slave client_id collisions in large environments

[mattbailey]

I'm running 64 locust slave processes per host, and getting client_id collisions regularly.

locust/locust/runners.py

Line 355 in 4bde2ca

self.client_id = socket.gethostname() + "_" + md5(str(time() + random.randint(0,10000)).encode('utf-8')).hexdigest()

The randomint here probably increases the chances of collisions, since it's added to a millisecond scale time function.

Recommend using uuid4() or uuid5(namespace, name) using the hostname as a namespace.