111 lines
2.3 KiB
Go
111 lines
2.3 KiB
Go
|
package prometheus
|
||
|
|
||
|
import (
|
||
|
"sync"
|
||
|
"syscall"
|
||
|
|
||
|
"github.com/Sirupsen/logrus"
|
||
|
"github.com/crosbymichael/cgroups"
|
||
|
metrics "github.com/docker/go-metrics"
|
||
|
)
|
||
|
|
||
|
func NewOOMCollector(ns *metrics.Namespace) (*OOMCollector, error) {
|
||
|
fd, err := syscall.EpollCreate1(syscall.EPOLL_CLOEXEC)
|
||
|
if err != nil {
|
||
|
return nil, err
|
||
|
}
|
||
|
c := &OOMCollector{
|
||
|
fd: fd,
|
||
|
memoryOOM: ns.NewLabeledGauge("memory_oom", "The number of times a container received an oom event", metrics.Total, "id"),
|
||
|
set: make(map[uintptr]*oom),
|
||
|
}
|
||
|
go c.start()
|
||
|
return c, nil
|
||
|
}
|
||
|
|
||
|
type OOMCollector struct {
|
||
|
mu sync.Mutex
|
||
|
|
||
|
memoryOOM metrics.LabeledGauge
|
||
|
fd int
|
||
|
set map[uintptr]*oom
|
||
|
}
|
||
|
|
||
|
type oom struct {
|
||
|
id string
|
||
|
c cgroups.Cgroup
|
||
|
}
|
||
|
|
||
|
func (o *OOMCollector) Add(id string, cg cgroups.Cgroup) error {
|
||
|
o.mu.Lock()
|
||
|
defer o.mu.Unlock()
|
||
|
fd, err := cg.OOMEventFD()
|
||
|
if err != nil {
|
||
|
return err
|
||
|
}
|
||
|
o.set[fd] = &oom{
|
||
|
id: id,
|
||
|
c: cg,
|
||
|
}
|
||
|
// set the gauge's default value
|
||
|
o.memoryOOM.WithValues(id).Set(0)
|
||
|
event := syscall.EpollEvent{
|
||
|
Fd: int32(fd),
|
||
|
Events: syscall.EPOLLHUP | syscall.EPOLLIN | syscall.EPOLLERR,
|
||
|
}
|
||
|
if err := syscall.EpollCtl(o.fd, syscall.EPOLL_CTL_ADD, int(fd), &event); err != nil {
|
||
|
return err
|
||
|
}
|
||
|
return nil
|
||
|
}
|
||
|
|
||
|
// Close closes the epoll fd
|
||
|
func (o *OOMCollector) Close() error {
|
||
|
return syscall.Close(int(o.fd))
|
||
|
}
|
||
|
|
||
|
func (o *OOMCollector) start() {
|
||
|
var events [128]syscall.EpollEvent
|
||
|
for {
|
||
|
n, err := syscall.EpollWait(o.fd, events[:], -1)
|
||
|
if err != nil {
|
||
|
if err == syscall.EINTR {
|
||
|
continue
|
||
|
}
|
||
|
logrus.WithField("error", err).Fatal("cgroups: epoll wait")
|
||
|
}
|
||
|
for i := 0; i < n; i++ {
|
||
|
o.process(uintptr(events[i].Fd), events[i].Events)
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
func (o *OOMCollector) process(fd uintptr, event uint32) {
|
||
|
// make sure to always flush the fd
|
||
|
flush(fd)
|
||
|
|
||
|
o.mu.Lock()
|
||
|
info, ok := o.set[fd]
|
||
|
if !ok {
|
||
|
o.mu.Unlock()
|
||
|
return
|
||
|
}
|
||
|
o.mu.Unlock()
|
||
|
// if we received an event but it was caused by the cgroup being deleted and the fd
|
||
|
// being closed make sure we close our copy and remove the container from the set
|
||
|
if info.c.State() == cgroups.Deleted {
|
||
|
o.mu.Lock()
|
||
|
delete(o.set, fd)
|
||
|
o.mu.Unlock()
|
||
|
syscall.Close(int(fd))
|
||
|
return
|
||
|
}
|
||
|
o.memoryOOM.WithValues(info.id).Inc(1)
|
||
|
}
|
||
|
|
||
|
func flush(fd uintptr) error {
|
||
|
buf := make([]byte, 8)
|
||
|
_, err := syscall.Read(int(fd), buf)
|
||
|
return err
|
||
|
}
|