h会像下面q个样子Q?br />
c:/temp
|
|--Blogging
|
|-repository
| |
| |-index
| |-meta
| |-namespaces
| |-nodetypes
|
|-version
|
|-workspace
|
|--default
|
|--Blogging
|
|-repository
| |
| |-index
| |-meta
| |-namespaces
| |-nodetypes
|
|-version
|
|-workspace
|
|--default
在上面的情况下，内容仓库ȝ录是c:/temp/Blogging.
2.内容仓库配置文gQ一个典型的配置文g如下Q?br />
<Repository>
<FileSystem class="org.apache.jackrabbit.core.fs.local.LocalFileSystem">
<param name="path" value="${rep.home}/repository"/>
</FileSystem>
<Security appName="Jackrabbit">
<AccessManager class="org.apache.jackrabbit.core.security.SimpleAccessManager"/>
<LoginModule class="org.apache.jackrabbit.core.security.SimpleLoginModule">
<param name="anonymousId" value="anonymous"/>
</LoginModule>
</Security>
<Workspaces rootPath="${rep.home}/workspaces" defaultWorkspace="default"/>
<Workspace name="${wsp.name}">
<FileSystem class="org.apache.jackrabbit.core.fs.local.LocalFileSystem">
<param name="path" value="${wsp.home}"/>
</FileSystem>
<PersistenceManager
class="org.apache.jackrabbit.core.state.db.DerbyPersistenceManager">
<param name="url" value="jdbc:derby:${wsp.home}/db;create=true"/>
<param name="schemaObjectPrefix" value="${wsp.name}_"/>
</PersistenceManager>
<SearchIndex class="org.apache.jackrabbit.core.query.lucene.SearchIndex">
<param name="path" value="${wsp.home}/index"/>
</SearchIndex>
</Workspace>
<Versioning rootPath="${rep.home}/version">
<FileSystem class="org.apache.jackrabbit.core.fs.local.LocalFileSystem">
<param name="path" value="${rep.home}/version" />
</FileSystem>
<PersistenceManager
class="org.apache.jackrabbit.core.state.db.DerbyPersistenceManager">
<param name="url" value="jdbc:derby:${rep.home}/version/db;create=true"/>
<param name="schemaObjectPrefix" value="version_"/>
</PersistenceManager>
</Versioning>
<SearchIndex class="org.apache.jackrabbit.core.query.lucene.SearchIndex">
<param name="path" value="${rep.home}/repository/index"/>
</SearchIndex>
</Repository>
<FileSystem class="org.apache.jackrabbit.core.fs.local.LocalFileSystem">
<param name="path" value="${rep.home}/repository"/>
</FileSystem>
<Security appName="Jackrabbit">
<AccessManager class="org.apache.jackrabbit.core.security.SimpleAccessManager"/>
<LoginModule class="org.apache.jackrabbit.core.security.SimpleLoginModule">
<param name="anonymousId" value="anonymous"/>
</LoginModule>
</Security>
<Workspaces rootPath="${rep.home}/workspaces" defaultWorkspace="default"/>
<Workspace name="${wsp.name}">
<FileSystem class="org.apache.jackrabbit.core.fs.local.LocalFileSystem">
<param name="path" value="${wsp.home}"/>
</FileSystem>
<PersistenceManager
class="org.apache.jackrabbit.core.state.db.DerbyPersistenceManager">
<param name="url" value="jdbc:derby:${wsp.home}/db;create=true"/>
<param name="schemaObjectPrefix" value="${wsp.name}_"/>
</PersistenceManager>
<SearchIndex class="org.apache.jackrabbit.core.query.lucene.SearchIndex">
<param name="path" value="${wsp.home}/index"/>
</SearchIndex>
</Workspace>
<Versioning rootPath="${rep.home}/version">
<FileSystem class="org.apache.jackrabbit.core.fs.local.LocalFileSystem">
<param name="path" value="${rep.home}/version" />
</FileSystem>
<PersistenceManager
class="org.apache.jackrabbit.core.state.db.DerbyPersistenceManager">
<param name="url" value="jdbc:derby:${rep.home}/version/db;create=true"/>
<param name="schemaObjectPrefix" value="version_"/>
</PersistenceManager>
</Versioning>
<SearchIndex class="org.apache.jackrabbit.core.query.lucene.SearchIndex">
<param name="path" value="${rep.home}/repository/index"/>
</SearchIndex>
</Repository>
在这个配|文仉Q?lt;Repository>元素是根元素Q它包含了下面这些元素：
aQ?lt;FileSystem>: 该元素配|了内容仓库的全局数据存储位置Q这些全局数据包括已注册的命名I间Q定制的节点cd{等。 ? JackRabbit 提供了几U选择Q一U是像上面例子里配置的存储在本地文g里，LocalFileSystem. 如果你想把它们存储在数据库里Q你可以使用 DbFileSystem.
