- Reduce memory usage/allocations during wait for volume attachment ([#126575](https://github.com/kubernetes/kubernetes/pull/126575), [@Lucaber](https://github.com/Lucaber)) [SIG Node and Storage]

- Reduce critical section in watchcache to fix kube-apiserver scalability under heavy load of list requests ([#122027](https://github.com/kubernetes/kubernetes/pull/122027), [@wojtek-t](https://github.com/wojtek-t)) [SIG API Machinery]

import okhttp3.internal.connection.RealConnectionPool
import okhttp3.internal.connection.RealRoutePlanner
import okhttp3.internal.connection.RouteDatabase

/**
 * Manages reuse of HTTP and HTTP/2 connections for reduced network latency. HTTP requests that
 * share the same [Address] may share a [Connection]. This class implements the policy
 * of which connections to keep open for future use.
 *

   ([#121958](https://github.com/kubernetes/kubernetes/pull/121958), [@neolit123](https://github.com/neolit123))
- Added an optimization to reduce stack memory usage for watch requests. It can be disabled with the feature gate: `APIServingWithRoutine=false` ([#120902](https://github.com/kubernetes/kubernetes/pull/120902), [@linxiulei](https://github.com/linxiulei))

* allow windows mount path ([#51240](https://github.com/kubernetes/kubernetes/pull/51240), [@andyzhangx](https://github.com/andyzhangx))
* Reduce log noise produced by prometheus-to-sd, by bumping it to version 0.2.2. ([#54635](https://github.com/kubernetes/kubernetes/pull/54635), [@loburm](https://github.com/loburm))

These are the primary classes involved with establishing, sharing, and terminating connections:

 * **RealConnectionPool** manages reuse of HTTP and HTTP/2 connections for reduced latency. Every OkHttpClient has one, and its lifetime spans the lifetime of the OkHttpClient.

- Services implemented a field selector for the ClusterIP and Type fields.
  The Kubelet uses this field selector to avoid monitoring Headless Services, which helps reduce memory consumption. ([#123905](https://github.com/kubernetes/kubernetes/pull/123905), [@aojea](https://github.com/aojea)) [SIG Apps, Node and Testing]

    if (tableSize <= BYTE_MAX_SIZE) {
      /*
       * Use 8 bits per entry. The value is unsigned to allow use up to a size of 2^8.
       *
       * The absent indicator of -1 signed becomes 2^8 - 1 unsigned, which reduces the actual max
       * size to 2^8 - 1. However, due to a load factor < 1 the limit is never approached.
       */
      byte[] hashTable = new byte[tableSize];
      Arrays.fill(hashTable, ABSENT);

  * Pods can now consume Secret data in environment variables and inject those
environment variables into a container’s command-line args.
  * Stable version of Heapster which scales up to 1000 nodes: more metrics, reduced
latency, reduced cpu/memory consumption (~4mb per monitored node).
  * Pods now have a security context which allows users to specify:
     * attributes which apply to the whole pod:
        * User ID

 * which is optimal, and <i>not</i> the size of the internal hashtable, which could be much larger
 * than {@code size()}. Furthermore, this structure places significantly reduced load on the garbage
 * collector by only using a constant number of internal objects.
 *
 * <p>If there are no removals, then iteration order for the {@link #entrySet}, {@link #keySet}, and