Re: [Owfs-developers] bus interface statistics

[Stefano M. <mo...@ic...>]

> On 13 Jul 2019, at 22:36, Mick Sulley <mi...@su...> wrote:
> 
> Thank you both for your input.  I agree, I do not like the scheduled power cycle option either and I continue to look for the root cause of the issue.
> 
Let me add just a few more comments: 100% reliability is of course not possible, therefore a watchdog timer may still be useful.
> The reason I considered a scheduled power cycle is that it seems that after a power cycle I do not see any errors, then over a few days or weeks I start to see errors, 85 reads, device not found, etc and then I get a lockup, although the timing is very variable.
> 
If you experience a system degradation before lockup, then you can setup a daemon that monitors the 1-wire vitals, and eventually restarts the system when some threshold value is reached.
> I do have heartbeat file which is monitored by another machine, so I will look at using that to power cycle it.
> 
The RPi chip has an hardware watchdog timer, which you can access via /dev/watchdog. Unfortunately the RPi is not able to power cycle itself, but only to reset, so some external board/system is needed to power-cycle the RPi.
> One other thought, I have separate power supplies for 1-wire and the Pi.  Can I just power cycle the 1-wire adapter and leave the Pi running?
> 
I have no experience with the DS2482-800 and with HW design, but I understand that you have to power-cycle the 1-wire adapter, no need to cold reboot the RPi host. But maybe I’m mistaken here.

S.

> 
> On 13/07/2019 19:58, Stefano Miccoli via Owfs-developers wrote:
>> 
>> 
>>> On 11 Jul 2019, at 23:10, Mick Sulley <mi...@su... <mailto:mi...@su...>> wrote:
>>> 
>>> The reason for the question is that I still have random bus lockups and I am considering creating something to power cycle the system, either on a time basis, e.g. 3am each day, or based on some early warning detection from the data in interface/statistics if that is possible.
>>> 
>>> Does anyone have an opinion on scheduled power cycle?  Good idea or not?
>> 
>> This make sense only if you are sure that the bus lockups are **not** random, but somehow occur only after some time has elapsed from the last power cycle, and this time is longer than one day.
>> 
>> On the contrary if the lookups are truly random, then a reboot every 24h just ensures that the longest down-time is less than 24h. If it is impossible to avoid random lookups then the most sensible solution would be a watchdog timer. This way you  can ensure that bus down time is shorter that the watchdog time interval itself.
>> 
>> Stefano
>> 
>> 
>> 
>> _______________________________________________
>> Owfs-developers mailing list
>> Owf...@li... <mailto:Owf...@li...>
>> https://lists.sourceforge.net/lists/listinfo/owfs-developers <https://lists.sourceforge.net/lists/listinfo/owfs-developers>
> _______________________________________________
> Owfs-developers mailing list
> Owf...@li...
> https://lists.sourceforge.net/lists/listinfo/owfs-developers