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
This commit is contained in:
Kubernetes Publisher
2017-11-16 03:09:21 -08:00
173 changed files with 3350 additions and 1180 deletions
+10 -9
View File
@@ -255,7 +255,7 @@ func Getgroups() (gids []int, err error) {
return nil, nil
}
// Sanity check group count. Max is 1<<16 on Linux.
// Sanity check group count. Max is 1<<16 on Linux.
if n < 0 || n > 1<<20 {
return nil, EINVAL
}
@@ -290,8 +290,8 @@ type WaitStatus uint32
// 0x7F (stopped), or a signal number that caused an exit.
// The 0x80 bit is whether there was a core dump.
// An extra number (exit code, signal causing a stop)
// is in the high bits. At least that's the idea.
// There are various irregularities. For example, the
// is in the high bits. At least that's the idea.
// There are various irregularities. For example, the
// "continued" status is 0xFFFF, distinguishing itself
// from stopped via the core dump bit.
@@ -926,7 +926,7 @@ func Recvmsg(fd int, p, oob []byte, flags int) (n, oobn int, recvflags int, from
msg.Namelen = uint32(SizeofSockaddrAny)
var iov Iovec
if len(p) > 0 {
iov.Base = (*byte)(unsafe.Pointer(&p[0]))
iov.Base = &p[0]
iov.SetLen(len(p))
}
var dummy byte
@@ -941,7 +941,7 @@ func Recvmsg(fd int, p, oob []byte, flags int) (n, oobn int, recvflags int, from
iov.Base = &dummy
iov.SetLen(1)
}
msg.Control = (*byte)(unsafe.Pointer(&oob[0]))
msg.Control = &oob[0]
msg.SetControllen(len(oob))
}
msg.Iov = &iov
@@ -974,11 +974,11 @@ func SendmsgN(fd int, p, oob []byte, to Sockaddr, flags int) (n int, err error)
}
}
var msg Msghdr
msg.Name = (*byte)(unsafe.Pointer(ptr))
msg.Name = (*byte)(ptr)
msg.Namelen = uint32(salen)
var iov Iovec
if len(p) > 0 {
iov.Base = (*byte)(unsafe.Pointer(&p[0]))
iov.Base = &p[0]
iov.SetLen(len(p))
}
var dummy byte
@@ -993,7 +993,7 @@ func SendmsgN(fd int, p, oob []byte, to Sockaddr, flags int) (n int, err error)
iov.Base = &dummy
iov.SetLen(1)
}
msg.Control = (*byte)(unsafe.Pointer(&oob[0]))
msg.Control = &oob[0]
msg.SetControllen(len(oob))
}
msg.Iov = &iov
@@ -1023,7 +1023,7 @@ func ptracePeek(req int, pid int, addr uintptr, out []byte) (count int, err erro
var buf [sizeofPtr]byte
// Leading edge. PEEKTEXT/PEEKDATA don't require aligned
// Leading edge. PEEKTEXT/PEEKDATA don't require aligned
// access (PEEKUSER warns that it might), but if we don't
// align our reads, we might straddle an unmapped page
// boundary and not get the bytes leading up to the page
@@ -1262,6 +1262,7 @@ func Getpgrp() (pid int) {
//sys PivotRoot(newroot string, putold string) (err error) = SYS_PIVOT_ROOT
//sysnb prlimit(pid int, resource int, newlimit *Rlimit, old *Rlimit) (err error) = SYS_PRLIMIT64
//sys Prctl(option int, arg2 uintptr, arg3 uintptr, arg4 uintptr, arg5 uintptr) (err error)
//sys Pselect(nfd int, r *FdSet, w *FdSet, e *FdSet, timeout *Timespec, sigmask *Sigset_t) (n int, err error) = SYS_PSELECT6
//sys read(fd int, p []byte) (n int, err error)
//sys Removexattr(path string, attr string) (err error)
//sys Renameat(olddirfd int, oldpath string, newdirfd int, newpath string) (err error)