Adding the required Kubernetes API so that the kubelet can start using
it. This patch also adds the corresponding alpha feature gate as
outlined in KEP 4639.
Signed-off-by: Sascha Grunert <sgrunert@redhat.com>
A Kubernetes Node requires to have at minimum one IP address
because those are used on the Pods field HostIPs and in some cases,
when pods uses hostNetwork: true, as PodIPs.
Nodes that use IP addresses as Hostname are interpreted as an IP
address, so it is possible that are nodes that don't hane any DNSname.
The feature gate AllowDNSOnlyNodeCSR will allow user to opt-in for
the old behavior.
Change-Id: I094531d87246f1e7a5ef4fe57bd5d9840cb1375d
This is the second and final step towards making kubelet independent of the
resource.k8s.io API versioning because it now doesn't need to copy structs
defined by that API from the driver to the API server.
This is a first step towards making kubelet independent of the resource.k8s.io
API versioning because it now doesn't need to copy structs defined by that API
from the driver to the API server. The next step is removing the other
direction (reading ResourceClaim status and passing the resource handle to
drivers).
The drivers must get deployed so that they have their own connection to the API
server. Securing at least the writes via a validating admission policy should
be possible.
As before, the kubelet removes all ResourceSlices for its node at startup, then
DRA drivers recreate them if (and only if) they start up again. This ensures
that there are no orphaned ResourceSlices when a driver gets removed while the
kubelet was down.
While at it, logging gets cleaned up and updated to use structured, contextual
logging as much as possible. gRPC requests and streams now use a shared,
per-process request ID and streams also get logged.
The manual deep comparison code is hard to maintain (would need to be updated
in https://github.com/kubernetes/kubernetes/pull/125488) and error prone.
In fact, one test case failed when doing a full automatic comparison with
cmp.Diff because it wasn't setting allMemory.
Recording the expected and actual checksum in the error makes it possible to
provide that information, for example in a failed test like the ones for DRA.
Otherwise developers have to manually step through the test with a debugger to
figure out what the new checksum is.
This reverts commit 0c0e19b343.
During stress test for SVM controller, the controller is unable to
make a list call due to following error:
resourceversion.go:155: I0716 21:49:26.973127] storage-version-migrator-controller: Error syncing SVM resource, retrying svm="crdsvm" err="error getting latest resourceVersion for stable.example.com/v1, Resource=testcrds: Timeout: Too large resource version: 28976, current: 20349"
With the feature disabled, the stress test passes.
Signed-off-by: Monis Khan <mok@microsoft.com>
