moby/daemon/update.go
Qiang Huang 08c7075c40 Soften limitation of update kernel memory
Kernel memory is not allowed to be updated if container is
running, it's not actually a precise kernel limitation.

Before kernel version 4.6, kernel memory will not be accounted
until kernel memory limit is set, if a container created with
kernel memory initialized, kernel memory is accounted as soon
as process created in container, so kernel memory limit update
is allowed afterward. If kernel memory is not initialized,
kernel memory consumed by processes in container will not be
accounted, so we can't update the limit because the account
will be wrong.

So update kernel memory of a running container with kernel memory
initialized is allowed, we should soften the limitation by docker.

Signed-off-by: Qiang Huang <h.huangqiang@huawei.com>
2016-07-12 08:07:24 +08:00

90 lines
2.3 KiB
Go

package daemon
import (
"fmt"
"github.com/docker/engine-api/types/container"
)
// ContainerUpdate updates configuration of the container
func (daemon *Daemon) ContainerUpdate(name string, hostConfig *container.HostConfig, validateHostname bool) ([]string, error) {
var warnings []string
warnings, err := daemon.verifyContainerSettings(hostConfig, nil, true, validateHostname)
if err != nil {
return warnings, err
}
if err := daemon.update(name, hostConfig); err != nil {
return warnings, err
}
return warnings, nil
}
// ContainerUpdateCmdOnBuild updates Path and Args for the container with ID cID.
func (daemon *Daemon) ContainerUpdateCmdOnBuild(cID string, cmd []string) error {
if len(cmd) == 0 {
return nil
}
c, err := daemon.GetContainer(cID)
if err != nil {
return err
}
c.Path = cmd[0]
c.Args = cmd[1:]
return nil
}
func (daemon *Daemon) update(name string, hostConfig *container.HostConfig) error {
if hostConfig == nil {
return nil
}
container, err := daemon.GetContainer(name)
if err != nil {
return err
}
restoreConfig := false
backupHostConfig := *container.HostConfig
defer func() {
if restoreConfig {
container.Lock()
container.HostConfig = &backupHostConfig
container.ToDisk()
container.Unlock()
}
}()
if container.RemovalInProgress || container.Dead {
return errCannotUpdate(container.ID, fmt.Errorf("Container is marked for removal and cannot be \"update\"."))
}
if err := container.UpdateContainer(hostConfig); err != nil {
restoreConfig = true
return errCannotUpdate(container.ID, err)
}
// if Restart Policy changed, we need to update container monitor
container.UpdateMonitor(hostConfig.RestartPolicy)
// If container is not running, update hostConfig struct is enough,
// resources will be updated when the container is started again.
// If container is running (including paused), we need to update configs
// to the real world.
if container.IsRunning() && !container.IsRestarting() {
if err := daemon.containerd.UpdateResources(container.ID, toContainerdResources(hostConfig.Resources)); err != nil {
restoreConfig = true
return errCannotUpdate(container.ID, err)
}
}
daemon.LogContainerEvent(container, "update")
return nil
}
func errCannotUpdate(containerID string, err error) error {
return fmt.Errorf("Cannot update container %s: %v", containerID, err)
}