Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Bug 1985073: use 1m resolution for control plane cpu alerts #1201

Merged
merged 1 commit into from
Aug 9, 2021
Merged

Bug 1985073: use 1m resolution for control plane cpu alerts #1201

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
4 changes: 2 additions & 2 deletions bindata/assets/alerts/cpu-utilization.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ spec:
To fix this, increase the CPU and memory on your control plane nodes.
expr: |
sum(
100 - (avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100)
100 - (avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[1m])) * 100)
AND on (instance) label_replace( kube_node_role{role="master"}, "instance", "$1", "node", "(.+)" )
)
/
Expand All @@ -42,7 +42,7 @@ spec:
kube-apiservers are also under-provisioned.
To fix this, increase the CPU and memory on your control plane nodes.
expr: |
100 - (avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100) > 90 AND on (instance) label_replace( kube_node_role{role="master"}, "instance", "$1", "node", "(.+)" )
100 - (avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[1m])) * 100) > 90 AND on (instance) label_replace( kube_node_role{role="master"}, "instance", "$1", "node", "(.+)" )
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

While we're refactoring this, I think it's easier to read if we rephrase from 100 - (avg idle) > 90 to avg idle < 10.

Also, this still doesn't account for holes in the node_cpu_seconds_total metric. My understanding of the rate call is that if the covered minute has any node_cpu_seconds_total data, but node_cpu_seconds_total is ticking up at 20% for 10s, while node_cpu_seconds_total is missing for the other 50s, it will look like 0.2 * 0.1 = 0.02 = 2% idle, so the expr would match despite 20% being > 10%

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it will look like 0.2 * 0.1 = 0.02 = 2% idle

rate does extrapolation based on the slope of the first and the last sample under the window.
rate also avoids extrapolating too far, extrapolation extends to half the sample interval when the first or the last sample is too far away from the window
so prometheus does not handle the missing data/gap as above. since rate already extrapolates for us i don't see it's necessary for the alert to take into account any gap in its calculation.

(slack thread where we discussed it - https://coreos.slack.com/archives/C01CQA76KMX/p1626752724386400)

for: 5m
labels:
severity: critical