基于Redis的CAS集群
单点登录(SSO)是复杂应用系统的基本需求,Yale CAS是目前常用的开源解决方案。CAS认证中心,基于其特殊作用,自然会成为整个应用系统的核心,所有应用系统的认证工作,都将请求到CAS来完成。因此CAS服务器是整个应用的关键节点,CAS发生故障,所有系统都将陷入瘫痪。同时,CAS的负载能力要足够强,能够承担所有的认证请求响应。利用负载均衡和集群技术,不仅能克服CAS单点故障,同时将认证请求分布到多台CAS服务器上,有效减轻单台CAS服务器的请求压力。下面将基于CAS 3.4.5来讨论下CAS集群。
CAS的工作原理,主要是基于票据(Ticket)来实现的(参见 CAS基本原理)。CAS票据,存储在TicketRegistry中,因此要想实现CAS Cluster, 必须要多台CAS之间共享所有的Ticket,采用统一的TicketRegistry,可以达到此目的。 缺省的CAS实现中,TicketRegistry在内存中实现,不同的CAS服务器有自己单独的TicketRegistry,因此是不支持分布式集群的。但CAS提供了支持TicketRegistry分布式的接口 org.jasig.cas.ticket.registry.AbstractDistributedTicketRegistry,我们可以实现这个接口实现多台CAS服务器TicketRegistry共享,从而实现CAS集群。
同时,较新版本CAS使用SpringWebFlow作为认证流程,而webflow需要使用session存储流程相关信息,因此实现CAS集群,我们还得需要让不同服务器的session进行共享。
我们采用内存数据库Redis来实现TicketRegistry,让多个CAS服务器共用同一个TicketRegistry。同样方法,我们让session也存储在Redis中,达到共享session的目的。下面就说说如何用 Redis来实现TicketRegistry,我们使用Java调用接口Jedis来操作Redis,代码如下: import java.io.ByteArrayInputStream; import java.io.ByteArrayOutputStream; import java.io.ObjectInputStream; import java.io.ObjectOutputStream; import java.util.Collection; import org.jasig.cas.ticket.Ticket; import org.jasig.cas.ticket.TicketGrantingTicket; import org.jasig.cas.ticket.registry.AbstractDistributedTicketRegistry; import redis.clients.jedis.Jedis; import redis.clients.jedis.JedisPool; import redis.clients.jedis.JedisPoolConfig; /* * TicketRegistry using Redis, to solve CAS Cluster. * * @author ZL * */ public class RedisTicketRegistry extends AbstractDistributedTicketRegistry { private static int redisDatabaseNum; private static String hosts; private static int port; private static int st_time; //ST最大空闲时间 private static int tgt_time; //TGT最大空闲时间 private static JedisPool cachePool; static { redisDatabaseNum = PropertiesConfigUtil.getPropertyInt("redis_database_num"); hosts = PropertiesConfigUtil.getProperty("hosts"); port = PropertiesConfigUtil.getPropertyInt("port"); st_time = PropertiesConfigUtil.getPropertyInt("st_time"); tgt_time = PropertiesConfigUtil.getPropertyInt("tgt_time"); cachePool = new JedisPool(new JedisPoolConfig(), hosts, port); } public void addTicket(Ticket ticket) { Jedis jedis = cachePool.getResource(); jedis.select(redisDatabaseNum); int seconds = 0; String key = ticket.getId() ; if(ticket instanceof TicketGrantingTicket){ //key = ((TicketGrantingTicket)ticket).getAuthentication().getPrincipal().getId(); seconds = tgt_time/1000; }else{ seconds = st_time/1000; } ByteArrayOutputStream bos = new ByteArrayOutputStream(); ObjectOutputStream oos = null; try{ oos = new ObjectOutputStream(bos); oos.writeObject(ticket); }catch(Exception e){ log.error("adding ticket to redis error."); }finally{ try{ if(null!=oos) oos.close(); }catch(Exception e){ log.error("oos closing error when adding ticket to redis."); } } jedis.set(key.getBytes(), bos.toByteArray()); jedis.expire(key.getBytes(), seconds); cachePool.returnResource(jedis); } public Ticket getTicket(final String ticketId) { return getProxiedTicketInstance(getRawTicket(ticketId)); } private Ticket getRawTicket(final String ticketId) { if(null == ticketId) return null; Jedis jedis = cachePool.getResource(); jedis.select(redisDatabaseNum); Ticket ticket = null; ByteArrayInputStream bais = new ByteArrayInputStream(jedis.get(ticketId.getBytes())); ObjectInputStream ois = null; try{ ois = new ObjectInputStream(bais); ticket = (Ticket)ois.readObject(); }catch(Exception e){ log.error("getting ticket to redis error."); }finally{ try{ if(null!=ois) ois.close(); }catch(Exception e){ log.error("ois closing error when getting ticket to redis."); } } cachePool.returnResource(jedis); return ticket; } public boolean deleteTicket(final String ticketId) { if (ticketId == null) { return false; } Jedis jedis = cachePool.getResource(); jedis.select(redisDatabaseNum); jedis.del(ticketId.getBytes()); cachePool.returnResource(jedis); return true; } public Collection<Ticket> getTickets() { throw new UnsupportedOperationException("GetTickets not supported."); } protected boolean needsCallback() { return false; } protected void updateTicket(final Ticket ticket) { addTicket(ticket); } }
同时,我们在ticketRegistry.xml配置文件中,将TicketRegistry实现类指定为上述实现。即修改下面的class值 <!-- Ticket Registry --> <bean id="ticketRegistry" class="org.jasig.cas.util.RedisTicketRegistry" /> <!-- <bean id="ticketRegistry" class="org.jasig.cas.ticket.registry.DefaultTicketRegistry" /> -->
因为使用了Redis的expire功能,注释掉如下代码:
<!-- TICKET REGISTRY CLEANER --> lt;!-- <bean id="ticketRegistryCleaner" class="org.jasig.cas.ticket.registry.support.DefaultTicketRegistryCleaner" p:ticketRegistry-ref="ticketRegistry" /> <bean id="jobDetailTicketRegistryCleaner" class="org.springframework.scheduling.quartz.MethodInvokingJobDetailFactoryBean" p:targetObject-ref="ticketRegistryCleaner" p:targetMethod="clean" /> <bean id="triggerJobDetailTicketRegistryCleaner" class="org.springframework.scheduling.quartz.SimpleTriggerBean" p:jobDetail-ref="jobDetailTicketRegistryCleaner" p:startDelay="20000" p:repeatInterval="5000000" /> -->通过上述实现TicketRegistry,多台CAS服务器就可以共用同一个 TicketRegistry。对于如何共享session,我们可以采用现成的第三方工具tomcat-redis-session-manager直接集成即可。对于前端web服务器(如nginx),做好负载均衡配置,将认证请求分布转发给后面多台CAS,实现负载均衡和容错目的。