Thanks for the replies ref hardware and report intervals - I was running on a Pi A+ - turns out they halved the ram so its now been moved to a Pi3 and its running ok so far. I will look into a longer term upgrade but glad the Pi3 seems to be coping.
Andy On Tuesday, 5 February 2019 08:34:54 UTC, Andrew Milner wrote: > > Gary - the point I was really trying to make was that - especially where > an RPi is involved - it is likely that one cannot necessarily do all that > one wishes - but compromising does not necessarily mean a significant loss > of information. What people often forget is that midnight in particular > results in more load for reports as the yearly plots and pages are > regenerated together with the monthly, weekly, current etc also and this > can result in more issues when small archive intervals are being used. > > I think we are really on the same page Gary, even if I may appear a > luddite and a pedant at times. > > Andrew > > > On Tuesday, 5 February 2019 07:49:34 UTC+2, gjr80 wrote: >> >> I beg to disagree. I learnt a long time ago to not impose my views on a >> user's requirements; maybe the user needs to generate a report every >> minute, maybe they need to record the number of widgets every minute, maybe >> they just like lots of dots in their plots. At the end of the day it does >> not matter, WeeWX supports a 1 minute archive interval. Now if a user >> chooses to use a 1 minute archive interval that may impose some limits on >> the user, but that is a different matter. I know the old chestnut about >> wind speeds and directions gets dragged out every now and then. I think we >> need to keep in mind that the period over which an observation is made does >> not necessarily bear any relationship to how often you may wish to record >> the value. >> >> Gary >> >> On Tuesday, 5 February 2019 14:22:12 UTC+10, Andrew Milner wrote: >>> >>> I continue to find it almost impossible to conceive a situation where an >>> archive interval of under 5 minutes serves any useful purpose. As long as >>> a gust value is recorded for the 5 minute period the average speed >>> value/direction over the period is more than sufficient - even for an >>> airfield - and the other readings are unlikely to alter over 5 minutes >>> anyway. So with a Pi a 5 minute interval is more than sufficient. >>> >>> I found the following whilst googling wind gust which I thought useful >>> since the implication is that wind measurements over short periods of time >>> (if the rest of the world is right) are error prone anyway: >>> >>> "Wind gusts (which last only a few seconds) make it hard to determine >>> the overall wind speed of storms whose winds don't always blow at constant >>> speeds. This is especially the case for tropical cyclones and hurricanes. >>> To estimate the overall wind speed, the wind and wind gusts are measured >>> over some period of time (typically 1 minute) and are then averaged >>> together. The result is the highest average wind observed within the >>> weather event, also called the *maximum sustained wind speed*. >>> >>> Here in the U.S., maximum sustained winds are always measured by >>> anemometers at a standard height of 33 feet (10 m) above ground for a >>> duration of 1 minute. The rest of the world averages their winds over a >>> period of 10 minutes. This difference is significant because measurements >>> averaged over just one minute are about 14% higher than those averaged over >>> the course of ten minutes." >>> >>> >>> On Tuesday, 5 February 2019 01:29:23 UTC+2, gjr80 wrote: >>>> >>>> Yes the NoneType error is certainly due to the realtime clientraw >>>> extension not handling a None value correctly, by the looks of it a wind >>>> related field so quite possibly brought on by your battery/connectivity >>>> issue. That extension was written over 2 years ago and never formally >>>> released (and judging by the lack of user questions I assume not used by >>>> too many). So that means it is probably not too robust. It maybe easiest >>>> to >>>> just disable it until I can get to have a look at it in the next 2-3 weeks >>>> (as per other thread). >>>> >>>> Some good advice above about getting a machine to work reliably, the >>>> only thing I would add is chose your archive interval carefully as that >>>> also has an impact on stability when you start to add in a number of >>>> extensions. A 5 minute archive period means there is 5 minutes between >>>> report cycles so arguably WeeWX has around 5 minutes to get all of its >>>> report processing completed, cut that down to 1 minute and WeeWX now only >>>> has 1/5 the time. You don't want your reports taking longer to run than >>>> your archive interval, that would be bad on many fronts. Some folks think >>>> that having the shortest possible archive interval is essential, that may >>>> be the case in some circumstances but there are other secondary effects >>>> that need to be kept in mind. Basically, to run a 1 minute archive WeeWX >>>> needs to be fairly lean extension wise. >>>> >>>> Gary >>>> >>>> On Tuesday, 5 February 2019 08:19:58 UTC+10, mwall wrote: >>>>> >>>>> On Monday, February 4, 2019 at 4:42:52 PM UTC-5, Andy Hudson-Smith >>>>> wrote: >>>>>> >>>>>> Noted the use of skins and mqtt on a pi, i kind of assumed it was up >>>>>> to the job, i do like the weewx system so may well move up to a more >>>>>> useful >>>>>> machine - coming from Windows into this world means i know little beyond >>>>>> Dells and Win 10 which is what i have been trying to avoid due to >>>>>> frequent >>>>>> updates and crashes of other systems. >>>>>> >>>>>> Any advice on what sort of system weewx needs to run would be good >>>>>> (again i should Google this first!). >>>>>> >>>>> >>>>> here are some suggestions: >>>>> >>>>> https://github.com/weewx/weewx/wiki/hardware >>>>> >>>>> data collection is almost never a problem. uploading data is not cpu, >>>>> i/o, or network intensive. most reports will run just fine on a pi or >>>>> other low-power machine - you should have no problems with 1 or 2 >>>>> reports. >>>>> however, you might have issues if you try to run 3 or 4 query-intensive >>>>> reports with a short archive interval. >>>>> >>>>> in some configurations, the crt (cumulus realtime) and wdcr >>>>> (clientraw) extensions can be problematic on low-end hardware, since they >>>>> query the database each report cycle (archive interval) to get the data >>>>> they need. if you run them on archive intervals you should be ok, but if >>>>> you run them on loop packets you can easily run into the 'database >>>>> locked' >>>>> problems. >>>>> >>>>> the NoneType error is almost certainly due to a bug in rtcr.py - it is >>>>> getting a value of None but does not handle it properly. >>>>> >>>>> m >>>>> >>>> -- You received this message because you are subscribed to the Google Groups "weewx-user" group. To unsubscribe from this group and stop receiving emails from it, send an email to weewx-user+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
