Why is my grafana site missing data?

I’m having issues with intermittent data loss. Voltage measurements seem to be better than current measurements, but still not very consistent. Not sure if this is something that can be easily debugged for where the issue lies.

It might be explained by the data flow not being a normalised refresh, but a push on update. So if there is no change, then there is no new data sent.

The ‘per second’ is just an indication, some data points are quite infrequent, and some are just sporadic.

It can also be that your GX device CPU isn’t keeping up. Check the dbus round trip times to make sure there aren’t any spikes there.

Thanks, Ill check round trip times. Is that info available in the remote console of the GX device, or somewhere else?

It is logged in VRM and accessible as a custom widget in the Advanced section.

It should also be available in grafana in realtime, but it’s calculated by VRM Logger, hence no D-Bus path so I am not exactly sure. Easier to use VRM advanced for the clue if that is the issue.

So it does look like there are some spikes. Is this an issue?

No that is fine, 2ms is perfectly acceptable.

If it gets to the 100+ ms and beyond then there is an issue.

If that time period of 1-2ms dbus round trip times co-incides with when you’re seeing data missing it is not because of CPU load. But most likely either expected variation from when the data is pushed (it’s not periodic like VRM samples), or something else.

okay, so it might just be a limitation of grafana? Or do you think that this behavior is abnormal for grafana?

Sporadic reporting intervals is an inherent design decision of grafana / mqtt

It’s just passing the data straight through as it gets it without the GX device collating it, aggregating it, and normalising it in a packet like the system does for VRM.

There might be other reasons that data isn’t appearing, but it’s normal that it’s not a consistent flow and never will be.

You can use the features of grafana to provide averages and other statistical methods to infill the gaps if you want, but the fundamental data points elements are the way they are, as realtime as possible as they are changing.