Persist container state to disk

Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
This commit is contained in:
Michael Crosby 2015-02-11 16:45:23 -08:00
parent 31327166e5
commit 5c246d038f
17 changed files with 211 additions and 197 deletions

View File

@ -9,46 +9,46 @@ const (
)
type Cgroup struct {
Name string `json:"name,omitempty"`
Name string `json:"name"`
// name of parent cgroup or slice
Parent string `json:"parent,omitempty"`
Parent string `json:"parent"`
// If this is true allow access to any kind of device within the container. If false, allow access only to devices explicitly listed in the allowed_devices list.
AllowAllDevices bool `json:"allow_all_devices,omitempty"`
AllowAllDevices bool `json:"allow_all_devices"`
AllowedDevices []*Device `json:"allowed_devices,omitempty"`
AllowedDevices []*Device `json:"allowed_devices"`
// Memory limit (in bytes)
Memory int64 `json:"memory,omitempty"`
Memory int64 `json:"memory"`
// Memory reservation or soft_limit (in bytes)
MemoryReservation int64 `json:"memory_reservation,omitempty"`
MemoryReservation int64 `json:"memory_reservation"`
// Total memory usage (memory + swap); set `-1' to disable swap
MemorySwap int64 `json:"memory_swap,omitempty"`
MemorySwap int64 `json:"memory_swap"`
// CPU shares (relative weight vs. other containers)
CpuShares int64 `json:"cpu_shares,omitempty"`
CpuShares int64 `json:"cpu_shares"`
// CPU hardcap limit (in usecs). Allowed cpu time in a given period.
CpuQuota int64 `json:"cpu_quota,omitempty"`
CpuQuota int64 `json:"cpu_quota"`
// CPU period to be used for hardcapping (in usecs). 0 to use system default.
CpuPeriod int64 `json:"cpu_period,omitempty"`
CpuPeriod int64 `json:"cpu_period"`
// CPU to use
CpusetCpus string `json:"cpuset_cpus,omitempty"`
CpusetCpus string `json:"cpuset_cpus"`
// MEM to use
CpusetMems string `json:"cpuset_mems,omitempty"`
CpusetMems string `json:"cpuset_mems"`
// Specifies per cgroup weight, range is from 10 to 1000.
BlkioWeight int64 `json:"blkio_weight,omitempty"`
BlkioWeight int64 `json:"blkio_weight"`
// set the freeze value for the process
Freezer FreezerState `json:"freezer,omitempty"`
Freezer FreezerState `json:"freezer"`
// Parent slice to use for systemd TODO: remove in favor or parent
Slice string `json:"slice,omitempty"`
Slice string `json:"slice"`
}

