From: Gregg L. <gr...@li...> - 2009-04-18 19:08:49
|
Marc MERLIN wrote: > I'm trying to debug my house and I'm honestly not sure what the subjective > threshold between ok and not ok is > Note, my setup makes a full bus rescan (as in loop that requests state of > each device) every 15mn so I have more chance to generate those than you :) > > Also, to debug, I change stuff and just tell mh to ask all the devices for > their state above. Then I could how many erros I got. > > I just have no idea how many background noise/errors are reasonably > acceptable, if any. So, here's the deal... If you barrage the powerline w/ a lot of traffic, you will get collisions and those will result in resends. The trick is to try to avoid it to the extent possible. The initial start-up status scan is deliberately done asynchronously--which means that the opportunity for collisions is greatest. I don't worry about that as I may get a couple of collisions, but the resend(s) always gets through. Now, what I would be *very* concerned about is any significant number of resends while doing memory operations--which are synchronous. I'm mentioning the above because your request for data would me meaningless unless it was properly qualified w/ the conditions in which the resends occur (and, those can't be easily extracted from logs). I can tell you anecdotally that I never get resends when doing any memory operations and routinely get 2-3 resends when the "barrage" status scans are performed. If you see the same results, then you are fine. Also, realize that I run w/ the fastest settings possible (i.e., no throttling and minimal, inter-send delays). One area that I frequently see a resend is associated w/ the triggering of a light based upon an insteon motion sensor. This is because (for some who knows why reason), the motion sensor sends two reports w/ a slight delay (less than a second) between them. The state is set on the first report. So, if I then immediately turn on a light, then there is an opportunity for the light's insteon ACK to collide w/ the second motion sensor report. That's quite annoying. Possibly, I'll work on something to address it as a part of the second phase of refactoring. Gregg |