Using the "normal" logic for a feature gated field simplifies the
implementation of the feature gate.
There is one (entirely theoretic!) problem with updating from 1.31: if a claim
was allocated in 1.31 with admin access, the status field was not set because
it didn't exist yet. If a driver now follows the current definition of "unset =
off", then it will not grant admin access even though it should. This is
theoretic because drivers are starting to support admin access with 1.32, so
there shouldn't be any claim where this problem could occur.
Kubernetes-commit: 4419568259590c35f1dab69aabec3d740944a51d
Drivers need to know that because admin access may also grant additional
permissions. The allocator needs to ignore such results when determining which
devices are considered as allocated.
In both cases it is conceptually cleaner to not rely on the content of the
ClaimSpec.
Kubernetes-commit: f3fef01e79a75ebc4c327afb7d05d6fd350e08fa
As with the apiserver feature gate for CBOR as a serving and storage encoding, the client feature
gates for CBOR are being initially added through a test-only feature gate instance that is not wired
to environment variables or to command-line flags and is intended only to be enabled
programmatically from integration tests. The test-only instance will be removed as part of alpha
graduation and replaced by conventional client feature gating.
Kubernetes-commit: ea13190d8bd3a4bb3e82055b529aa7599ae5c6e1
This removes the DRAControlPlaneController feature gate, the fields controlled
by it (claim.spec.controller, claim.status.deallocationRequested,
claim.status.allocation.controller, class.spec.suitableNodes), the
PodSchedulingContext type, and all code related to the feature.
The feature gets removed because there is no path towards beta and GA and DRA
with "structured parameters" should be able to replace it.
Kubernetes-commit: f84eb5ecf894fa0fc4e0d05da52ef51d4cd723d9
With this change, we can typically avoid an extra heap allocation when
calling AddAfter with a positive duration (which causes the creation of
a waitFor object). This is because workqueues are typically used with
string keys, and casting a string (or more generally, non-pointer types)
to an `interface{}` will cause an heap escape / allocation.
Ater this change, there is no longer any usage of `type t interface{}`,
which was creating some confusion after the switch to generics in the
workqueue package.
Co-authored-by: Quan Tian <quan.tian@broadcom.com>
Signed-off-by: Antonin Bas <antonin.bas@broadcom.com>
Kubernetes-commit: 53ddffb55709857fec6bddbd5ca4ca1b03f7da97
I frequently find myself in the situation of not remembering which of
QPS/Burst I have to set. This change adds a small go doc to clarify
that.
Kubernetes-commit: cd1645ff2c195fbb353cfabcbc36e3c3b883c3c5
The workqueue implementation was recently updated to be strongly typed,
using Go generics. However the metrics implementation was not updated,
and continued using interface{}. This translated to unnecessary memory
allocations when invoking the queueMetrics interface methods to track
queue operation. We can avoid these extra heap allocations by using
generics for the metrics implementation as well.
Signed-off-by: Antonin Bas <antonin.bas@broadcom.com>
Kubernetes-commit: 1aec7568e111f5855121e3afacacf431e5f95948
The azure and gcp plugins no longer do anything other than point to
the corresponding external credential plugins. Client code should no
longer try to load them, so they should be removed from the examples.
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Kubernetes-commit: 4531c6de760ec42941d9f509103f1b29c88bfae0