View File

@ -3,98 +3,98 @@ package configs
import "fmt"
type Rlimit struct {
Type int `json:"type,omitempty"`
Hard uint64 `json:"hard,omitempty"`
Soft uint64 `json:"soft,omitempty"`
Type int `json:"type"`
Hard uint64 `json:"hard"`
Soft uint64 `json:"soft"`
}
// IDMap represents UID/GID Mappings for User Namespaces.
type IDMap struct {
ContainerID int `json:"container_id,omitempty"`
HostID int `json:"host_id,omitempty"`
Size int `json:"size,omitempty"`
ContainerID int `json:"container_id"`
HostID int `json:"host_id"`
Size int `json:"size"`
}
// Config defines configuration options for executing a process inside a contained environment.
type Config struct {
// NoPivotRoot will use MS_MOVE and a chroot to jail the process into the container's rootfs
// This is a common option when the container is running in ramdisk
NoPivotRoot bool `json:"no_pivot_root,omitempty"`
NoPivotRoot bool `json:"no_pivot_root"`
// ParentDeathSignal specifies the signal that is sent to the container's process in the case
// that the parent process dies.
ParentDeathSignal int `json:"parent_death_signal,omitempty"`
ParentDeathSignal int `json:"parent_death_signal"`
// PivotDir allows a custom directory inside the container's root filesystem to be used as pivot, when NoPivotRoot is not set.
// When a custom PivotDir not set, a temporary dir inside the root filesystem will be used. The pivot dir needs to be writeable.
// This is required when using read only root filesystems. In these cases, a read/writeable path can be (bind) mounted somewhere inside the root filesystem to act as pivot.
PivotDir string `json:"pivot_dir,omitempty"`
PivotDir string `json:"pivot_dir"`
// Path to a directory containing the container's root filesystem.
Rootfs string `json:"rootfs,omitempty"`
Rootfs string `json:"rootfs"`
// Readonlyfs will remount the container's rootfs as readonly where only externally mounted
// bind mounts are writtable.
Readonlyfs bool `json:"readonlyfs,omitempty"`
Readonlyfs bool `json:"readonlyfs"`
// Mounts specify additional source and destination paths that will be mounted inside the container's
// rootfs and mount namespace if specified
Mounts []*Mount `json:"mounts,omitempty"`
Mounts []*Mount `json:"mounts"`
// The device nodes that should be automatically created within the container upon container start. Note, make sure that the node is marked as allowed in the cgroup as well!
Devices []*Device `json:"devices,omitempty"`
Devices []*Device `json:"devices"`
MountLabel string `json:"mount_label,omitempty"`
MountLabel string `json:"mount_label"`
// Hostname optionally sets the container's hostname if provided
Hostname string `json:"hostname,omitempty"`
Hostname string `json:"hostname"`
// Console is the path to the console allocated to the container.
Console string `json:"console,omitempty"`
Console string `json:"console"`
// Namespaces specifies the container's namespaces that it should setup when cloning the init process
// If a namespace is not provided that namespace is shared from the container's parent process
Namespaces Namespaces `json:"namespaces,omitempty"`
Namespaces Namespaces `json:"namespaces"`
// Capabilities specify the capabilities to keep when executing the process inside the container
// All capbilities not specified will be dropped from the processes capability mask
Capabilities []string `json:"capabilities,omitempty"`
Capabilities []string `json:"capabilities"`
// Networks specifies the container's network setup to be created
Networks []*Network `json:"networks,omitempty"`
Networks []*Network `json:"networks"`
// Routes can be specified to create entries in the route table as the container is started
Routes []*Route `json:"routes,omitempty"`
Routes []*Route `json:"routes"`
// Cgroups specifies specific cgroup settings for the various subsystems that the container is
// placed into to limit the resources the container has available
Cgroups *Cgroup `json:"cgroups,omitempty"`
Cgroups *Cgroup `json:"cgroups"`
// AppArmorProfile specifies the profile to apply to the process running in the container and is
// change at the time the process is execed
AppArmorProfile string `json:"apparmor_profile,omitempty"`
AppArmorProfile string `json:"apparmor_profile"`
// ProcessLabel specifies the label to apply to the process running in the container. It is
// commonly used by selinux
ProcessLabel string `json:"process_label,omitempty"`
ProcessLabel string `json:"process_label"`
// RestrictSys will remount /proc/sys, /sys, and mask over sysrq-trigger as well as /proc/irq and
// /proc/bus
RestrictSys bool `json:"restrict_sys,omitempty"`
RestrictSys bool `json:"restrict_sys"`
// Rlimits specifies the resource limits, such as max open files, to set in the container
// If Rlimits are not set, the container will inherit rlimits from the parent process
Rlimits []Rlimit `json:"rlimits,omitempty"`
Rlimits []Rlimit `json:"rlimits"`
// AdditionalGroups specifies the gids that should be added to supplementary groups
// in addition to those that the user belongs to.
AdditionalGroups []int `json:"additional_groups,omitempty"`
AdditionalGroups []int `json:"additional_groups"`
// UidMappings is an array of User ID mappings for User Namespaces
UidMappings []IDMap `json:"uid_mappings,omitempty"`
UidMappings []IDMap `json:"uid_mappings"`
// GidMappings is an array of Group ID mappings for User Namespaces
GidMappings []IDMap `json:"gid_mappings,omitempty"`
GidMappings []IDMap `json:"gid_mappings"`
}
// Gets the root uid for the process on host which could be non-zero

