deployment view about grafana-dashboards-kubernetes HOT 7 CLOSED

Hi @pid42,

I don't think a deployment dashboard would be relevant, especially because it would mean make one for Statefulsets and Daemonsets too.

You already have the number or replicas available and unavailable in the "namespace" view.
If this is not enough, can you please give more information on your use-case?

from grafana-dashboards-kubernetes.

Closing this issue

from grafana-dashboards-kubernetes.

HI @dotdc, thanks for these charts, they amazing!

Would such a view not be beneficial to track the behaviour/performance of a specific application which spans multiple pods? Small use-case is to verify load is correctly being distributed across all pods in a deployments and not a single pod having massive spikes or consuming higher resources.

I currently use the namespace view to achieve this, but since I have thousands of pods, the CPU & Memory usage by pod is just a massive colouring board :) I have to manually select all the pods linked to a deployment in both charts to investigate certain application pods

from grafana-dashboards-kubernetes.

@Piroddi @pid42 I'll see what I can do for this use-case.
Re-opening

from grafana-dashboards-kubernetes.

I added a new created_by variable in k8s-views-namespaces.json to select a specific set of pods.
This will only affect the "Resources" and "Kubernetes Resources" rows.

The implemented solution should work for Deployment, StatefulSet and Daemonset.

@Piroddi @pid42 Let me know what you think and if you need anything else.

from grafana-dashboards-kubernetes.

Closing

from grafana-dashboards-kubernetes.

Hi @dotdc, I'm just testing out how well this can be used to drill down on a particular type of application.
It turns out, because of our rapid development cycle, there are constantly new revision rolled out for a single deployment.

There are two problems we're facing:

1. the created_by filter has a lot of options because of the number of replicas and it's hard to view it for one deployment to view. We have to select all of the replicasets tied to the deployment.
2. the graphs are very difficult to read because they are currently summed by pod, and if we have many replicas, it is pretty much not viewable. (See screenshots)

Not sure how we should go about 1.

For 2, if we want to get a better sense of usage of one single deployment (doesn't matter how many pods), shouldn't it sum the timeseries up by (created_by) instead?

CPU

Memory

from grafana-dashboards-kubernetes.