Bug: Data source names seem to be used as UID's in code if data source is set in a dashboard by a dashboard variable

[GrgDev]

I tried using grafana-wtf recently with my instance of Grafana 11. However, I will get a giant list of warnings whenever I run any explore commands. They all mention basically the same thing, that a data source could not be found. (I had to replace the real names with placeholders here, but other than that, this is from an actual production instance.)

2025-03-28 15:48:16,170 [grafana_wtf.core                    ] WARNING : Data source 'datasource-1' not found
2025-03-28 15:48:16,170 [grafana_wtf.core                    ] WARNING : Data source 'datasource-2' not found
2025-03-28 15:48:16,170 [grafana_wtf.core                    ] WARNING : Data source 'datasource-3' not found

I went into the source code to see where this error is coming from and it seems to be coming from here.

grafana-wtf/grafana_wtf/core.py

Line 719 in 77aec2b

log.warning(f"Data source '{ds_uid}' not found")

I see this line

log.warning(f"Data source '{ds_uid}' not found")

So it should be printing the UID, not the name, right? In my examples above, datasource-3 should have a UID of -iifrMA4k while datasource-2 should have a UID of YN2vilf7k. Yet that's not what is being printed. Is this a Grafana 11 specific issue?

[GrgDev]

I just took a quick second look at the grafana-wtf code where this warning comes from. I now realize it is likely only giving me these errors on dashboards that use a dashboard variable to define the data source. Almost all of my dashboards do this. I set a data source type of "Prometheus" for the dashboard data source variable and use a regex to match the set of data source names I want to be able to use for the dashboard.

[GrgDev]

In the meantime, is there a way to make grafana-wtf basically ignore data sources if all I care about is doing a string search?

[amotl]

Hi Greg,

thanks for writing in. What do you think about this little patch, created without thinking too much about it?

• Relax logging when dashboards defined as variables can not be found #188

With kind regards,
Andreas.

[amotl]

@nikodemas and @cronosnull: Do you think the patch GH-188 is viable for this occasion, or just plain wrong, and we would need to dig deeper?

[cronosnull]

@amotl what about using datasource_by_name as a fallback when datasource_by_uid gets None?

grafana-wtf/grafana_wtf/core.py

Line 717 in 644eedc

datasource = self.datasource_by_uid.get(ds_uid)

I think the values in the variables are data source names, but not completely sure, so I would support both cases.

[nikodemas]

@amotl I am not sure if I agree with #188 - if someone only has dashboards that use dashboard variable as described by @GrgDev, explore would not return anything and wouldn't even print any warning this way confusing the user even more. Looking at the issue description I would agree that the suggestion from @cronosnull might make sense, but I haven't tested it out.