View File

@ -11,28 +11,28 @@ const (
type Device struct {
// Device type, block, char, etc.
Type rune `json:"type,omitempty"`
Type rune `json:"type"`
// Path to the device.
Path string `json:"path,omitempty"`
Path string `json:"path"`
// Major is the device's major number.
Major int64 `json:"major,omitempty"`
Major int64 `json:"major"`
// Minor is the device's minor number.
Minor int64 `json:"minor,omitempty"`
Minor int64 `json:"minor"`
// Cgroup permissions format, rwm.
Permissions string `json:"permissions,omitempty"`
Permissions string `json:"permissions"`
// FileMode permission bits for the device.
FileMode os.FileMode `json:"file_mode,omitempty"`
FileMode os.FileMode `json:"file_mode"`
// Uid of the device.
Uid uint32 `json:"uid,omitempty"`
Uid uint32 `json:"uid"`
// Gid of the device.
Gid uint32 `json:"gid,omitempty"`
Gid uint32 `json:"gid"`
}
func (d *Device) CgroupString() string {

View File

@ -1,11 +1,11 @@
package configs
type Mount struct {
Type string `json:"type,omitempty"`
Source string `json:"source,omitempty"` // Source path, in the host namespace
Destination string `json:"destination,omitempty"` // Destination path, in the container
Writable bool `json:"writable,omitempty"`
Relabel string `json:"relabel,omitempty"` // Relabel source if set, "z" indicates shared, "Z" indicates unshared
Private bool `json:"private,omitempty"`
Slave bool `json:"slave,omitempty"`
Type string `json:"type"`
Source string `json:"source"` // Source path, in the host namespace
Destination string `json:"destination"` // Destination path, in the container
Writable bool `json:"writable"`
Relabel string `json:"relabel"` // Relabel source if set, "z" indicates shared, "Z" indicates unshared
Private bool `json:"private"`
Slave bool `json:"slave"`
}

View File

@ -19,7 +19,7 @@ const (
// alternate path that is able to be joined via setns.
type Namespace struct {
Type NamespaceType `json:"type"`
Path string `json:"path,omitempty"`
Path string `json:"path"`
}
func (n *Namespace) Syscall() int {

View File

@ -6,42 +6,42 @@ package configs
// container to be setup with the host's networking stack
type Network struct {
// Type sets the networks type, commonly veth and loopback
Type string `json:"type,omitempty"`
Type string `json:"type"`
// Name of the network interface
Name string `json:"name,omitempty"`
Name string `json:"name"`
// The bridge to use.
Bridge string `json:"bridge,omitempty"`
Bridge string `json:"bridge"`
// MacAddress contains the MAC address to set on the network interface
MacAddress string `json:"mac_address,omitempty"`
MacAddress string `json:"mac_address"`
// Address contains the IPv4 and mask to set on the network interface
Address string `json:"address,omitempty"`
Address string `json:"address"`
// Gateway sets the gateway address that is used as the default for the interface
Gateway string `json:"gateway,omitempty"`
Gateway string `json:"gateway"`
// IPv6Address contains the IPv6 and mask to set on the network interface
IPv6Address string `json:"ipv6_address,omitempty"`
IPv6Address string `json:"ipv6_address"`
// IPv6Gateway sets the ipv6 gateway address that is used as the default for the interface
IPv6Gateway string `json:"ipv6_gateway,omitempty"`
IPv6Gateway string `json:"ipv6_gateway"`
// Mtu sets the mtu value for the interface and will be mirrored on both the host and
// container's interfaces if a pair is created, specifically in the case of type veth
// Note: This does not apply to loopback interfaces.
Mtu int `json:"mtu,omitempty"`
Mtu int `json:"mtu"`
// TxQueueLen sets the tx_queuelen value for the interface and will be mirrored on both the host and
// container's interfaces if a pair is created, specifically in the case of type veth
// Note: This does not apply to loopback interfaces.
TxQueueLen int `json:"txqueuelen,omitempty"`
TxQueueLen int `json:"txqueuelen"`
// HostInterfaceName is a unique name of a veth pair that resides on in the host interface of the
// container.
HostInterfaceName string `json:"host_interface_name,omitempty"`
HostInterfaceName string `json:"host_interface_name"`
}
// Routes can be specified to create entries in the route table as the container is started
@ -53,14 +53,14 @@ type Network struct {
// destination of 0.0.0.0(or *) when viewed in the route table.
type Route struct {
// Sets the destination and mask, should be a CIDR. Accepts IPv4 and IPv6
Destination string `json:"destination,omitempty"`
Destination string `json:"destination"`
// Sets the source and mask, should be a CIDR. Accepts IPv4 and IPv6
Source string `json:"source,omitempty"`
Source string `json:"source"`
// Sets the gateway. Accepts IPv4 and IPv6
Gateway string `json:"gateway,omitempty"`
Gateway string `json:"gateway"`
// The device to set this route up for, for example: eth0
InterfaceName string `json:"interface_name,omitempty"`
InterfaceName string `json:"interface_name"`
}

