Consider removing stacking from stacked graphs

[jdbaldry]

I find the arguments in https://everydayanalytics.ca/2014/08/stacked-area-graphs-are-not-your-friend.html to be very convincing and think it can be hard to interpret stacked graphs beyond seeing what the total value is. Some of our stacked graphs are more readable than others. It can be a useful visual cue when in the "Blocks loaded / sec" panel in the "Queries" dashboard is all red but it is difficult to interpret others like "Data fetched / sec" panel in the same dashboard.

Below is a quick grep for all the $.stack references in the repo.

$ rg  --no-heading --line-number --with-filename --color=always --ignore-case -- \\\$.stack .
./cortex-mixin/dashboards/config.libsonnet:13:        $.stack +
./cortex-mixin/dashboards/config.libsonnet:22:        $.stack +
./cortex-mixin/dashboards/writes-resources.libsonnet:64:        $.stack +
./cortex-mixin/dashboards/writes-resources.libsonnet:73:        $.stack +
./cortex-mixin/dashboards/reads.libsonnet:150:        $.stack +
./cortex-mixin/dashboards/object-store.libsonnet:12:        $.stack +
./cortex-mixin/dashboards/object-store.libsonnet:26:        $.stack +
./cortex-mixin/dashboards/alertmanager-resources.libsonnet:48:        $.stack +
./cortex-mixin/dashboards/alertmanager-resources.libsonnet:54:        $.stack +
./cortex-mixin/dashboards/alertmanager-resources.libsonnet:66:        $.stack +
./cortex-mixin/dashboards/alertmanager-resources.libsonnet:75:        $.stack +
./cortex-mixin/dashboards/reads-resources.libsonnet:111:        $.stack +
./cortex-mixin/dashboards/reads-resources.libsonnet:120:        $.stack +
./cortex-mixin/dashboards/writes.libsonnet:238:        $.stack + {
./cortex-mixin/dashboards/dashboard-utils.libsonnet:109:    $.stack + {
./cortex-mixin/dashboards/dashboard-utils.libsonnet:120:    $.stack + {
./cortex-mixin/dashboards/dashboard-utils.libsonnet:263:      $.stack +
./cortex-mixin/dashboards/dashboard-utils.libsonnet:305:      $.stack +
./cortex-mixin/dashboards/compactor-resources.libsonnet:24:        $.stack +
./cortex-mixin/dashboards/compactor-resources.libsonnet:30:        $.stack +
./cortex-mixin/dashboards/compactor-resources.libsonnet:42:        $.stack +
./cortex-mixin/dashboards/compactor-resources.libsonnet:51:        $.stack +
./cortex-mixin/dashboards/queries.libsonnet:68:        $.stack,
./cortex-mixin/dashboards/queries.libsonnet:192:        $.stack +
./cortex-mixin/dashboards/queries.libsonnet:198:        $.stack +

Ripgrep finished at Mon May 17 11:38:59

[pracucci]

I think it depends on a case by case basis. I agree with you that in some cases the stacked panel makes reading the data harder (eg. when you look for outliers) but I just would avoid doing a "blind" global replacement. I'm 💯 up for checking current stacked graphs and change them when appropriate.