Memcached一致性哈希测试
1. 准备
- 两台已经装好memached的机器,假如ip分别为192.168.1.2和192.168.1.3
- spy memcached 客户端jar包
- 启动memached
2. 编写Java测试程序
- 测试哈希一致性的写入
@Test
public void testSetWithConsistentHashing() throws IOException, InterruptedException {
ConnectionFactory connFactory = new ConnectionFactoryBuilder().setLocatorType(Locator.CONSISTENT).build();
MemcachedClient clusterClient = new MemcachedClient(
connFactory,
AddrUtil.getAddresses("192.168.6.201:11211 192.168.6.205:11211"));
MemcachedClient clientOne = new MemcachedClient(AddrUtil.getAddresses("192.168.6.201:11211"));
MemcachedClient clientTwo = new MemcachedClient(AddrUtil.getAddresses("192.168.6.205:11211"));
// 放1000个
for (int i = 0; i < 1000; i++) {
clusterClient.set("test" + i, 300, "test" + i);
}
while (true) {
System.out.println("===================== in one =================================");
// 从one中获取,看是否存在,存在的key
int oneCount = 0;
for (int i = 0; i < 1000; i++) {
Object value = clientOne.get("test" + i);
if (value != null) {
oneCount++;
System.out.println(value);
}
}
Assert.assertTrue(oneCount > 0);
Thread.sleep(3 * 1000);
System.out.println("======================= in two ===============================");
// 从one中获取,看是否存在,存在的key
int twoCount = 0;
for (int i = 0; i < 1000; i++) {
Object value = clientTwo.get("test" + i);
if (value != null) {
twoCount++;
System.out.println(value);
}
}
Assert.assertTrue(twoCount > 0);
}
}
- 改造程序一,测试写入和读取过程中启动或者停掉加入的节点机器
@Test
public void testGetValueWithConsistentHashing() throws IOException, InterruptedException {
ConnectionFactory connFactory = new ConnectionFactoryBuilder().setLocatorType(Locator.CONSISTENT).build();
MemcachedClient clusterClient = new MemcachedClient(
connFactory,
AddrUtil.getAddresses("192.168.6.201:11211 192.168.6.205:11211"));
MemcachedClient clientOne = new MemcachedClient(AddrUtil.getAddresses("192.168.6.201:11211"));
MemcachedClient clientTwo = new MemcachedClient(AddrUtil.getAddresses("192.168.6.205:11211"));
while (true) {
try {
// 存数据
for (int i = 0; i < 1000; i++) {
clusterClient.set("test" + i, 300, "test" + i);
}
// 校验one中是否有数据
System.out.println("===================== in one =================================");
int oneCount = 0;
for (int i = 0; i < 1000; i++) {
Object value = clientOne.get("test" + i);
if (value != null) {
oneCount++;
if (oneCount > 3) {
continue;
}
System.out.println(value);
}
}
System.out.println("one count: " + oneCount);
// //校验two中是否有数据
System.out.println("===================== in two =================================");
int twoCount = 0;
for (int i = 0; i < 1000; i++) {
Object value = clientTwo.get("test" + i);
if (value != null) {
twoCount++;
if (twoCount > 3) {
continue;
}
System.out.println(value);
}
}
System.out.println("two count: " + twoCount);
Thread.sleep(3 * 1000);
} catch (Exception e) {
e.printStackTrace();
}
}
}
3. 测试
同时开启两台memcached,然后执行第一个测试程序,结果执行通过,发现两台机器的memcached中都有要设置的数据,而且是不同的.
同时开启两台memcached,然后执行第二个测试程序,开始执行征程.然后停掉其中一台,没多长时间,测试程序抛出异常
2014-08-29 19:27:57.752 INFO net.spy.memcached.MemcachedConnection: Reconnecting {QA sa=/192.168.1.2:11211, #Rops=0, #Wops=23, #iq=0, topRop=null, topWop=net.spy.memcached.protocol.ascii.StoreOperationImpl@567a380b, toWrite=0, interested=0}
2014-08-29 19:27:57.753 INFO net.spy.memcached.MemcachedConnection: Connection state changed for sun.nio.ch.SelectionKeyImpl@5e94fd79
2014-08-29 19:27:57.754 INFO net.spy.memcached.MemcachedConnection: Reconnecting due to failure to connect to {QA sa=/192.168.1.2:11211, #Rops=0, #Wops=23, #iq=0, topRop=null, topWop=net.spy.memcached.protocol.ascii.StoreOperationImpl@567a380b, toWrite=0, interested=0}
java.net.ConnectException: 拒绝连接
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:712)
at net.spy.memcached.MemcachedConnection.handleIO(MemcachedConnection.java:313)
at net.spy.memcached.MemcachedConnection.handleIO(MemcachedConnection.java:199)
at net.spy.memcached.MemcachedClient.run(MemcachedClient.java:1622)
2014-08-29 19:27:57.754 WARN net.spy.memcached.MemcachedConnection: Closing, and reopening {QA sa=/192.168.1.2:11211, #Rops=0, #Wops=23, #iq=0, topRop=null, topWop=net.spy.memcached.protocol.ascii.StoreOperationImpl@567a380b, toWrite=0, interested=0}, attempt 1.
然后就隔一段时间抛出一次异常,且每次抛出异常的时间是不断推迟的,具体推迟到多久这个尚未测试.同时所要写入的1000个值全部写入到了另一台机器中.
- 开启停掉的那台机器,客户端自动重新链接了,同时恢复了对刚刚恢复的机器的写入和读取
2014-08-29 20:03:06.157 WARN net.spy.memcached.MemcachedConnection: Could not redistribute to another node, retrying primary node for test0.
2014-08-29 20:03:06.234 INFO net.spy.memcached.MemcachedConnection: Reconnecting {QA sa=/192.168.1.2:11211, #Rops=0, #Wops=22, #iq=0, topRop=null, topWop=net.spy.memcached.protocol.ascii.StoreOperationImpl@6d2cb32a, toWrite=0, interested=0}
2014-08-29 20:03:06.235 INFO net.spy.memcached.MemcachedConnection: Connection state changed for sun.nio.ch.SelectionKeyImpl@51906e2
2014-08-29 20:03:06.476 INFO net.spy.memcached.MemcachedConnection: Reconnecting {QA sa=/192.168.1.2:11211, #Rops=0, #Wops=3, #iq=0, topRop=null, topWop=net.spy.memcached.protocol.ascii.GetOperationImpl@59b3d57, toWrite=0, interested=0}
2014-08-29 20:03:06.477 INFO net.spy.memcached.MemcachedConnection: Connection state changed for sun.nio.ch.SelectionKeyImpl@399931f9
4. 结果
spymemcached 客户端对memcached哈希一致性比较i理想,在有多台memcached的情况下,挂掉一台,不会造成全部损失,但会失去这台机器的缓存,导致这部分数据可能会落在数据库上,相比单节点挂掉,情况要理想的多.
5. 探索
如果支持主备方案,那么情况会更加理想.