Some dashboard panel selectors missing

[applike-ss]

Describe the bug
The following panels are missing any selector, which makes the dashboard consider dragonfly instances from other namespaces:

• FiberSwitchDelay
• Pipeline length
• Pipeline latency

To Reproduce
Steps to reproduce the behavior:

1. Deploy multiple dragonfly instances (different namespaces)
2. Deploy a prometheus + grafana
3. Add a monitor to fetch the metrics
4. Install Dashboard from repo
5. Check panels

Expected behavior
I should not see metrics from different namespaces, as those must by definiton be a different cluster.

Would be nice to have dragonfly clusters emit a label cluster which would be used to select metrics of a specific cluster.
That would make having multiple clusters in the same namespace possible too.

Screenshots

[romange]

Hi, thanks for filing the issue. Can you send a pr, please?

[applike-ss]

I'd fix the dashboard, which selectors do you find fitting? Should the advanced metrics panel group also filter based on the instance label?

To me it would be more usable if i could see the metrics of all pods of the same cluster at the same time.

What do you think?

[romange]

Where do you take this dashboard, actually? Is it from https://github.com/dragonflydb/dragonfly/tree/main/tools/local/monitoring ?

[applike-ss]

yes, that is the one

[romange]

So this one it totally for the community to decide. I mainly use it for local development, so I do not have strong opinions.

[romange]

@Pothulapati , @Abhra303 - what do you think?

[applike-ss]

pls check it out if you are interested: https://github.com/dragonflydb/dragonfly/pull/4165/files

[applike-ss]

I would still be interested to know from community side if there is anything against adjusting the dashboard so that we see metrics for the whole cluster and don't have to select a specific pod from the statefulset.

IMHO that makes sense to get a good overview about what is going on in the cluster (but still grouping metrics by pod).