Graph labels question

Grafana Dashboards
,
1

  • What Grafana version and what operating system are you using? 12.0, Linux

  • What are you trying to achieve? Failure mode diagnosis in time series data (system CAN bus logs) for service, with basic anomaly detection and future integration into a customer facing agent.

  • How are you trying to achieve it? Host logs in database, review in Grafana dashboard with basic system state/metrics, preset graphs and DTC analysis. Results may be used directly for manual responses, or later parsed by an agent to create a response.

As I review my options with panels, I’m seeing a couple of issues -

  1. Editing Graph labels seems to be difficult. For instance, I cannot seem to figure out how to club multiple column values into the same label for a datapoint. How do developers get around such limitations?
  2. I aim to present the various system nodes as separate live blocks with system parameters either color coded or written inside based on cursor position in other existing graphs. What are my options for custom combinations of various graph types? For instance, rectangular gauge to represent how “full” the node is, with color showing temperature and text showing operation state.
  3. What are my options for using anomaly detection as input to generate a summary using an LLM agent? I understand Grafana Assistant might fulfil this, but would like to review all my options.

I apologize in advance if this is the wrong forum or format for this issue.

2

Welcome @GR_afana to the Grafana forum.

Re: your first 2 questions, what is your datasource, and can you share some of your sample data and queries you have created for this data?

Re: your 3rd question, have you read through this?

3

Hey there grant,

Currently experimenting with infinity datasource, but I’d like to explore all my offline options before committing. The data itself is just CAN messages in log files (ASC/BLF converted to CSV using python-can). And yes I’m aware of the Grafana Cloud option for outlier detection, but that will not be accessible to me in a week, and that doesn’t seem to allow contextual anomaly detection. Hope that helps.

4

So as you probably know, Infinity datasource is one of those datasource where the options are very extensive and grouping multiple values into the same label for a datapoint is likely possible using one of the tools. However, since you are using python already to convert the messages, perhaps doing the grouping there would be easier.

Can you share any sample data here so that me or others can paste into play.grafana.org and see what pans out?