View File

@ -1,18 +0,0 @@
package configs
// The status of a container.
type Status int
const (
// The container exists and is running.
Running Status = iota + 1
// The container exists, it is in the process of being paused.
Pausing
// The container exists, but all its processes are paused.
Paused
// The container does not exist.
Destroyed
)

View File

@ -9,21 +9,44 @@ import (
"github.com/docker/libcontainer/configs"
)
// The status of a container.
type Status int
const (
// The container exists and is running.
Running Status = iota + 1
// The container exists, it is in the process of being paused.
Pausing
// The container exists, but all its processes are paused.
Paused
// The container does not exist.
Destroyed
)
// State represents a running container's state
type State struct {
// ID is the container ID.
ID string `json:"id"`
// InitProcessPid is the init process id in the parent namespace.
InitProcessPid int
InitProcessPid int `json:"init_process_pid"`
// InitProcessStartTime is the init process start time.
InitProcessStartTime string
InitProcessStartTime string `json:"init_process_start"`
// Path to all the cgroups setup for a container. Key is cgroup subsystem name
// with the value as the path.
CgroupPaths map[string]string
CgroupPaths map[string]string `json:"cgroup_paths"`
// NamespacePaths are filepaths to the container's namespaces. Key is the namespace name
// with the value as the path.
NamespacePaths map[string]string
NamespacePaths map[string]string `json:"namespace_paths"`
// Config is the container's configuration.
Config configs.Config `json:"config"`
}
// A libcontainer container object.
@ -40,7 +63,7 @@ type Container interface {
// errors:
// ContainerDestroyed - Container no longer exists,
// Systemerror - System error.
Status() (configs.Status, error)
Status() (Status, error)
// State returns the current container's state information.
//

View File

@ -20,6 +20,9 @@ File: {{$frame.File}}{{end}}
`))
func newGenericError(err error, c ErrorCode) Error {
if le, ok := err.(Error); ok {
return le
}
return &GenericError{
Timestamp: time.Now(),
Err: err,
@ -28,6 +31,18 @@ func newGenericError(err error, c ErrorCode) Error {
}
}
func newSystemError(err error) Error {
if le, ok := err.(Error); ok {
return le
}
return &GenericError{
Timestamp: time.Now(),
Err: err,
ECode: SystemError,
Stack: stacktrace.Capture(2),
}
}
type GenericError struct {
Timestamp time.Time
ECode ErrorCode

View File

@ -383,7 +383,7 @@ func TestFreeze(t *testing.T) {
if err := container.Resume(); err != nil {
t.Fatal(err)
}
if state != configs.Paused {
if state != libcontainer.Paused {
t.Fatal("Unexpected state: ", state)
}

