"data stopped" rollover problem
Posted: Thu 10 Oct 2019 11:02 am
Running MX build 3052 on GNU/Linux, attaching to a Davis Vantage Pro2.
Connection to the VP dropped this morning at 08:32 UTC. I didn't notice until 09:30 UTC. On noticing the problem I made sure that the device was present (ls -l /dev/ttyU*) and restarted MX. On checking in the interface it appeared that all was well when looking at the 3-day graphs (temperature, humidity, wind, etc.). However when I looked at the daily temperature and rainfall graphs I noticed that there were two entries for 9th October 2019.
As my rollover is at 09:00 UTC I suspected a problem in dayfile.txt, and sure enough there were two entries for the 9th. The first entry contained data for 09:00 on the 9th until 08:31 on the 10th. The second entry contained data from 08:32 on the 10th until 09:00 on the 10th. I inferred this from the max/min values of temperature and pressure.
Looking in MXDiags files for before the restart I could see that, despite the lack of data coming in, the end-of-day rollover still occurred at 09:00 UTC. Looking in MXDiags files for after the restart there was another end-of-day rollover when the missing data from 08:32 to 09:00 was extracted from the logger and processed. It looks as if this is the reason for the two entries for the 9th in dayfile.txt.
I remedied the situation by stopping MX, copying the backup data files from the daily backup made at 09:00 on the 9th to the data folder, and restarting MX. This gave a very good test of the changes made recently to the serial protocol code, as 25 hours of 1-minute logger data were extracted and processed very quickly and flawlessly. Thanks @mcrossley for these recent changes, as I would've probably lost a day of data from my records otherwise.
Connection to the VP dropped this morning at 08:32 UTC. I didn't notice until 09:30 UTC. On noticing the problem I made sure that the device was present (ls -l /dev/ttyU*) and restarted MX. On checking in the interface it appeared that all was well when looking at the 3-day graphs (temperature, humidity, wind, etc.). However when I looked at the daily temperature and rainfall graphs I noticed that there were two entries for 9th October 2019.
As my rollover is at 09:00 UTC I suspected a problem in dayfile.txt, and sure enough there were two entries for the 9th. The first entry contained data for 09:00 on the 9th until 08:31 on the 10th. The second entry contained data from 08:32 on the 10th until 09:00 on the 10th. I inferred this from the max/min values of temperature and pressure.
Looking in MXDiags files for before the restart I could see that, despite the lack of data coming in, the end-of-day rollover still occurred at 09:00 UTC. Looking in MXDiags files for after the restart there was another end-of-day rollover when the missing data from 08:32 to 09:00 was extracted from the logger and processed. It looks as if this is the reason for the two entries for the 9th in dayfile.txt.
I remedied the situation by stopping MX, copying the backup data files from the daily backup made at 09:00 on the 9th to the data folder, and restarting MX. This gave a very good test of the changes made recently to the serial protocol code, as 25 hours of 1-minute logger data were extracted and processed very quickly and flawlessly. Thanks @mcrossley for these recent changes, as I would've probably lost a day of data from my records otherwise.