bQ?lt;Security>:内容仓库的安全配|，它有两个子元素：<AccessManager>?lt;LoginModule>?lt;AccessManager>配置的类用来判断用户有没有权限来对特定数据执行特定的操作?br /> cQ?lt;Workspaces>:q个元素的配|对所有的workspace都通用。它的rootPath 属性是所有workspace文g夹的根目录，在我们的例子里它是c:/temp/Blogging/WorkspaceQdefaultWorkspace 属性则包含了workspace的默认名?br /> dQ?lt;Workspace>:q个元素是所有workspace的默认配|模ѝ去每个workspace文g夹下你都会发C个workspace.xml文gQ这个文件和q个元素的配|一模一栗三个子元素Q?lt;FileSystem>Q和q个workspace相关数据的存储位|；<PersistenceManager> Q这个workspace内容节点存储{略Q?lt;SearchIndex>Q可选，全文索?br /> eQ?lt;Versioning>:配置一个版本相关的对象。其实JackRabbit也是把它作ؓ节点来处理的?br />
q两个参数可以通过两种方式讄Q一U是在仓库实例创建时直接传到Jackrabbit里去Q一U是间接的通过讄JNDI object factory?br />你可以设|org.apache.jackrabbit.repository.home q个pȝ属性的值来指定你的内容仓库ȝ录；也可以设|?br />org.apache.jackrabbit.repository.conf q个pȝ属性的值来指定你的内容仓库配置文grepository.xml。如果你不设定这两个
参数QJackrabbit会把当前目录作ؓ内容仓库ȝ录，同时Q它有一个默认的内容仓库配置文g?img src ="http://www.aygfsteel.com/RongHao/aggbug/95637.html" width = "1" height = "1" />
]]>
问题:我们真的需要这么多的filter?q些filterqo的范围认真考虑q吗,可以~小?
]]>
]]>
2. 解压~?jbpm-3.0.zip ?'temp' 目录
3. 使用 eclipse, ?'temp\jbpm-3.0' 作ؓ an existing project into workspace 导入
配置q接 MySQL
1. ?'jbpm-3.0\lib' 目录?创徏 'mysql' 目录
2. ?mysql数据库驱?(mysql-connector-java-3.1.7-bin.jar) 拯?'mysql' 目录
3. ?mysql 中创Z个数据库Q数据库名字
5. 把两个配|文?(create.db.hibernate.properties, identity.db.xml) ?'hsqldb' 目录?拯?'mysql' 目录
6. 按下面所C编?'create.db.hibernate.properties' 文g: hibernate.dialect=org.hibernate.dialect.MySQLDialect
hibernate.connection.driver_class=com.mysql.jdbc.Driver
hibernate.connection.url=jdbc:mysql://localhost:3306/
hibernate.connection.username=
hibernate.show_sql=true hibernate.query.substitutions=true 1, false 0
hibernate.c3p0.min_size=1 hibernate.c3p0.max_size=3
7. 另一个文?'identity.db.xml'不做改动
8. ?'jbpm-3.0' 根目? ~辑ANT的脚?'build.deploy.xml' 扑ֈ target name="create.db", 删除 db.start, db.stop 在这个目标块中将所有的'hsqldb' 替换?'mysql'
9. q行ANT ant create.db -buildfile build.deploy.xml q行完毕后就会发现mysql中多出很多表Q这是jbpm保持状态用?
创徏 jbpm.war 使其在tomcat中运?/STRONG>
默认的打war包时Q掉了一些库文g
1. ?eclipse? ~辑ant脚本 'build.deploy.xml' 在目标块 target name="build.webapp" 中在
<copy todir="build/jbpm.war.dir/WEB-INF/lib">
<fileset dir="build" includes="jbpm-webapp-${jbpm.version}.jar" />
<fileset dir="build" includes="jbpm*.jar" />
<fileset dir="lib/hibernate" includes="*.jar" />
<fileset dir="lib/bsh" includes="*.jar" />
3. 打开源文?JbpmSessionFactory.java, ?getInstance() Ҏ? 删除下面代码
InitialContext initialContext = new InitialContext(); Object o = initialContext.lookup(jndiName);
下面这?
instance = (JbpmSessionFactory) PortableRemoteObject.narrow
(o, JbpmSessionFactory.class);
替换?instance = (JbpmSessionFactory) PortableRemoteObject.narrow
(new JbpmSessionFactory(createConfiguration()), JbpmSessionFactory.class);
4.?createConfiguration(String configResource) Ҏ? 注释掉这D代?
String hibernatePropertiesResource = JbpmConfiguration.getString("jbpm.hibernate.properties");
if (hibernatePropertiesResource!=null) { Properties hibernateProperties =
new Properties();
try { hibernateProperties.load( ClassLoaderUtil.getStream(hibernatePropertiesResource) ); }
catch (IOException e) {
e.printStackTrace();
throw new RuntimeException
("couldn't load the hibernate properties from resource '"hibernatePropertiesResource"'", e);
}
log.debug("overriding hibernate properties with "+ hibernateProperties); configuration.setProperties(hibernateProperties);
}
同时加入下面的代?
configuration.setProperty("hibernate.dialect", "org.hibernate.dialect.MySQLDialect");
configuration.setProperty("hibernate.connection.driver_class", "com.mysql.jdbc.Driver");
configuration.setProperty("hibernate.connection.url", "jdbc:mysql://localhost:3306/
configuration.setProperty("hibernate.connection.username", "
configuration.setProperty("hibernate.connection.password", "
configuration.setProperty("hibernate.connection.pool_size", "15");
5. q行脚本命o ant build ant build.webapp -buildfile build.deploy.xml
6. jbpm.war ?'jbpm-3.0\build' 下拷贝到 'tomcat.home\webapps'
7. 启动 tomcat
8. 打开览?'http://localhost:8080/jbpm'
]]>