Issue with windows timesync and "restart if data stops"

[SamiS]

I just noticed a minor issue with windows os automatic clock adjustment and "restart if data stops" -setting on cumulus. For some reason my pc clock had drifted so much forward, that windows adjusted it 10 seconds backwards. Bad luck was that the adjustment happened at 1:00:02, and rolled back to 0:59:52. This caused cumulus to think that it did not receive any new data at 1:00 and caused it to restart even while it already had received the "missing" data.

Diags folder attached.

[steve]

Yes, Mark Crossley spotted this recently - https://cumulus.hosiene.co.uk/viewtopic.php?f=4&t=11858

[SamiS]

Ok, thanks for a quick answer. I just recently switched from XP to 7 and had not noticed that win7 has a scheduled event of running timesync, but that definitely can cause issues. Maybe this could be worth mentioning on the readme? I think it is likely that this starts to raise questions more now when people are replacing XP:s with newer OSes.


PS. I'm not a real programmer, but depending on the structure of a program, shouldn't it be quite easy to "log" time when data was last read, and if clock jumps backwards while running, then do not trigger data missing until clock has gone over the "logged" time. Of course this depends on the internal structure of the code whether it is easy or not.

[steve]

It always makes me laugh when people tell me "it should be easy"

I'll see what I can do.

P.S. Your suggested algorithm is flawed, which kind of proves my point

[water01]

Alternatively do as I did. Go to the Task Scheduler (Programs/Accessories/System Tools/Task Scheduler, expand out the Task Scheduler Library, expand out Microsoft, expand out Windows, double click on Time Synchronization, right click on the Synchronization task in the top middle pane to Open a Window for editing, select the Triggers tab and change Weekly to Daily and select your time (I would suggest say 01.50.00 to stop it rolling back past an hour) and then OK your way out.

It is most unlikely that a daily update is going to allow the computer time to drift to more than a couple of seconds, whereas a weekly scheduler only means the clock has to vary by 1.45 seconds a day to exceed a 10 second change.

[steve]

Build 1094 should fix.

[SamiS]

It always makes me laugh when people tell me "it should be easy"

I'll see what I can do.

P.S. Your suggested algorithm is flawed, which kind of proves my point

Glad to make someone happy. But since you managed to publish a fixed version in less than 7 hours that kind of proves my point also.

Anyway, thank you once again.