sergeskor / ssvlongtime Goto Github PK

We are using this library on 50+ devices (ESP32 and STM32)

Just occasionally we see a jump in uptime, e.g, I just observed one ESP32 device going from 2907s (~48mins) to 4297876s (~7weeks)

Looking at the code I see a similar number const uint32_t multiplier = 4294967; and if I add that to my original value of 2907 I get 4297874 which is very close to the inflated uptime I'm seeing!

I'm guessing the multiplier is being applied when it should not be!?

We do not see this behaviour often, maybe seen 20 times over the 50 devices over the last year or so.

This device in question gets reset several times a day, avg uptime is between 1 and 15 hours, this has not happened on this device in its 90days of operation.

Some device run almost constantly while others are reset quite frequently.

Looking back through our last 90 days data, I see another ESP32 jumping form 20 weeks to 27 weeks. Some seem to jump just at random times, others not jumping at all, and I see one device has done this 4 times always around the 2 day mark.

Each of our boards contain one ESP32 and one STM32. This issue seems isolated to the ESP32s. I can not find this happening on any STM32s.

Very strange. Any ideas?

I am maybe thinking power supply / noise issues? All these devices are situated around combustion engines powered from the alternator/battery. Although we do have good regulators.

Or maybe a threading issue. The ESP32 is dual core and we are using threads. Could a thread sync/timing issue be causing newMillis < lastMillis just by 1 and triggering overflow when it shouldn't be?

Thanks.