View File

@ -3,9 +3,11 @@
package libcontainer
import (
"encoding/json"
"fmt"
"os"
"os/exec"
"path/filepath"
"syscall"
"github.com/docker/libcontainer/cgroups"
@ -32,21 +34,21 @@ func (c *linuxContainer) Config() configs.Config {
return *c.config
}
func (c *linuxContainer) Status() (configs.Status, error) {
func (c *linuxContainer) Status() (Status, error) {
if c.initProcess == nil {
return configs.Destroyed, nil
return Destroyed, nil
}
// return Running if the init process is alive
if err := syscall.Kill(c.initProcess.pid(), 0); err != nil {
if err == syscall.ESRCH {
return configs.Destroyed, nil
return Destroyed, nil
}
return 0, err
return 0, newSystemError(err)
}
if c.config.Cgroups != nil && c.config.Cgroups.Freezer == configs.Frozen {
return configs.Paused, nil
return Paused, nil
}
return configs.Running, nil
return Running, nil
}
func (c *linuxContainer) State() (*State, error) {
@ -54,14 +56,16 @@ func (c *linuxContainer) State() (*State, error) {
if err != nil {
return nil, err
}
if status == configs.Destroyed {
if status == Destroyed {
return nil, newGenericError(fmt.Errorf("container destroyed"), ContainerNotExists)
}
startTime, err := c.initProcess.startTime()
if err != nil {
return nil, err
return nil, newSystemError(err)
}
state := &State{
ID: c.ID(),
Config: *c.config,
InitProcessPid: c.initProcess.pid(),
InitProcessStartTime: startTime,
CgroupPaths: c.cgroupManager.GetPaths(),
@ -96,7 +100,7 @@ func (c *linuxContainer) Processes() ([]int, error) {
glog.Info("fetch container processes")
pids, err := c.cgroupManager.GetPids()
if err != nil {
return nil, newGenericError(err, SystemError)
return nil, newSystemError(err)
}
return pids, nil
}
@ -108,14 +112,14 @@ func (c *linuxContainer) Stats() (*Stats, error) {
stats = &Stats{}
)
if stats.CgroupStats, err = c.cgroupManager.GetStats(); err != nil {
return stats, newGenericError(err, SystemError)
return stats, newSystemError(err)
}
for _, iface := range c.config.Networks {
switch iface.Type {
case "veth":
istats, err := getNetworkInterfaceStats(iface.HostInterfaceName)
if err != nil {
return stats, newGenericError(err, SystemError)
return stats, newSystemError(err)
}
stats.Interfaces = append(stats.Interfaces, istats)
}
@ -128,20 +132,20 @@ func (c *linuxContainer) Start(process *Process) (int, error) {
if err != nil {
return -1, err
}
doInit := status == configs.Destroyed
doInit := status == Destroyed
parent, err := c.newParentProcess(process, doInit)
if err != nil {
return -1, err
return -1, newSystemError(err)
}
if err := parent.start(); err != nil {
// terminate the process to ensure that it properly is reaped.
if err := parent.terminate(); err != nil {
glog.Warning(err)
}
return -1, err
return -1, newSystemError(err)
}
if doInit {
c.initProcess = parent
c.updateState(parent)
}
return parent.pid(), nil
}
@ -149,11 +153,11 @@ func (c *linuxContainer) Start(process *Process) (int, error) {
func (c *linuxContainer) newParentProcess(p *Process, doInit bool) (parentProcess, error) {
parentPipe, childPipe, err := newPipe()
if err != nil {
return nil, err
return nil, newSystemError(err)
}
cmd, err := c.commandTemplate(p, childPipe)
if err != nil {
return nil, err
return nil, newSystemError(err)
}
if !doInit {
return c.newSetnsProcess(p, cmd, parentPipe, childPipe), nil
@ -171,7 +175,10 @@ func (c *linuxContainer) commandTemplate(p *Process, childPipe *os.File) (*exec.
cmd.SysProcAttr = &syscall.SysProcAttr{}
}
cmd.ExtraFiles = []*os.File{childPipe}
cmd.SysProcAttr.Pdeathsig = syscall.SIGKILL
if c.config.ParentDeathSignal > 0 {
cmd.SysProcAttr.Pdeathsig = syscall.Signal(c.config.ParentDeathSignal)
}
return cmd, nil
}
@ -254,11 +261,19 @@ func (c *linuxContainer) Destroy() error {
if err != nil {
return err
}
if status != configs.Destroyed {
if status != Destroyed {
return newGenericError(nil, ContainerNotStopped)
}
// TODO: remove cgroups
return os.RemoveAll(c.root)
if !c.config.Namespaces.Contains(configs.NEWPID) {
if err := killCgroupProcesses(c.cgroupManager); err != nil {
glog.Warning(err)
}
}
err = c.cgroupManager.Destroy()
if rerr := os.RemoveAll(c.root); err == nil {
err = rerr
}
return err
}
func (c *linuxContainer) Pause() error {
@ -270,11 +285,23 @@ func (c *linuxContainer) Resume() error {
}
func (c *linuxContainer) Signal(signal os.Signal) error {
glog.Infof("sending signal %d to pid %d", signal, c.initProcess.pid())
return c.initProcess.signal(signal)
}
// TODO: rename to be more descriptive
func (c *linuxContainer) NotifyOOM() (<-chan struct{}, error) {
return NotifyOnOOM(c.cgroupManager.GetPaths())
}
func (c *linuxContainer) updateState(process parentProcess) error {
c.initProcess = process
state, err := c.State()
if err != nil {
return err
}
f, err := os.Create(filepath.Join(c.root, stateFilename))
if err != nil {
return err
}
defer f.Close()
return json.NewEncoder(f).Encode(state)
}

