Redis跟MySQL同样,拥有很是强大的主从复制功能,并且还支持一个master能够拥有多个slave,而一个slave又能够拥有多个slave,从而造成强大的多级服务器集群架构。
redis的主从复制是异步进行的,它不会影响master的运行,因此不会下降redis的处理性能。主从架构中,能够考虑关闭Master的数据持久化功能,只让Slave进行持久化,这样能够提升主服务器的处理性能。同时Slave为只读模式,这样能够避免Slave缓存的数据被误修改。html
1.配置redis
实际生产中,主从架构是在几个不一样服务器上安装相应的Redis服务。为了测试方便,我这边的主从备份的配置,都是在我Windows 本机上测试。缓存
1. 安装两个Redis 实例,Master和Slave。将Master端口设置为6379,Slave 端口设置为6380 。bind 都设置为:127.0.0.1。 服务器
2. 在Slave 实例 ,增长:slaveof 127.0.0.1 6379 配置。以下图所示:session
配置完成以后,启动这两个实例,若是输出以下内容,说明主从复制的架构已经配置成功了。架构
注意:在同一台电脑上测试,Master和Slave的端口不要同样,不然是不能同时启动两个实例的。并发
2.测试app
在命令行,分别链接上Master服务器和Slave 服务器。而后在Master 写入缓存,而后在Slave 中读取。以下图所示:异步
3.C#中调用性能
主从架构的Redis的读写其实和单台Redis 的读写差很少,只是部分配置和读取区分了主从,若是不清楚C#中如何使用redis,请参考我这篇文章 《Redis总结(二)C#中如何使用redis》。
须要注意的是:ServiceStack.Redis 中GetClient()方法,只能拿到Master redis中获取链接,而拿不到slave 的readonly链接。这样 slave起到了冗余备份的做用,读的功能没有发挥出来,若是并发请求太多的话,则Redis的性能会有影响。
因此,咱们须要的写入和读取的时候作一个区分,写入的时候,调用client.GetClient() 来获取writeHosts的Master的redis 连接。读取,则调用client.GetReadOnlyClient()来获取的readonlyHost的 Slave的redis连接。
或者能够直接使用client.GetCacheClient() 来获取一个链接,他会在写的时候调用GetClient获取链接,读的时候调用GetReadOnlyClient获取链接,这样能够作到读写分离,从而利用redis的主从复制功能。
1. 配置文件 app.config
<!-- redis Start --> <add key="SessionExpireMinutes" value="180" /> <add key="redis_server_master_session" value="127.0.0.1:6379" /> <add key="redis_server_slave_session" value="127.0.0.1:6380" /> <add key="redis_max_read_pool" value="300" /> <add key="redis_max_write_pool" value="100" /> <!--redis end-->
2. Redis操做的公用类RedisCacheHelper
using System; using System.Collections.Generic; using System.Configuration; using System.Linq; using System.Text; using System.Web; using ServiceStack.Common.Extensions; using ServiceStack.Redis; using ServiceStack.Logging; namespace Weiz.Redis.Common { public class RedisCacheHelper { private static readonly PooledRedisClientManager pool = null; private static readonly string[] writeHosts = null; private static readonly string[] readHosts = null; public static int RedisMaxReadPool = int.Parse(ConfigurationManager.AppSettings["redis_max_read_pool"]); public static int RedisMaxWritePool = int.Parse(ConfigurationManager.AppSettings["redis_max_write_pool"]); static RedisCacheHelper() { var redisMasterHost = ConfigurationManager.AppSettings["redis_server_master_session"]; var redisSlaveHost = ConfigurationManager.AppSettings["redis_server_slave_session"]; if (!string.IsNullOrEmpty(redisMasterHost)) { writeHosts = redisMasterHost.Split(','); readHosts = redisSlaveHost.Split(','); if (readHosts.Length > 0) { pool = new PooledRedisClientManager(writeHosts, readHosts, new RedisClientManagerConfig() { MaxWritePoolSize = RedisMaxWritePool, MaxReadPoolSize = RedisMaxReadPool, AutoStart = true }); } } } public static void Add<T>(string key, T value, DateTime expiry) { if (value == null) { return; } if (expiry <= DateTime.Now) { Remove(key); return; } try { if (pool != null) { using (var r = pool.GetClient()) { if (r != null) { r.SendTimeout = 1000; r.Set(key, value, expiry - DateTime.Now); } } } } catch (Exception ex) { string msg = string.Format("{0}:{1}发生异常!