fix(dashboards): Port compute/node (pods) dashboard to new grafonnet #912
Conversation
dashboards/resources/node.libsonnet
Outdated
| // This effectively "hides" max capacity from the panel. It shows only in the legend, showing the max capacity. In the "legacy" graph panel | ||
| // the max capacity could be reselected and thus shown on the timeseries with a dotted line. This is no longer possible. | ||
| + fieldOverride.byName.withProperty('custom.hideFrom', { tooltip: true, viz: true, legend: false }) |
There was a problem hiding this comment.
Seeking feedback on this from @skl and maintainers.
For many of the timeseries panels which show requests and limits as visible dotted line thresholds on the panel, the "capacity" on this dashboard, and the compute/namespace (pod) are only shown in the legend, and can not be "unhid" to show on the timeseries.
Is this desirable? Should the capacity be shown in the graph?
In under utilized environments (like mine) the capacity is dramatically higher than the stacked usage, so you would want to hide it to be able to reason on the actual utilization.
There was a problem hiding this comment.
I'm ok with just seeing the max cpu/memory capacity of the node in the legend like this, especially as that value probably changes rarely.
For k8s requests/limits/quotas though, I think it's useful to see them plotted as series on the graph as these can change much more frequently.
@povilasv be good to get your opinion here, too.
There was a problem hiding this comment.
For me the max capacity is very useful, but I usually run very utilized environments 🤔
There was a problem hiding this comment.
Given that disabling the visualization in the dashboard code makes it impossible to show the capacity on the timeseries, I've brought it back as a default (see below).
If a user wants to see a specific pods utilization, they can select it in the legend, or jump to it from the table(s).
I think this visualization makes the most sense as "how much of my capacity is used", so a giant gap between the utilization and the upper capacity threshold is just as meaninful.
|
@skl could you also review when you have time? 🙇 |
Part of:
Following on from:
BEFORE
AFTER