View File

@ -18,7 +18,6 @@ import (
)
const (
configFilename = "config.json"
stateFilename = "state.json"
)
@ -65,23 +64,9 @@ func (l *linuxFactory) Create(id string, config *configs.Config) (Container, err
} else if !os.IsNotExist(err) {
return nil, newGenericError(err, SystemError)
}
data, err := json.MarshalIndent(config, "", "\t")
if err != nil {
return nil, newGenericError(err, SystemError)
}
if err := os.MkdirAll(containerRoot, 0700); err != nil {
return nil, newGenericError(err, SystemError)
}
f, err := os.Create(filepath.Join(containerRoot, configFilename))
if err != nil {
os.RemoveAll(containerRoot)
return nil, newGenericError(err, SystemError)
}
defer f.Close()
if _, err := f.Write(data); err != nil {
os.RemoveAll(containerRoot)
return nil, newGenericError(err, SystemError)
}
return &linuxContainer{
id: id,
root: containerRoot,
@ -96,13 +81,7 @@ func (l *linuxFactory) Load(id string) (Container, error) {
return nil, newGenericError(fmt.Errorf("invalid root"), ConfigInvalid)
}
containerRoot := filepath.Join(l.root, id)
glog.Infof("loading container config from %s", containerRoot)
config, err := l.loadContainerConfig(containerRoot)
if err != nil {
return nil, err
}
glog.Infof("loading container state from %s", containerRoot)
state, err := l.loadContainerState(containerRoot)
state, err := l.loadState(containerRoot)
if err != nil {
return nil, err
}
@ -110,12 +89,12 @@ func (l *linuxFactory) Load(id string) (Container, error) {
processPid: state.InitProcessPid,
processStartTime: state.InitProcessStartTime,
}
cgroupManager := cgroups.LoadCgroupManager(config.Cgroups, state.CgroupPaths)
cgroupManager := cgroups.LoadCgroupManager(state.Config.Cgroups, state.CgroupPaths)
glog.Infof("using %s as cgroup manager", cgroupManager)
return &linuxContainer{
initProcess: r,
id: id,
config: config,
config: &state.Config,
initArgs: l.initArgs,
cgroupManager: cgroupManager,
root: containerRoot,
@ -155,23 +134,7 @@ func (l *linuxFactory) StartInitialization(pipefd uintptr) (err error) {
return i.Init()
}
func (l *linuxFactory) loadContainerConfig(root string) (*configs.Config, error) {
f, err := os.Open(filepath.Join(root, configFilename))
if err != nil {
if os.IsNotExist(err) {
return nil, newGenericError(err, ContainerNotExists)
}
return nil, newGenericError(err, SystemError)
}
defer f.Close()
var config *configs.Config
if err := json.NewDecoder(f).Decode(&config); err != nil {
return nil, newGenericError(err, ConfigInvalid)
}
return config, nil
}
func (l *linuxFactory) loadContainerState(root string) (*State, error) {
func (l *linuxFactory) loadState(root string) (*State, error) {
f, err := os.Open(filepath.Join(root, stateFilename))
if err != nil {
if os.IsNotExist(err) {

