MemCache分布式缓存的一个bug

Memcached分布式缓存策略不是由服务器端至支持的，多台服务器之间并不知道彼此的存在。分布式的实现是由客户端代码（Memcached.ClientLibrary）通过缓存key-server映射来实现的，基本原理就是对缓存key求hash值，用hash值对服务器数量进行模运算，该key值被分配到模运算结果为索引的那台server上。

Memcached.ClientLibrary对缓存key计算hashcode的核心算法如下：

 /// <summary>
 /// Returns appropriate SockIO object given
 /// string cache key and optional hashcode.
 ///
 /// Trys to get SockIO from pool.  Fails over
 /// to additional pools in event of server failure.
 /// </summary>
 /// <param name="key">hashcode for cache key</param>
 /// <param name="hashCode">if not null, then the int hashcode to use</param>
 /// <returns>SockIO obj connected to server</returns>
 public SockIO GetSock(string key, object hashCode)
 {
     string hashCodeString = "<null>";
     if(hashCode != null)
         hashCodeString = hashCode.ToString();

     if(Log.IsDebugEnabled)
     {
         Log.Debug(GetLocalizedString("cache socket pick").Replace("$$Key$$", key).Replace("$$HashCode$$", hashCodeString));
     }

     if (key == null || key.Length == )
     {
         if(Log.IsDebugEnabled)
         {
             Log.Debug(GetLocalizedString("null key"));
         }
         return null;
     }

     if(!_initialized)
     {
         if(Log.IsErrorEnabled)
         {
             Log.Error(GetLocalizedString("get socket from uninitialized pool"));
         }
         return null;
     }

     // if no servers return null
     if(_buckets.Count == )
         return null;

     // if only one server, return it
     if(_buckets.Count == )
         return GetConnection((string)_buckets[]);

     int tries = ;

     // generate hashcode
     int hv;
     if(hashCode != null)
     {
         hv = (int)hashCode;
     }
     else
     {

         // NATIVE_HASH = 0
         // OLD_COMPAT_HASH = 1
         // NEW_COMPAT_HASH = 2
         switch(_hashingAlgorithm)
         {
             case HashingAlgorithm.Native:
                 hv = key.GetHashCode();
                 break;

             case HashingAlgorithm.OldCompatibleHash:
                 hv = OriginalHashingAlgorithm(key);
                 break;

             case HashingAlgorithm.NewCompatibleHash:
                 hv = NewHashingAlgorithm(key);
                 break;

             default:
                 // use the native hash as a default
                 hv = key.GetHashCode();
                 _hashingAlgorithm = HashingAlgorithm.Native;
                 break;
         }
     }

     // keep trying different servers until we find one
     while(tries++ <= _buckets.Count)
     {
         // get bucket using hashcode
         // get one from factory
         int bucket = hv % _buckets.Count;
         if(bucket < )
             bucket += _buckets.Count;

         SockIO sock = GetConnection((string)_buckets[bucket]);

         if(Log.IsDebugEnabled)
         {
             Log.Debug(GetLocalizedString("cache choose").Replace("$$Bucket$$", _buckets[bucket].ToString()).Replace("$$Key$$", key));
         }

         if(sock != null)
             return sock;

         // if we do not want to failover, then bail here
         if(!_failover)
             return null;

         // if we failed to get a socket from this server
         // then we try again by adding an incrementer to the
         // current key and then rehashing
         switch(_hashingAlgorithm)
         {
             case HashingAlgorithm.Native:
                 hv += ((string)("" + tries + key)).GetHashCode();
                 break;

             case HashingAlgorithm.OldCompatibleHash:
                 hv += OriginalHashingAlgorithm("" + tries + key);
                 break;

             case HashingAlgorithm.NewCompatibleHash:
                 hv += NewHashingAlgorithm("" + tries + key);
                 break;

             default:
                 // use the native hash as a default
                 hv += ((string)("" + tries + key)).GetHashCode();
                 _hashingAlgorithm = HashingAlgorithm.Native;
                 break;
         }
     }

     return null;
 }

根据缓存key得到服务器的核心代码

从源码中（62--82行代码）可以发现，计算hashcode的算法共三种：

（1）HashingAlgorithm.Native: 即使用.NET本身的hash算法，速度快，但与其他client可能不兼容，例如需要和java、ruby的client共享缓存的情况；

（2）HashingAlgorithm.OldCompatibleHash: 可以与其他客户端兼容，但速度慢；

（3）HashingAlgorithm.NewCompatibleHash: 可以与其他客户端兼容，据称速度快。

进一步分析发现，Memcached.ClientLibrary默认计算缓存key的hashcode的方式就是HashingAlgorithm.Native，而HashingAlgorithm.Native计算hashcode的算法为“hv = key.GetHashCode()”，即用了.net类库string类型自带的GetHashCode()方法。

