No need to (re-)build documentation when a release is published.
Great simplification of the Helm repo index update script: do not scan
all releases but just get the assets from the release that was
published.
This separation should make the maintenance of scripts and workflows
easier.
(cherry picked from commit 72bf84c4fa)
Add context to functions that now require it. Also, replace the
deprecated wait.Poll* calls with wait.PollUntilContextTimeout.
(cherry picked from commit 87371e2df0)
Eliminate all context.TODO() from the e2e tests and use ginkgo context
instead. This ensures that calls involving context are properly
cancelled and return fast in case the tests get aborted.
(cherry picked from commit ad8bd057b7)
Update mocked implementation of
k8s.io/kubelet/pkg/apis/podresources/v1.PodResourcesListerClient. The
mocked implementation is moved to a separate "mocks" subpackage as it's
for an external interface.
This patch also adds code for auto-generation for the mocked interface.
(cherry picked from commit e2d5ba1a2b)
Let's refactor part of the getCgroupMiscCapacity() out to its own
retrieveCgroupMiscCapacityValue(), for the legibility sake.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
We've been only considering cgroupsv2 when trying to read misc.capacity.
However, there are still a bunch of systems out there relying on
cgroupsv1.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
Don't try to update all nodes on config change when the NodeFeature API
is disabled. In this case we rely on gRPC and only act on gRPC requests
from the worker.
Validate labels before trying to update the node. Makes us fail early
nad prevent useless retries in case invalid labels are tried.
(backported from commit 2a3c7e4c93)
Make it possible to disable kubelet state tracking with
--set topologyUpdater.kubeletStateFiles="" as the documentation
suggests.
Also, fix the documentation regarding the default value of
topologyUpdater.kubeletStateFiles parameter.
Change the configuration so that, by default, we use a dedicated
serviceaccount for topology-updater (similar to topology-gc, nfd-master
and nfd-worker).
Fix the templates so that the serviceaccount and clusterrolebinding are
only created when topology-updater is enabled (clusterrole was already
handled this way).
This patch also correctly documents the default value of rbac.create
parameter of topology-updater and topology-gc.
Mount kubelet podresources socket on an independent path, not under
with the kubelet state directory. Otherwise container creation may fail
on mount creation if topologyUpdater.kubeletPodResourcesSockPath and/or
topologyUpdater.kubeletConfigPath Helm parameters are specified in a
certain way.
Volume/mount setup for the ConfigMap was erroneously inside conditionals
so it was not mounted unless TLS was enabled.
(cherry picked from commit b016def8a3)
