Splunk Apps » Insteon

No relevant logs. The input just seems to stop. The connection to the hub appears down too.

Theories:

1. The socket times out due to no activity
I might be able to handle with a TCP keep alive. I would expect that an exception would have been generated though.

2. An exception is being generated but not handled
I would think that the exception would be outputted.

3. The input is in a infinite loop
CPU usage seems fine though.

I suspect the issue is in the TCP interface read function that swallows socket exceptions.

Still happens.

Trying this method to detect down sockets: http://stackoverflow.com/questions/17386487/python-detect-when-a-socket-disconnects-for-any-reason

http://stackoverflow.com/questions/667640/how-to-tell-if-a-connection-is-dead-in-python

2014-12-27 17:12:03,006 ERROR Execution failed: Traceback (most recent call last):
  File "/Library/Splunk/splunk_sp/etc/apps/insteon/bin/insteon_app/modular_input.py", line 1127, in execute
    self.do_run(in_stream, log_exception_and_continue=True)
  File "/Library/Splunk/splunk_sp/etc/apps/insteon/bin/insteon_app/modular_input.py", line 1027, in do_run
    input_config)
  File "/Library/Splunk/splunk_sp/etc/apps/insteon/bin/insteon_plm.py", line 189, in run
    ready_to_read, ready_to_write, in_error = select.select([self.interface.__s,], [self.interface.__s,], [], 5)
AttributeError: 'TCP' object has no attribute '_InsteonPLMInput__s'

Monitoring the connection with this:

netstat -n | grep 9761

To test, I disconnected from the network for several minutes. I see that a connection is still established but no packets are being sent back and forth. It seems like the connection is broken from the hub-side and isn't being detected by Python.

Re-establishing a connection from another hosts succeeds which indicates that the connection really was dead. I'm going to leave the input running to see if it eventually detects the connection failure using the select.select method.

It's been 4 hours and my host still shows a connection to the PLM. I'm going to look into using a keep alive to help detect connection failures.

http://stackoverflow.com/questions/12248132/how-to-change-tcp-keepalive-timer-using-python-script

https://delog.wordpress.com/2013/08/16/handling-tcp-keepalive/

Implemented a method that bounces the connection if no activity is observed within a given time frame.