IPv6 should also be checked if it is globally enabled. On nftables, today this
is hardcoded, so if a Linux Kernel disables IPv6 during its boot or doesn't
have IPv6 compiled, it will still try to use IPv6, which can lead to some
unexpected errors.
This change verifies if IPv6 is enabled by checking if the IPv6 network interfaces
proc file is available
With StreamingCollectionEncodingToJSON and
StreamingCollectionEncodingToProtobuf, the WatchList must re-justify its
necessity. To prevent an ecosystem from building around a feature that
may not be promoted, we will stop serving list-via-watch until
performance numbers can justify its inclusion.
This also stops the kube-controller-manager from using the
list-via-watch by default. The fallback is a regular list, so during
the skew during an upgrade the "right" thing will happen and the new
StreamingCollectionEncoding will be used.
Choosing a port in advance is racy. A better solution is to use a Unix Domain
socket in the per-etcd-instance data directory. Then the name can be determined
in advance and there's no risk of conflicts with other etcd instances.
With unix:// for the endpoint, we have to be a bit more careful about
passing a TLS config to the etcd client library because for unix://, in
contrast to http://, it tries to use an incomplete config which
then fails to establish the connection.
When the kube-apiserver has --anonymous-auth=false,
the regular http.Client.Get() that WaitForAllControlPlaneComponents
does will not work.
Always use the discovery client when checking the health status
of the kube-apiserver.
Do a minor rework of struct fields and unit tests.
Replace nil client in cmd/phases/join/waitcontrolplane.go.
There was one error path that led to a "controller has shut down" log
message. Other errors caused different log entries or are so unlikely (event
handler registration failure!) that they weren't checked at all.
It's clearer to let Run return an error in all cases and then log the
"controller has shut down" error at the call site. This also enables tests to
mark themselves as failed, should that ever happen.
This change introduces improvements to the component compatibility registry:
- Modify the kube-scheduler test server to create a separate ComponentGlobalsRegistry
- Update the compatibility registry to handle multiple flag configurations
- Enhance test cases to support emulation version mapping between components
Various parts of kube-proxy passed around a "hostname", but it is
actually the name of the *node* kube-proxy is running on, which is not
100% guaranteed to be exactly the same as the hostname. Rename it
everywhere to make it clearer that (a) it is definitely safe to use
that name to refer to the Node, (b) it is not necessarily safe to use
that name with DNS, etc.
The controller is derived from the node taint eviction controller.
In contrast to that controller it tracks the UID of pods to prevent
deleting the wrong pod when it got replaced.
The tests and comments have also been updated because while
VolumeCapacityPriority preferred a node with the least allocatable,
StorageCapacityScoring preferred a node with the maximum allocatable.
Basically all callers want dual-stack-if-possible, so simplify that.
Also, tweak the startup-time checking in kubelet to treat "no iptables
support" as interesting but not an error.
It was there so you could mock the results via a FakeExec, but these
days any unit tests outside of pkg/util/iptables that want to mock
iptables results use a FakeIPTables instead of a real
utiliptables.Interface with a FakeExec.
Historically it took an exec argument so you could pass a FakeExec to
mock its behavior in unit tests, but it has a fake implementation now
that is much more useful for unit tests than trying to use the real
implementation with a fake exec. (The unit tests still use fake execs,
but they don't need to use a public constructor.) So remove the exec
args from the public constructors.
