Description
1. Proposed title of this feature request
What problem/issue/behavior are you having trouble with? What do you expect to see?
Kubelet's --runtime-request-timeout defaults to 2m0s
https://kubernetes.io/docs/reference/command-line-tools-reference/kubelet/
Crio's container create timeout is 4 mintues
https://github.com/cri-o/cri-o/blob/d57d2915beeb7f2d32813f4586342d860d710b4b/internal/oci/oci.go
// ContainerCreateTimeout represents the value of container creating timeout
ContainerCreateTimeout = 240 * time.Second
When we try to update this value via a KubeletConfig object (special type of MachineConfig) we get the following error.
Error: KubeletConfiguration: runtimeRequestTimeout is not allowed to be set, but contains: 4m0s
What is the business impact? Please also provide timeframe information.
Too many pods starting up on a single node cause many of them to hang or fail to be created.
2. What is the nature and description of the request?
We would expect these values to be the same or configurable.
3. Why does the customer need this? (List the business requirements here)
We have seen a number of times on very busy nodes kubelet logs stating that it's gives up on a container creation while crio is still starting it up. Unfortunately I don't have the logs handy because all clusters and pods are currently healthy but I will attach them next time we see it happening.
4. List any affected packages or components.