Merge pull request #36779 from sjenning/fix-memory-leak-via-terminated-pods

Automatic merge from submit-queue

fix leaking memory backed volumes of terminated pods

Currently, we allow volumes to remain mounted on the node, even though the pod is terminated.  This creates a vector for a malicious user to exhaust memory on the node by creating memory backed volumes containing large files.

This PR removes memory backed volumes (emptyDir w/ medium Memory, secrets, configmaps) of terminated pods from the node.

@saad-ali @derekwaynecarr
This commit is contained in:
Kubernetes Submit Queue 2016-11-17 21:29:51 -08:00 committed by GitHub
commit eca9e989a3
3 changed files with 148 additions and 3 deletions

View File

@ -129,10 +129,18 @@ func (dswp *desiredStateOfWorldPopulator) populatorLoopFunc() func() {
}
}
func isPodTerminated(pod *api.Pod) bool {
return pod.Status.Phase == api.PodFailed || pod.Status.Phase == api.PodSucceeded
}
// Iterate through all pods and add to desired state of world if they don't
// exist but should
func (dswp *desiredStateOfWorldPopulator) findAndAddNewPods() {
for _, pod := range dswp.podManager.GetPods() {
if isPodTerminated(pod) {
// Do not (re)add volumes for terminated pods
continue
}
dswp.processPodVolumes(pod)
}
}
@ -144,9 +152,18 @@ func (dswp *desiredStateOfWorldPopulator) findAndRemoveDeletedPods() {
runningPodsFetched := false
for _, volumeToMount := range dswp.desiredStateOfWorld.GetVolumesToMount() {
if _, podExists :=
dswp.podManager.GetPodByUID(volumeToMount.Pod.UID); podExists {
continue
pod, podExists := dswp.podManager.GetPodByUID(volumeToMount.Pod.UID)
if podExists {
// Skip running pods
if !isPodTerminated(pod) {
continue
}
// Skip non-memory backed volumes belonging to terminated pods
volume := volumeToMount.VolumeSpec.Volume
if (volume.EmptyDir == nil || volume.EmptyDir.Medium != api.StorageMediumMemory) &&
volume.ConfigMap == nil && volume.Secret == nil {
continue
}
}
// Once a pod has been deleted from kubelet pod manager, do not delete

View File

@ -68,6 +68,7 @@ go_test(
"restart_test.go",
"runtime_conformance_test.go",
"summary_test.go",
"volume_manager_test.go",
],
library = "go_default_library",
tags = ["automanaged"],

View File

@ -0,0 +1,127 @@
/*
Copyright 2016 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package e2e_node
import (
"time"
"k8s.io/kubernetes/pkg/api"
"k8s.io/kubernetes/pkg/util/uuid"
"k8s.io/kubernetes/test/e2e/framework"
"fmt"
. "github.com/onsi/ginkgo"
. "github.com/onsi/gomega"
)
var _ = framework.KubeDescribe("Kubelet Volume Manager", func() {
f := framework.NewDefaultFramework("kubelet-volume-manager")
Describe("Volume Manager", func() {
Context("On terminatation of pod with memory backed volume", func() {
It("should remove the volume from the node", func() {
var (
memoryBackedPod *api.Pod
volumeName string
)
By("Creating a pod with a memory backed volume that exits success without restart", func() {
volumeName = "memory-volume"
memoryBackedPod = f.PodClient().Create(&api.Pod{
ObjectMeta: api.ObjectMeta{
Name: "pod" + string(uuid.NewUUID()),
Namespace: f.Namespace.Name,
},
Spec: api.PodSpec{
RestartPolicy: api.RestartPolicyNever,
Containers: []api.Container{
{
Image: "gcr.io/google_containers/busybox:1.24",
Name: "container" + string(uuid.NewUUID()),
Command: []string{"sh", "-c", "echo"},
VolumeMounts: []api.VolumeMount{
{
Name: volumeName,
MountPath: "/tmp",
},
},
},
},
Volumes: []api.Volume{
{
Name: volumeName,
VolumeSource: api.VolumeSource{
EmptyDir: &api.EmptyDirVolumeSource{Medium: api.StorageMediumMemory},
},
},
},
},
})
err := framework.WaitForPodSuccessInNamespace(f.ClientSet, memoryBackedPod.Name, f.Namespace.Name)
Expect(err).NotTo(HaveOccurred())
})
By("Verifying the memory backed volume was removed from node", func() {
volumePath := fmt.Sprintf("/tmp/%s/volumes/kubernetes.io~empty-dir/%s", string(memoryBackedPod.UID), volumeName)
var err error
for i := 0; i < 10; i++ {
// need to create a new verification pod on each pass since updates
//to the HostPath volume aren't propogated to the pod
pod := f.PodClient().Create(&api.Pod{
ObjectMeta: api.ObjectMeta{
Name: "pod" + string(uuid.NewUUID()),
Namespace: f.Namespace.Name,
},
Spec: api.PodSpec{
RestartPolicy: api.RestartPolicyNever,
Containers: []api.Container{
{
Image: "gcr.io/google_containers/busybox:1.24",
Name: "container" + string(uuid.NewUUID()),
Command: []string{"sh", "-c", "if [ -d " + volumePath + " ]; then exit 1; fi;"},
VolumeMounts: []api.VolumeMount{
{
Name: "kubelet-pods",
MountPath: "/tmp",
},
},
},
},
Volumes: []api.Volume{
{
Name: "kubelet-pods",
VolumeSource: api.VolumeSource{
// TODO: remove hardcoded kubelet volume directory path
// framework.TestContext.KubeVolumeDir is currently not populated for node e2e
HostPath: &api.HostPathVolumeSource{Path: "/var/lib/kubelet/pods"},
},
},
},
},
})
err = framework.WaitForPodSuccessInNamespace(f.ClientSet, pod.Name, f.Namespace.Name)
gp := int64(1)
f.PodClient().Delete(pod.Name, &api.DeleteOptions{GracePeriodSeconds: &gp})
if err == nil {
break
}
<-time.After(10 * time.Second)
}
Expect(err).NotTo(HaveOccurred())
})
})
})
})
})