On 02/12/2010 10:08, Arjen de Korte wrote:
Citeren Arnaud Quette <[email protected]>:

Thanks for the suggestions, I've added the flush statement as well as some
debugging information. As this is a intermittent issue I decided to try
overloading the UPS by sending it repeated beeper commands while watching the debug output. What appears to happen is that the UPS returns an unknown
"~00R000" response. This means get_belkin_reply() returns -1, causing a
datastale state is set when called from do_status().

you should remove the datastale() call since upsd will automatically flag
the device as stalled if it has failed to update its data for 15 seconds
(default of MAXAGE).

Not at all!

The upsd server will only declare the *driver* stale if it fails to respond within MAXAGE seconds. However, as long as it keeps answering the PING from the server, it will not be declared stale. This mechanism is something completely different from what happens if the driver calls dstate_datastale(). In that case the driver tells the upsd server that the *UPS* fails to respond. See the chapter on "Staleness control" in docs/new-drivers.txt.

What really needs to be done, is that the driver doesn't treat the "~00R000" reply as an error condition. Apparently the UPS acknowledges the receipt of data, without further response (indicating that 0 bytes follow). The belkin driver doesn't accept this at the moment and requires that a reply follows. This is what needs to be changed.
Ah, I had actually thought about doing exactly that by allowing for a zero-length response, but wasn't sure what the affect of doing so would be. I'm going to have a look at the way in which other drivers operate.


Last but not least, in most drivers, we allow a couple of missed replies before we call dstate_datastale() so that glitches don't lead to automatic reconnects.
Can you suggest what driver would be a good template to use?


Best regards, Arjen


_______________________________________________
Nut-upsuser mailing list
[email protected]
http://lists.alioth.debian.org/mailman/listinfo/nut-upsuser

Reply via email to