Open jonnystorm opened 7 years ago
Case in point: once an :snmpm
agent is registered, we provide no API to unregister it, even if we registered with incorrect credentials. This makes it impossible to test multiple credentials without restarting the entire application each time. Untenable.
What are all those agents doing? Are they being used? Should we deregister them? When might we do that? Are they even named sensibly?
Maybe all that registering should be handled through a maintenance task. Being in-line, it would have all the information necessary to act on our behalf. Aging policy would need to be configurable. My gut tells me we should kill old agent processes based on average number of requests per minute, to avoid killing a much (albeit burstily) used agent. But for now, just checking for a non-zero request count would suffice.