flat lined

[daj]

I had an odd problem this morning -- one I have never seen before in Cumulus.

At around 4:50am my temperature flat lined (as did all the other data). A few hours later everything else seemed to recover but the Temperature stuck at 5C. I only noticed it late in the day when I thought it was odd the temperature had been the same most of the day.

At first I assumed my sensor had died, but on checking the console for the station the data on screen was correct, Cumulus was just not reading it. I closed down Cumulus, restarted and it read the correct data.

To recover, I went back to a backup at 4am and restarted Cumulus which read all the old data from the console so I did not lose anything -- phew

Any idea why it would do this?
(Some graphs attached)

[werribee_au]

David,

This happens to me randomly (albeit not too often lately) and I haven't been able to nail the cause. Although I have never seen some data recover and the temp staying flat-lined, that is strange! I do have a daily scheduled reboot each morning at 2am so the reboot would have got things going again before that may have happened.

View thread: https://cumulus.hosiene.co.uk/viewtopic.php?f=13&t=2298

The only thing I haven't tried is a reinstall of the OS on the notebook as it has a lot of redundant software on it from it's previous life as my work notebook (Mostly uninstalled) and now it is the dedicated station computer.

This problem only started after moving the notebook and station console to another part of the house which is about 5m closer to the transmitter (now about 8m away). At the same time I lifted the anemometer and weather vane another 5m so I don't know if this has been the cause (Due to the extra 5m of cable acting as an antenna picking up interference).

[daj]

Gerard

I don't think it is a problem with the location as the console was quite happy receiving data, Cumulus was just not reading it. My Cumulus runs relatively well. It has run on the same PC for almost two years without issue (I dont use the PC for anything else). I did have a run where the console lost communication for several hours at a time a few months ago but that seems to have stopped.

This one just seemed strange as the console was happy but Cumulus was getting nothing.

[steve]

All I can think of is that one or more of the data values were reading as invalid, so Cumulus was ignoring them all. Can you zip up and attach the diags folder, please?

Regarding the temperature value not recovering when the others did - are you using the EWtempthreshold setting by any chance? I think that could well be a bug in this situation as the temperature difference is calculated from the last good value. If it suddenly changed a lot when it recovered, it would trigger the threshold. It needs a bit more intelligence.

[daj]

Steve

Spot on with the EWtempthreshold so that explains the recovery of all but the temp around noon.

here is an extract of my debug log, which may be more useful that the diags folder.

It seems to go wrong around this line

-133.0203624 : 04:46:12 request EW data block, written = 9, addr = 000000

the addr continues as 000000 until

-130.4547280 : 11:53:48 request EW data block, written = 9, addr = 00A1B0

At which point it lost sensor contact for a short period which seemed to reset things

http://dl.dropbox.com/u/33488/debug-extract.log.zip
http://dl.dropbox.com/u/33488/Diags.zip

[nitrx]

are you using the EWtempthreshold setting by any chance?

EWtempthreshold ? I can't find anything about this and what does this mean..

[daj]

EWtempthreshold ? I can't find anything about this and what does this mean..

Slightly typo, sorry!

EWTempdiff
see http://wiki.sandaysoft.com/a/Cumulus.ini

Mine is set to 2.5, so if the temp jumps more than 2.5 degrees between readings it assumes an error. I was flatlining at 5 and then when it did get a reading it was 13 so it ignored it

[nitrx]

Thanks I was't aware of this feature , I really missed but it's availble , but you think it's buggy ?

[daj]

.... but you think it's buggy ?

Not at all, it does exactly what it is supposed to. My post here was more to find out why Cumulus suddenly stopped reading any data from the Console, but when it did it obeys my rule to ignore the spike in the temp (which in this case was valid, but Cumulus has missed it creeping up from 5 to 13).


There is another handy one....EWpressureoffset=x.y

We really should get one for the rain too

We all know that EW stations can be prone to a spurious chunk of data so this just means Cumulus ignores the data and hopefully the next one is good

[nitrx]

Oke its clear to me I wish I had found this in an earlier stage , still learning anyway

[steve]

I'm not sure whether it's related or not, but this is a problem:

-133.0203624

You're running with the debug log switched on permanently. The logging component logs the number of seconds (minutes?) since it started, this is the first number in each line. This value eventually wraps round (I can't remember how many bits it uses) and the logging component crashes - or in this case it seems it didn't crash but perhaps caused some other problem in Cumulus. But it does seem that it would have gone negative some time before the problem occurred, so may be a red herring.

The station apparently stops responding, interestingly at the point where it is moving on from address 9F00, so maybe it had some sort of internal problem. Cumulus is trying to read the first block of memory and nothing is coming back.

At 11:53:48 it starts responding again, but the data blocks are out of sync, and at 11:53:52 it starts behaving again and all is back to normal (except for the issue with the temperature difference filter).

[daj]

thanks Steve -- I guess we will probably never know. I just thought i would post in case of interest.

I didn't realise Debugging was running until I checked later (a 380Mb file). I rarely look at the PC running Cumulus and it has been running for weeks without the need for me to do anything (which is a good thing of course).

[steve]

thanks Steve -- I guess we will probably never know. I just thought i would post in case of interest.

Yes, it's always worth looking at as there are a number of ways that a 'flat line' can occur, and investigation might lead to an improvement in Cumulus. But in this case I think something just went funny in the station (or the USB somewhere). At some point I need to have a look at detecting and signalling this sort of problem.