Hi,
I’ve got a setup going now with multiple input nodes posting to several feeds in a self-hosted emoncms instance.
On the feeds page, I can see data going in and if I view a graph of feeds, I get some wiggly lines. However, hitting the API using feeds/data.json with the key, a valid feed ID and some sensible times (millisecond linux format) returns no data. If I “damage” the GET request, I get expected errors (invalid API key, non-existant feed id, too many data points - more than 8928) so it seems the API itself is working and can see datapoints, but won’t return them.
I can see on the initial summary screen that the “feedwriter” service is not running - is that going to have a bearing on this? If so, any clues as to why it may have stopped and how to kick it into life again?
How have you installed the system? If you can see data on a graph then data has reached the ‘feed’.
Generally I’d expect feedwriter to be running. It buffers data so reducing write load on the SD Card but graphs etc pull data off the written to file data.
Thanks Brian,
I did try seconds initially (because the time converter I found defaulted to seconds since 1970). When I didn’t get anything from the API I went looking and found an old archived thread on openenergymonitor.org that said to use millisconds (so add three zeros) so assumed that was correct. archived thread
I’ve just tried seconds again and found the same issue.
Hi (again) Brian,
I did a standard install on a LAMP stack running on a Ubuntu Core virtual HyperV server. Maybe feedwriter doesn’t need to run if it’s not on an SD card and it’s clever enough to know that.
If I use the times from the CSV view on the graph page, I do NOT get any data.
If I use the feed/list.json or feed/timevalue.json for the same feed ID, I DO get the expected value back.
I have indeed been through the feeds API help page (it’s how I got my API read key). timevalue, value and fetch all seem to work, but data.json does not. It’s only time parameters that are added, so I assume something is up with them.
As I said, I can get error messages from the API by including ridiculous values in the time parameters and/or interval parameter (so I can get the “too many values” error by either decreasing the start timestamp sufficiently, or decreasing the interval value). It’s just when the timestamp parameters are valid values, I get nothing back.
Ok great. What is the install config you used? I created the advice on an Ubuntu install, and I have it running here on a VM with Feedwriter enabled.
Can you click on the button Copy as Markdown next to Server Information on the Admin page and paste into a reply here please (no further formatting required).
Hi Brian,
Sorry, thought I’d try to start the service and editted my reply after you’d asked this. The result of using “restart” instead of “start” is the same though. The /bin/chown ${USER}… process is failing.
Ah, that looks like the badger! The username in the .conf file was an “interim” one I’d used while setting up and then renamed to match our naming convention. I’ve editted the .conf to include the new name and the service started OK. It’s showing as green on the admin dashboard now too.