Describe the feature
When running in kubernetes at scale, mittens makes the pod objects gigantic. This causes k8s clusters to fail/timeout/delay actions at high scale.
Here is one example of a use-case and how much data flows thru all cluster operators + api server itself.
kubectl -n my-namespace mypod -o yaml | yq .spec.containers[].args |wc -c
133100
Even when not degrading clusters, this leads to high cpu use and in some cases memory use on operators adding cost to operating.
It looks like this:
|
flag.StringVar(&t.HTTPProtocol, "target-http-protocol", string(http.HTTP1), "Protocol used for HTTP requests") |
could be dropped in favor of viper and allow both.
Rational
To keep clusters stable when running at high scale with a lot of churn.
Example Scenario
If it applies, use cases where the feature be used.
Describe the feature
When running in kubernetes at scale, mittens makes the pod objects gigantic. This causes k8s clusters to fail/timeout/delay actions at high scale.
Here is one example of a use-case and how much data flows thru all cluster operators + api server itself.
Even when not degrading clusters, this leads to high cpu use and in some cases memory use on operators adding cost to operating.
It looks like this:
mittens/cmd/flags/target.go
Line 47 in 6446db6
Rational
To keep clusters stable when running at high scale with a lot of churn.
Example Scenario
If it applies, use cases where the feature be used.