View File

@ -82,14 +82,12 @@ func TestFactoryLoadContainer(t *testing.T) {
}
expectedState = &State{
InitProcessPid: 1024,
Config: *expectedConfig,
}
)
if err := os.Mkdir(filepath.Join(root, id), 0700); err != nil {
t.Fatal(err)
}
if err := marshal(filepath.Join(root, id, configFilename), expectedConfig); err != nil {
t.Fatal(err)
}
if err := marshal(filepath.Join(root, id, stateFilename), expectedState); err != nil {
t.Fatal(err)
}

View File

@ -9,11 +9,13 @@ import (
"strings"
"syscall"
"github.com/docker/libcontainer/cgroups"
"github.com/docker/libcontainer/configs"
"github.com/docker/libcontainer/netlink"
"github.com/docker/libcontainer/system"
"github.com/docker/libcontainer/user"
"github.com/docker/libcontainer/utils"
"github.com/golang/glog"
)
type initType string
@ -226,3 +228,35 @@ func setupRlimits(config *configs.Config) error {
}
return nil
}
// killCgroupProcesses freezes then itterates over all the processes inside the
// manager's cgroups sending a SIGKILL to each process then waiting for them to
// exit.
func killCgroupProcesses(m cgroups.Manager) error {
var procs []*os.Process
if err := m.Freeze(configs.Frozen); err != nil {
glog.Warning(err)
}
pids, err := m.GetPids()
if err != nil {
m.Freeze(configs.Thawed)
return err
}
for _, pid := range pids {
if p, err := os.FindProcess(pid); err == nil {
procs = append(procs, p)
if err := p.Kill(); err != nil {
glog.Warning(err)
}
}
}
if err := m.Freeze(configs.Thawed); err != nil {
glog.Warning(err)
}
for _, p := range procs {
if _, err := p.Wait(); err != nil {
glog.Warning(err)
}
}
return nil
}

View File

@ -11,7 +11,6 @@ import (
"syscall"
"github.com/docker/libcontainer/cgroups"
"github.com/docker/libcontainer/configs"
"github.com/docker/libcontainer/system"
"github.com/golang/glog"
)
@ -184,27 +183,7 @@ func (p *initProcess) wait() (*os.ProcessState, error) {
}
// we should kill all processes in cgroup when init is died if we use host PID namespace
if p.cmd.SysProcAttr.Cloneflags&syscall.CLONE_NEWPID == 0 {
// TODO: this will not work for the success path because libcontainer
// does not wait on the process. This needs to be moved to destroy or add a Wait()
// method back onto the container.
var procs []*os.Process
p.manager.Freeze(configs.Frozen)
pids, err := p.manager.GetPids()
if err != nil {
return nil, err
}
for _, pid := range pids {
// TODO: log err without aborting if we are unable to find
// a single PID
if p, err := os.FindProcess(pid); err == nil {
procs = append(procs, p)
p.Kill()
}
}
p.manager.Freeze(configs.Thawed)
for _, p := range procs {
p.Wait()
}
killCgroupProcesses(p.manager)
}
return state, nil
}

View File

@ -12,9 +12,6 @@ import (
var initCommand = cli.Command{
Name: "init",
Usage: "runs the init process inside the namespace",
Flags: []cli.Flag{
cli.IntFlag{Name: "fd", Value: 0, Usage: "internal pipe fd"},
},
Action: func(context *cli.Context) {
runtime.GOMAXPROCS(1)
runtime.LockOSThread()
@ -22,11 +19,7 @@ var initCommand = cli.Command{
if err != nil {
log.Fatal(err)
}
if context.Int("fd") == 0 {
log.Fatal("--fd must be specified for init process")
}
fd := uintptr(context.Int("fd"))
if err := factory.StartInitialization(fd); err != nil {
if err := factory.StartInitialization(3); err != nil {
log.Fatal(err)
}
panic("This line should never been executed")

View File

@ -24,7 +24,7 @@ func loadConfig(context *cli.Context) (*configs.Config, error) {
}
func loadFactory(context *cli.Context) (libcontainer.Factory, error) {
return libcontainer.New(context.GlobalString("root"), []string{os.Args[0], "init", "--fd", "3", "--"})
return libcontainer.New(context.GlobalString("root"), []string{os.Args[0], "init"})
}
func getContainer(context *cli.Context) (libcontainer.Container, error) {