Ship a default Grafana dashboard for Kubeflow Trainer
- Dominant language
- Go
- Stars
- 2.2k
- Forks
- 1.1k
- Avg merge
- 3d 22h
- Merged PRs (30d)
- 39
Description
### What you would like to be added?
Ship a default Grafana dashboard as part of the Helm chart that provides out-of-the-box visibility
into controller health and TrainJob lifecycle.
This revives [#1376](https://github.com/kubeflow/trainer/issues/1376), filed in 2021 for V1 but
closed as stale.
### Challenges
- Dashboard depends on which metrics are available (controller-runtime defaults today, custom
metrics once the companion metrics issue lands)
- Needs to cover two personas: **platform operators** (controller health, reconcile backlogs)
and **ML engineers** (TrainJob lifecycle, time-to-running, failure rates)
- Should be gated in Helm values to avoid installing for clusters without Grafana
### Why is this needed?
### Current State
- No dashboard shipped with Trainer
- Operators must manually discover available metrics, write PromQL, and build panels
- Peer projects (Kueue, Argo Workflows) bundle dashboards - Trainer is a gap
### Related Issues
- https://github.com/kubeflow/trainer/issues/1376
- https://github.com/kubeflow/trainer/issues/1340
- https://github.com/kubeflow/trainer/issues/2648
- https://github.com/kubeflow/trainer/issues/2779
### Love this feature?
Give it a 👍 We prioritize the features with most 👍
Contributor guide
Assessment
This issue has not been assessed yet.