[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [xmlblaster-devel] how to diagnose subscriber that stops getting messages after a while?



Brad Clements wrote:
Ok, these clues have helped me track this down.

My Python socket protocol client is apparently not 100% finished.

If an exception is raised in application code handling an message unit, what should the client layer do?
See http://www.xmlblaster.org/xmlBlaster/doc/requirements/interface.update.html#exception

I think you should catch everything and do whatever is appropriate for your clients use case.
In your case i would send a special note to nagios that the heartbeat client module is down.
Currently it falls down. the socket pumper thread is killed and inbound messages are dead. meanwhile outgoing messages keep working, hence heartbeat survives and the client is "alive" from xmlblaster's standpoint.

--

from an api preference standpoint, if the application code raises an exception while processing a message, should the client layer generate an exception message unit and send it back to xmlblaster?
See above. What could xmlBlaster do with your exception?
It is a client side problem, so i think you need to address it there.

regards
Marcel
Or should it print some diags and kill the entire connection (as if the server had terminated it).. in which case the application detects that and restarts.

I'm not sure what the best default behaviour should be. I think sending an exception back to the server is correct, but who'se going to notice that?