Add Kepler to gather energy usage data #530
Conversation
petewall
force-pushed
the
feat/kepler
branch
4 times, most recently
from
476db3a to
b282d00
Compare
|
Ignore the failing |
petewall
requested review from
jjo,
nikimanoledaki,
a team,
nevermind89x and
jewbetcha
and removed request for
a team
Signed-off-by: Pete Wall <[email protected]>
|
|
||
| # @ignored -- This skips including these values in README.md | ||
| canMount: | ||
| usrSrc: false |
There was a problem hiding this comment.
values.yaml does have a nodeSelector field defaulting to
nodeSelector:
kubernetes.io/os: linux
, suggest adding also amd64 constraint, as per https://kubernetes.io/docs/reference/labels-annotations-taints/#kubernetes-io-arch
kubernetes.io/arch: amd64
There was a problem hiding this comment.
Yeah, I really thought about adding this. My hesitation is that there are ARM platforms that are supported, and I don't want to cut them out.
Is it easier to have the amd64 restriction and have to explain how to remove if you have a supported arm node, or don't have the restriction and have a doc about why pods are crashing on graviton nodes. I'm leaning towards the latter, but happy to discuss.
There was a problem hiding this comment.
I looked a long time to determine if I can detect if the pod is on a graviton node, but I don't think it's possible at pod scheduling time.
Signed-off-by: Pete Wall <[email protected]>
This PR adds Kepler as another subchart dependency. This means that we can optionally deploy Kepler and tell Alloy to scrape metrics from it.
Notes to reviewers:
metrics.keplersection controls how we configure alloy to scrape keplerkeplersection controls how kepler is deployed..githubare for actions, mostly around the tooling that keeps kepler helm chart up to date.charts/k8s-monitoring/ci/*.values.yamlare the test files for our CI tooling. Kepler is enabled in one of those files and we add a check for a known metric to ensure that end-to-end delivery is workingexamplesare automatically generated