在某些自动化单元测试期间，我一直在使用Fabric在EC2实例上运行setup命令。这些测试运行好几个月，然后几天前，突然开始失败并出现错误：
Traceback (most recent call last):
File "tests.py", line 207, in test_setup
run('./setup_server')
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/network.py", line 682, in host_prompting_wrapper
return func(*args, **kwargs)
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/operations.py", line 1091, in run
shell_escape=shell_escape, capture_buffer_size=capture_buffer_size,
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/operations.py", line 934, in _run_command
capture_buffer_size=capture_buffer_size)
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/operations.py", line 816, in _execute
worker.raise_if_needed()
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/thread_handling.py", line 26, in raise_if_needed
six.reraise(e[0], e[1], e[2])
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/thread_handling.py", line 13, in wrapper
callable(*args, **kwargs)
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/io.py", line 31, in output_loop
OutputLooper(*args, **kwargs).loop()
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/io.py", line 152, in loop
self._flush(end_of_line + "\n")
File "/usr/local/myproject/src/buildbot/worker3/myproject_runtests/build/.env/local/lib/python2.7/site-packages/fabric/io.py", line 57, in _flush
self.stream.flush()
IOError: [Errno 32] Broken pipe
我正在运行的命令是pip install -r requirements.txt，它似乎运行得很好。如果我在本地运行测试，它会完成而不会出错。但是，每次在AWS上运行时，它都会失败一半。
这会导致什么？由于这是一个IOError，这些可能是由几乎任何类型的小网络中断引起的，我不确定如何诊断它。如果Fabric暂时失去连接，那可以解释它，但它无法解释为什么它是可重复的。我已经重新运行了这个脚本几十次，并且在最初完全连接到EC2实例后每次都失败了。
Fabric中是否有某种配置我可以改变以改善连接错误处理？