Testing improvements

Fix bugs in DeltaFIFO

federation-apiserver test: Increasing timeout for clusters to be ready

federation-apiserver test: Cleaning up clusters after the test is run

One more place we need a secret.

Automatic merge from submit-queue

Fixes and improvements to Photon Controller backend for kube-up

- Improve reliability of network address detection by using MAC
  address. VMware has a MAC OUI that reliably distinguishes the VM's
  NICs from the other NICs (like the CBR). This doesn't rely on the
  unreliable reporting of the portgroup.
- Persist route changes. We configure routes on the master and nodes,
  but previously we didn't persist them so they didn't last across
  reboots. This persists them in /etc/network/interfaces
- Fix regression that didn't configure auth for kube-apiserver with
  Photon Controller.
- Reliably run apt-get update: Not doing this can cause apt to fail.
- Remove unused nginx config in salt

Update the latestReleaseBranch to release-1.3 in the munger.

Automatic merge from submit-queue

Fixed e2e gke test "scale up correct target pool".

[![Analytics](https://kubernetes-site.appspot.com/UA-36037335-10/GitHub/.github/PULL_REQUEST_TEMPLATE.md?pixel)]()

Fixed e2e gke test "scale up correct target pool" by adding creation of
cluster autoscaler for the new pool.

Automatic merge from submit-queue

Change default value of deleting-pods-burst to 1

Fix. #27413

Automatic merge from submit-queue

Kubelet Volume Attach/Detach/Mount/Unmount Redesign

This PR redesigns the Volume Attach/Detach/Mount/Unmount in Kubelet as proposed in https://github.com/kubernetes/kubernetes/issues/21931

```release-note
A new volume manager was introduced in kubelet that synchronizes volume mount/unmount (and attach/detach, if attach/detach controller is not enabled).

This eliminates the race conditions between the pod creation loop and the orphaned volumes loops. It also removes the unmount/detach from the `syncPod()` path so volume clean up never blocks the `syncPod` loop.
```

Automatic merge from submit-queue

Update default value of CLAIM_BINDER_SYNC_PERIOD.

It's 15 seconds now, see #26414 and #24236.

Automatic merge from submit-queue

federation: choosing a default federation name in test instead of failing

The tests are failing right now:
http://kubekins.dls.corp.google.com/job/kubernetes-e2e-gce-federation/

```
[k8s.io] Service [Feature:Federation] should be able to discover a non-local federated service

/go/src/k8s.io/kubernetes/_output/dockerized/go/src/k8s.io/kubernetes/test/e2e/federated-service.go:130 Jun 14 12:40:35.091: FEDERATION_NAME environment variable must be set


[k8s.io] Service [Feature:Federation] should be able to discover a federated service

/go/src/k8s.io/kubernetes/_output/dockerized/go/src/k8s.io/kubernetes/test/e2e/federated-service.go:130 Jun 14 12:40:40.802: FEDERATION_NAME environment variable must be set
```

This is to fix them.

cc @kubernetes/sig-cluster-federation @mml 

Fixes https://github.com/kubernetes/kubernetes/pull/27410

This commit adds a new volume manager in kubelet that synchronizes
volume mount/unmount (and attach/detach, if attach/detach controller
is not enabled).

This eliminates the race conditions between the pod creation loop
and the orphaned volumes loops. It also removes the unmount/detach
from the `syncPod()` path so volume clean up never blocks the
`syncPod` loop.

Rename UniqueDeviceName to UniqueVolumeName and move helper functions
from attacherdetacher to volumehelper package.
Introduce UniquePodName alias

Automatic merge from submit-queue

Fix failures in up and down services in large clusters

Ref #26687

Automatic merge from submit-queue

Make timeout for starting system pods configurable

Context: in 2000-node clusters (if only one node is big enough to fit heapster, which is our testing configuration), heapster won't be scheduled until that node has route. However, creating routes is pretty expensive and currently can take even 2 hours.

@zmerlynn @gmarek 

Fixed e2e gke test "scale up correct target pool" by adding creation of
cluster autoscaler for the new pool.

Revert "Downward api node defaults"

It's 15 seconds now, see #26414.

Changed constans to reduce hpa e2e test flakiness.

Turn off mitlinode pool test in GKE until it is fixed

Changed constans to reduce hpa e2e test flakiness.

Automatic merge from submit-queue

Add Google Cloud DNS auth scope to GCE VMs when they are started in a federation.

cc @colhom @kubernetes/sig-cluster-federation 

[![Analytics](https://kubernetes-site.appspot.com/UA-36037335-10/GitHub/.github/PULL_REQUEST_TEMPLATE.md?pixel)]()

Automatic merge from submit-queue

Downward api node defaults

Fixes #27062

Replaces https://github.com/kubernetes/kubernetes/pull/27107

Automatic merge from submit-queue

Update PVC.Selector docs: it's ignored when VolumeName is set

@kubernetes/sig-storage 

Automatic merge from submit-queue

GKE provider: Add a HEAPSTER_MACHINE_TYPE option

[![Analytics](https://kubernetes-site.appspot.com/UA-36037335-10/GitHub/.github/PULL_REQUEST_TEMPLATE.md?pixel)]()

Allows heterogeneous cluster configurations for large cluster
synthetic testing.

Supersedes #27389 

Automatic merge from submit-queue

Fix a debugging line

A trivial update. @k8s-oncall can we manually merge it?

- Improve reliability of network address detection by using MAC
  address. VMware has a MAC OUI that reliably distinguishes the VM's
  NICs from the other NICs (like the CBR). This doesn't rely on the
  unreliable reporting of the portgroup.
- Persist route changes. We configure routes on the master and nodes,
  but previously we didn't persist them so they didn't last across
  reboots. This persists them in /etc/network/interfaces
- Fix regression that didn't configure auth for kube-apiserver with
  Photon Controller.
- Reliably run apt-get update: Not doing this can cause apt to fail.
- Remove unused nginx config in salt