Bug就要浮现出来了，根据微软（http://msdn.microsoft.com/zh-cn/library/system.object.gethashcode.aspx）对GetHashCode的解释：the .NET Framework does not guarantee the default implementation of the GetHashCode method, and the value this method returns may differ between .NET Framework versions and platforms, such as 32-bit and 64-bit platforms。string类型的GetHashCode()函数并不能保证不同平台同一个字符串返回的hash值相同，这样问题就出来了，对于不同服务器的同一缓存key来说，产生的hashcode可能不同，同一key对应的数据可能缓存到了不同的MemCache服务器上，数据的一致性无法保证，清除缓存的代码也可能失效。

// 64位 4.0
[__DynamicallyInvokable, ReliabilityContract(Consistency.WillNotCorruptState, Cer.MayFail), SecuritySafeCritical]
public unsafe override int GetHashCode()
{
    if (HashHelpers.s_UseRandomizedStringHashing)
    {
        return string.InternalMarvin32HashString(this, this.Length, 0L);
    }
    IntPtr arg_25_0;
    IntPtr expr_1C = arg_25_0 = this;
    if (expr_1C != )
    {
        arg_25_0 = (IntPtr)((int)expr_1C + RuntimeHelpers.OffsetToStringData);
    }
    char* ptr = arg_25_0;
    int num = ;
    int num2 = num;
    char* ptr2 = ptr;
    int num3;
    while ((num3 = (int)(*(ushort*)ptr2)) != )
    {
        num = ((num << ) + num ^ num3);
        num3 = (int)(*(ushort*)(ptr2 + (IntPtr) / ));
        if (num3 == )
        {
            break;
        }
        num2 = ((num2 << ) + num2 ^ num3);
        ptr2 += (IntPtr) / ;
    }
    return num + num2 * ;
}

// 64位 2.0
// string
[ReliabilityContract(Consistency.WillNotCorruptState, Cer.MayFail)]
public unsafe override int GetHashCode()
{
    IntPtr arg_0F_0;
    IntPtr expr_06 = arg_0F_0 = this;
    if (expr_06 != )
    {
        arg_0F_0 = (IntPtr)((int)expr_06 + RuntimeHelpers.OffsetToStringData);
    }
    char* ptr = arg_0F_0;
    int num = ;
    int num2 = num;
    char* ptr2 = ptr;
    int num3;
    while ((num3 = (int)(*(ushort*)ptr2)) != )
    {
        num = ((num << ) + num ^ num3);
        num3 = (int)(*(ushort*)(ptr2 + (IntPtr) / ));
        if (num3 == )
        {
            break;
        }
        num2 = ((num2 << ) + num2 ^ num3);
        ptr2 += (IntPtr) / ;
    }
    return num + num2 * ;
}

//32位 4.0
[__DynamicallyInvokable, ReliabilityContract(Consistency.WillNotCorruptState, Cer.MayFail), SecuritySafeCritical]
public unsafe override int GetHashCode()
{
    if (HashHelpers.s_UseRandomizedStringHashing)
    {
        return string.InternalMarvin32HashString(this, this.Length, 0L);
    }
    IntPtr arg_25_0;
    IntPtr expr_1C = arg_25_0 = this;
    if (expr_1C != )
    {
        arg_25_0 = (IntPtr)((int)expr_1C + RuntimeHelpers.OffsetToStringData);
    }
    char* ptr = arg_25_0;
    int num = ;
    int num2 = num;
    int* ptr2 = (int*)ptr;
    int i;
    for (i = this.Length; i > ; i -= )
    {
        num = ((num << ) + num + (num >> ) ^ *ptr2);
        num2 = ((num2 << ) + num2 + (num2 >> ) ^ ptr2[(IntPtr) / ]);
        ptr2 += (IntPtr) / ;
    }
    if (i > )
    {
        num = ((num << ) + num + (num >> ) ^ *ptr2);
    }
    return num + num2 * ;
}

GetHashCode几种版本的实现代码

解决问题的方法就是不要用MemCache默认的hash算法，实现方式有两种：

（1）初始化MemCache服务器的时候，指定为MemCahce自带其它的hash算法，代码为“this.pool.HashingAlgorithm = HashingAlgorithm.OldCompatibleHash;”。

（2）自定义hash算法，调用set()、get()、delete()等方式时传递hash值，这几个方法有参数传递hashcode的重载。

参考资料：分析Memcached客户端如何把缓存数据分布到多个服务器上(转)、memcached client - memcacheddotnet (Memcached.ClientLibrary) 1.1.5、memcache分布式实现、Object.GetHashCode 方法、关于 HashCode做key的可能性。