meteohub.de

AAMFlyer wrote:Couldn't you just do every 14 or every 16 minutes

Certainly - but I wondered if there was a neat way to achieve 'every 15 minutes' at times other than 0, 15, 30 & 45.

There's no shortcut method that I'm aware of -- I just did the 4,14,24,34,44,54 method for my own graphs.

However, the more that I think about this, there is one thing that is a bit perplexing about it all -- the Meteohub system is still drawing the same number of graphs over a 15 minute load period as it was when I had them all defined as */10. However, staggering seems to reduce the load despite the fact that with */10 there was more "idle time" between cron requests which should give the system time to catch up and finish the graphs anyway.

Perhaps the staggering method really just tricks our minds into thinking the load is more spread in the short term (over the 1 and 5 minute time-frame) while really the system is under the same overall pressure regardless of if you queue up a lot at once, followed by a long break to catch up, or if you queue things up slowly as other tasks are completed.

In the end, on the NSLU2 it'll always come down to a question of priorities in order to keep the system from becoming overloaded. For instance, I disable the drawing of daily solar and UV graphs late at night, and in their place I have the system update weekly, monthly and yearly graphs instead.

sevenless wrote:There's no shortcut method that I'm aware of -- I just did the 4,14,24,34,44,54 method for my own graphs.

Okay. Worth asking. :)

Perhaps the staggering method really just tricks our minds into thinking the load is more spread in the short term (over the 1 and 5 minute time-frame) while really the system is under the same overall pressure regardless of if you queue up a lot at once, followed by a long break to catch up, or if you queue things up slowly as other tasks are completed.

It may well be that there are other factors beyond out control, perhaps some internal 'load balancing' mechanism?
I've certainly noticed that html template uploads don't seem to take place right at the time specified by the CRON job. The file doesn't appear on the server till a few minutes later - though when it does it has the date/time you'd expect from the CRON settings.

I've certainly noticed that html template uploads don't seem to take place right at the time specified by the CRON job.

That puzzled me for a while too, but then I discovered this in section 2.11 of the manual:

An FTP Upload is done by a time scheduler every 5 minutes.

And sure enough, when I check the meteohub logs, ncftpput is called roughly that frequently and simply uploads everything in the Uploads folder. Thinking about it this makes a lot of sense as it reduces system load by queuing up files and then uploading them all at once, rather than sending each one individually with all the associated authentication and protocol chatter.

Presumably the time stamp of the files is correct because the file is created at the time specified by the cron schedule.

This of course makes for a second interesting observation -- a cron schedule of less than five minutes for most files is pointless! (Of course the manual is silent on how WD-Live files are handled, and since Boris initially designed that to be created every minute I suspect he put in a provision for those to be delivered more frequently, but...???)

sevenless wrote:Of course the manual is silent on how WD-Live files are handled, and since Boris initially designed that to be created every minute I suspect he put in a provision for those to be delivered more frequently, but...???

And there seem to be other internal schedules too - e.g. (when relevant) the WSWIN 'sync' data file import.csv gets produced every 10 minute in record blocks always ending with a '?7th' minute record - e.g. 15:08 to 15:17, then 15:18 to 15:27.

Just a quick update - had a little more rain and it updated overnight - but the annual/monthly records are still wrong.....

Data is uploaded every minute now (if there is something to upload)

when you are wondering about timing of the jobs, you can login via ssh and have a look at the crontab that does all the scheduling. Meteohub also puts in some "sleep" to avoid starting all jobs at exactly the same time, which would overload the NSLU2.

Are the rain rate records still wrong?
Or do you mean total rain reports in the WDL graphs?
There will be changes with both in 2.7a, so waiting a day for this to be released might be an option.
Things like "highest rainfall during a day" for this month/year are not supported by meteohub at the moment (even not with 2.7a).

That sounds great on the rain updates!!! For the records that are wrong/marked red - those look like what your referring too - Highest Daily Rain and Highest rain in 1 Hour....

A though - if Meteohub currently doesn't support those yet - perhaps they can just be zero'd out or taken out.... Just a thought - but I do notice that they have red dots next to them - does that mean not data.....?

Thanks

I my comment above per the red and green dots - perhaps just an easy legend on what a red dot meants etc.... Thanks!

Gregg