Merge pull request #55188 from mindprince/accelerator-monitoring

Automatic merge from submit-queue (batch tested with PRs 55798, 49579, 54862, 55188, 51990). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.

Add monitoring support for hardware accelerators

Currently only NVIDIA GPU monitoring is implemented.

Feature repo issue: https://github.com/kubernetes/features/issues/369
cAdvisor PR: https://github.com/google/cadvisor/pull/1762

/kind feature
/sig node
/sig instrumentation
/area hw-accelerators

**Release note**:
```release-note
Kubelet now exposes metrics for NVIDIA GPUs attached to the containers.
```

Kubernetes-commit: 779105673ab2f00af9e02dc9ac3c6413588aeae4

vendor/golang.org/x/sys/unix/syscall_darwin.go

@@ -54,7 +54,7 @@ func nametomib(name string) (mib []_C_int, err error) {
 
 	// NOTE(rsc): It seems strange to set the buffer to have
 	// size CTL_MAXNAME+2 but use only CTL_MAXNAME
-	// as the size.  I don't know why the +2 is here, but the
+	// as the size. I don't know why the +2 is here, but the
 	// kernel uses +2 for its own implementation of this function.
 	// I am scared that if we don't include the +2 here, the kernel
 	// will silently write 2 words farther than we specify
@@ -377,7 +377,6 @@ func IoctlGetTermios(fd int, req uint) (*Termios, error) {
 // Searchfs
 // Delete
 // Copyfile
-// Poll
 // Watchevent
 // Waitevent
 // Modwatch