var uploadedParts []minio.CompletePart
	// new multipart must not set mtime as it may lead to erroneous cleanups at various intervals.
	opts.Internal.SourceMTime = time.Time{} // this value is saved properly in CompleteMultipartUpload()
	var uploadID string
	attempts := 1
	for attempts <= 3 {

* Apiserver proxy rewrites URL when service returns absolute path with request's host. ([#52556](https://github.com/kubernetes/kubernetes/pull/52556), [@roycaihw](https://github.com/roycaihw))
* Logging cleanups ([#54443](https://github.com/kubernetes/kubernetes/pull/54443), [@bowei](https://github.com/bowei))
        * Updates kube-dns to use client-go 3
        * Updates containers to use alpine as the base image on all platforms

This is the driest thing I know.  Silence all round, if you please!
"William the Conqueror, whose cause was favoured by the pope, was
soon submitted to by the English, who wanted leaders, and had been
of late much accustomed to usurpation and conquest.  Edwin and
Morcar, the earls of Mercia and Northumbria--"'

  `Ugh!' said the Lory, with a shiver.

  - build: Update to k/repo-infra@v0.1.1 (supports go1.15.2)
  - build: Use go-runner:buster-v2.0.1 (built using go1.15.1)
  - bazel: Replace --features with Starlark build settings flag
  - hack/lib/util.sh: some bash cleanups
    
    - switched one spot to use kube::logging
    - make kube::util::find-binary return an error when it doesn't find
      anything so that hack scripts fail fast instead of with '' binary not

# Release 2.18.0

## TensorFlow

### Breaking Changes

* `tf.lite`
    * C API:

* Bump cluster autoscaler to v0.5.4, which fixes scale down issues with pods ignoring SIGTERM. ([#45483](https://github.com/kubernetes/kubernetes/pull/45483), [@mwielgus](https://github.com/mwielgus))
* Fixes and minor cleanups to pod (anti)affinity predicate ([#45098](https://github.com/kubernetes/kubernetes/pull/45098), [@wojtek-t](https://github.com/wojtek-t))

* Fix a bug in server printer that could cause kube-apiserver to panic. ([#79349](https://github.com/kubernetes/kubernetes/pull/79349), [@roycaihw](https://github.com/roycaihw))
* Fix a bug in the IPVS proxier where virtual servers are not cleaned up even though the corresponding Service object was deleted. ([#80942](https://github.com/kubernetes/kubernetes/pull/80942), [@gongguan](https://github.com/gongguan))

- Fixes a regression introduced in 1.29 where conntrack entries for UDP connections
  to deleted pods did not get cleaned up correctly, which could (among other things)
  cause DNS problems when DNS pods were restarted. ([#127807](https://github.com/kubernetes/kubernetes/pull/127807), [@danwinship](https://github.com/danwinship)) [SIG Network]

* Allow disable outbound SNAT when Azure standard load balancer is used together with outbound rules.  ([#75282](https://github.com/kubernetes/kubernetes/pull/75282), [@feiskyer](https://github.com/feiskyer))
* Ensure Azure load balancer cleaned up on 404 or 403 when deleting LoadBalancer services. ([#75256](https://github.com/kubernetes/kubernetes/pull/75256), [@feiskyer](https://github.com/feiskyer))

* etcd3: make gets for previous value in watch serialize-able ([#34089](https://github.com/kubernetes/kubernetes/pull/34089), [@wojtek-t](https://github.com/wojtek-t))
* etcd3: minor cleanups ([#34234](https://github.com/kubernetes/kubernetes/pull/34234), [@wojtek-t](https://github.com/wojtek-t))