Just one "funny" with the historic graph, caused by the perennial problem with my weather station: registering a bizarre rain rate due to spurious tipping when I clean the tipping spoon. I'd hand-edited the various log.txt files in ~/Cumulus/data, so it didn't affect the record values. But I'd not done the .json files: alldailyraindata.json still had the rogue value.
It wasn't until I looked at the historic graph (Rainfall, rainfall rate graph) that I spotted this. I've now corrected that file and uploaded it to the web server, but I'm still seeing the spike.
So where does historic charts get its data from? Is it one of those things that will need Cumulus to be restarted?
It affects both the local :8998 site and my public web site.
UPDATE: Because Cumulus has just stopped when I was running the :8998 site (see separate thread), I've had to restart Cumulus and the corrected json file is now being read correctly. So that answers my question: the graphs used a cached version of the file, not the live one in the /web folder on the Cumulus computer (a RasPi) or on the public web server.