上回测试失败（nodejs cluster测试 ）的原因总算找到了，不是nodejs的原因，nodejs官方cluster工作良好，问题在于测试方式上，在浏览器中同时开两个相同标签，浏览器只会先发送一个请求，等该请求返回后，在发送第二个，也就是说请求在浏览器端就已经被阻塞了。

所以《Node入门》一文中测试阻塞的方式不完全正确，文中描述的阻塞可能在浏览器端就被阻塞了。

正确的方法是在一个请求返回后不断刷新，或是开不同的浏览器进程。这是开十个woker的输出，可以看到十个woker完美并行执行。

A worker with #2 is now connected to 0.0.0.0:80
A worker with #4 is now connected to 0.0.0.0:80
A worker with #5 is now connected to 0.0.0.0:80
A worker with #8 is now connected to 0.0.0.0:80
A worker with #1 is now connected to 0.0.0.0:80
A worker with #3 is now connected to 0.0.0.0:80
A worker with #6 is now connected to 0.0.0.0:80
A worker with #7 is now connected to 0.0.0.0:80
A worker with #9 is now connected to 0.0.0.0:80
A worker with #10 is now connected to 0.0.0.0:80
Worker #1 has a request
Worker #7 has a request
Worker #6 has a request
Worker #3 has a request
Worker #8 has a request
Worker #5 has a request
Worker #4 has a request
Worker #2 has a request
Worker #10 has a request
Worker #9 has a request
Worker #1 makes a response
Time: 5003ms
Worker #1 has a request
Worker #7 makes a response
Time: 5005ms
Worker #6 makes a response
Time: 5008ms
Worker #3 makes a response
Time: 5004ms
Worker #8 makes a response
Time: 5008ms
Worker #5 makes a response
Time: 5008ms
Worker #4 makes a response
Time: 5004ms
Worker #2 makes a response
Time: 5001ms
Worker #10 makes a response
Time: 5000ms
Worker #9 makes a response
Time: 5000ms

另外，火狐浏览器在一个请求超过一秒并且失败时，会再次发送该请求，但不会发送第三次，所以在模拟宕机的测试中，开一个Firefox的标签，你会发现实际宕掉两个worker。