- [Additional changes](#additional-changes)
  - [External Dependencies](#external-dependencies)
  - [Bug Fixes](#bug-fixes)
      - [General Fixes and Reliability](#general-fixes-and-reliability)
  - [Non-user-facing changes](#non-user-facing-changes)
- [v1.11.0-rc.3](#v1110-rc3)
  - [Downloads for v1.11.0-rc.3](#downloads-for-v1110-rc3)
    - [Client Binaries](#client-binaries-11)
    - [Server Binaries](#server-binaries-11)

* Fix resync goroutine leak in ListAndWatch ([#35672](https://github.com/kubernetes/kubernetes/pull/35672), [@tatsuhiro-t](https://github.com/tatsuhiro-t))
* Kubelet will no longer set hairpin mode on every interface on the machine when an error occurs in setting up hairpin for a specific interface. ([#36990](https://github.com/kubernetes/kubernetes/pull/36990), [@bboreham](https://github.com/bboreham))

# LINUX ARM64 PYCPP
# In Linux Arm64 presubmit/continuous build, we cross-compile the binaries on
# Linux x86 so that we can use RBE. Since tests still need to run on the single
# host Arm64 machine, the build becomes too slow (~30 min) to be a presubmit.
# For testing purposes, we want to see the runtime performance of an
# experimental job that is build-only, i.e, we only build the test targets and

		VersionID:         versionID,
		MTime:             dobj.DeleteMarkerMTime.Time,
		DeleteReplication: drs,
		Versioned:         globalBucketVersioningSys.PrefixEnabled(bucket, dobj.ObjectName),
		// Objects matching prefixes should not leave delete markers,
		// dramatically reduces namespace pollution while keeping the
		// benefits of replication, make sure to apply version suspension
		// only at bucket level instead.

# CL 132000043 debug/elf: support arm64 relocations, Michael Hudson-Doyle <******@****.***>
pkg debug/elf, const EM_AARCH64 = 183
pkg debug/elf, const EM_AARCH64 Machine
pkg debug/elf, const R_AARCH64_ABS16 = 259
pkg debug/elf, const R_AARCH64_ABS16 R_AARCH64
pkg debug/elf, const R_AARCH64_ABS32 = 258
pkg debug/elf, const R_AARCH64_ABS32 R_AARCH64
pkg debug/elf, const R_AARCH64_ABS64 = 257

	if poolIdx < len(z.serverPools) && setIdx < len(z.serverPools[poolIdx].sets) {
		return z.serverPools[poolIdx].sets[setIdx].getDisks(), nil
	}
	return nil, fmt.Errorf("Matching pool %s, set %s not found", humanize.Ordinal(poolIdx+1), humanize.Ordinal(setIdx+1))
}

// Return the count of disks in each pool
func (z *erasureServerPools) SetDriveCounts() []int {

	if !p.more() {
		return EOF
	}
	tok := p.input[p.inputPos]
	p.inputPos++
	return tok
}

func (p *Parser) back() {
	if p.inputPos == 0 {
		p.errorf("internal error: backing up before BOL")
	} else {
		p.inputPos--
	}
}

func (p *Parser) peek() lex.ScanToken {
	if p.more() {
		return p.input[p.inputPos].ScanToken
	}
	return scanner.EOF
}

- The calculations for Pod topology spread skew now exclude nodes that
  don't match the node affinity/selector. This may lead to unschedulable pods if you previously had pods
  matching the spreading selector on those excluded nodes (not matching the node affinity/selector),
  especially when the `topologyKey` is not node-level. Revisit the node affinity and/or pod selector in the

			// far enough to be better than Rabin-Karp.
			// Experiments (using IndexPeriodic) suggest
			// the cutover is about 16 byte skips.
			// TODO: if large prefixes of sep are matching
			// we should cutover at even larger average skips,
			// because Equal becomes that much more expensive.
			// This code does not take that effect into account.
			j := bytealg.IndexRabinKarp(s[i:], sep)

- Added CEL runtime cost calculation into ValidatingAdmissionPolicy, matching the evaluation cost
  restrictions that already apply to CustomResourceDefinition.
  If rule evaluation uses more compute than the limit, the API server aborts the evaluation and the