{2}", "cache", "存储", key); } } public static void Add<T>(string key, T value, TimeSpan slidingExpiration) { if (value == null) { return; } if (slidingExpiration.TotalSeconds <= 0) { Remove(key); return; } try { if (pool != null) { using (var r = pool.GetClient()) { if (r != null) { r.SendTimeout = 1000; r.Set(key, value, slidingExpiration); } } } } catch (Exception ex) { string msg = string.Format("{0}:{1}发生异常!{2}", "cache", "存储", key); } } public static T Get<T>(string key) { if (string.IsNullOrEmpty(key)) { return default(T); } T obj = default(T); try { if (pool != null) { using (var r = pool.GetClient()) { if (r != null) { r.SendTimeout = 1000; obj = r.Get<T>(key); } } } } catch (Exception ex) { string msg = string.Format("{0}:{1}发生异常!{2}", "cache", "获取", key); } return obj; } public static void Remove(string key) { try { if (pool != null) { using (var r = pool.GetClient()) { if (r != null) { r.SendTimeout = 1000; r.Remove(key); } } } } catch (Exception ex) { string msg = string.Format("{0}:{1}发生异常!{2}", "cache", "删除", key); } } public static bool Exists(string key) { try { if (pool != null) { using (var r = pool.GetClient()) { if (r != null) { r.SendTimeout = 1000; return r.ContainsKey(key); } } } } catch (Exception ex) { string msg = string.Format("{0}:{1}发生异常!{2}", "cache", "是否存在", key); } return false; } public static IDictionary<string, T> GetAll<T>(IEnumerable<string> keys) where T : class { if (keys == null) { return null; } keys = keys.Where(k => !string.IsNullOrWhiteSpace(k)); if (keys.Count() == 1) { T obj = Get<T>(keys.Single()); if (obj != null) { return new Dictionary<string, T>() { { keys.Single(), obj } }; } return null; } if (!keys.Any()) { return null; } IDictionary<string, T> dict = null; if (pool != null) { keys.Select(s => new { Index = Math.Abs(s.GetHashCode()) % readHosts.Length, KeyName = s }) .GroupBy(p => p.Index) .Select(g => { try { using (var r = pool.GetClient(g.Key)) { if (r != null) { r.SendTimeout = 1000; return r.GetAll<T>(g.Select(p => p.KeyName)); } } } catch (Exception ex) { string msg = string.Format("{0}:{1}发生异常!{2}", "cache", "获取", keys.Aggregate((a, b) => a + "," + b)); } return null; }) .Where(x => x != null) .ForEach(d => { d.ForEach(x => { if (dict == null || !dict.Keys.Contains(x.Key)) { if (dict == null) { dict = new Dictionary<string, T>(); } dict.Add(x); } }); }); } IEnumerable<Tuple<string, T>> result = null; if (dict != null) { result = dict.Select(d => new Tuple<string, T>(d.Key, d.Value)); } else { result = keys.Select(key => new Tuple<string, T>(key, Get<T>(key))); } return result .Select(d => new Tuple<string[], T>(d.Item1.Split('_'), d.Item2)) .Where(d => d.Item1.Length >= 2) .ToDictionary(x => x.Item1[1], x => x.Item2); } } }
转自:https://www.cnblogs.com/zhangweizhong/p/4980639.html