Probably a race while retransmitting exchanges

[aellwein]

Hi @chrysn,

while doing some performance tests with observation of about 3000 LWM2M (CoAP) clients against the server, i've encountered following error:

2016-08-29 16:28:54,992 [ERROR] Exception in callback Context._schedule_retransmit.<locals>.retr() at /usr/local/lib/python3.5/site-packages/aiocoap-0.2-py3.5.egg/aiocoap/protocol.py:331
handle: <TimerHandle when=1745359.3182603796 Context._schedule_retransmit.<locals>.retr() at /usr/local/lib/python3.5/site-packages/aiocoap-0.2-py3.5.egg/aiocoap/protocol.py:331>
Traceback (most recent call last):
  File "/usr/local/lib/python3.5/asyncio/events.py", line 125, in _run
    self._callback(*self._args)
  File "/usr/local/lib/python3.5/site-packages/aiocoap-0.2-py3.5.egg/aiocoap/protocol.py", line 337, in retr
    self._retransmit(message, timeout, retransmission_counter)
  File "/usr/local/lib/python3.5/site-packages/aiocoap-0.2-py3.5.egg/aiocoap/protocol.py", line 344, in _retransmit
    exchange_monitor, next_retransmission = self._active_exchanges.pop(key)
KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 28534)

About 6 of 3000 clients failed to notify the server with observed changes. The error message is always from the same line, only the keys are always different, i've extracted them from the tracebacks:

KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 28534)
KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 27413)
KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 2950)
KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 7791)
KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 28881)
KeyError: (('::ffff:185.112.178.101', 5683, 0, 0), 13036)

I am not certain about it but i think it could be a race condition, if a key lookup fails (or maybe the exchange times out while so many clients are about to send the notifications to the server?).

Do you have any idea on this?

[aellwein]

just to mention, the CoAP retransmission parameters are set to their defaults (no changes there).

[aellwein]

The issue was produced by a wrong setup of my client. Not a bug.