Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

pkg/kubelet: recreate infra pod if the pod is changed #4563

Merged
merged 1 commit into from
Feb 20, 2015
Merged

pkg/kubelet: recreate infra pod if the pod is changed #4563

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
30 changes: 18 additions & 12 deletions pkg/kubelet/kubelet.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1047,6 +1047,7 @@ func (kl *Kubelet) syncPod(pod *api.BoundPod, dockerContainers dockertools.Docke
glog.Errorf("Unable to get pod with name %q and uid %q info with error(%v)", podFullName, uid, err)
}

podChanged := false
for _, container := range pod.Spec.Containers {
expectedHash := dockertools.HashContainer(&container)
dockerContainerName := dockertools.BuildDockerName(uid, podFullName, &container)
Expand All @@ -1055,8 +1056,8 @@ func (kl *Kubelet) syncPod(pod *api.BoundPod, dockerContainers dockertools.Docke
glog.V(3).Infof("pod %q container %q exists as %v", podFullName, container.Name, containerID)

// look for changes in the container.
podChanged := hash != 0 && hash != expectedHash
if !podChanged {
containerChanged := hash != 0 && hash != expectedHash
if !containerChanged {
// TODO: This should probably be separated out into a separate goroutine.
// If the container's liveness probe is unsuccessful, set readiness to false. If liveness is succesful, do a readiness check and set
// readiness accordingly. If the initalDelay since container creation on liveness probe has not passed the probe will return Success.
Expand Down Expand Up @@ -1088,23 +1089,14 @@ func (kl *Kubelet) syncPod(pod *api.BoundPod, dockerContainers dockertools.Docke
}
glog.Infof("pod %q container %q is unhealthy. Container will be killed and re-created.", podFullName, container.Name, live)
} else {
podChanged = true
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why would we restart the pod if just one container changed?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ugh, there is an escape here - if a container adds/removes/changes a host port, we won't restart the network container. I'm not too worried about that by itself, but are there any other aspects of a single container that might require the pod itself to be recreated?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@thockin previously, we kill the infra pod container whenever there is a container change. The infra container will not be recreated until the next sync happens. This pull request does not change the killing behavior. The only change is to recreate the killed infra container right after we finish the sync.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You're right. That's awful

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The only reason killing infra pod container when a container is changed is that we don't know if network related config is changed, such as port.

glog.Infof("pod %q container %q hash changed (%d vs %d). Container will be killed and re-created.", podFullName, container.Name, hash, expectedHash)
}
if err := kl.killContainer(dockerContainer); err != nil {
glog.V(1).Infof("Failed to kill container %q: %v", dockerContainer.ID, err)
continue
}
killedContainers[containerID] = empty{}

if podChanged {
// Also kill associated pod infra container if the pod changed.
if podInfraContainer, found, _ := dockerContainers.FindPodContainer(podFullName, uid, dockertools.PodInfraContainerName); found {
if err := kl.killContainer(podInfraContainer); err != nil {
glog.V(1).Infof("Failed to kill pod infra container %q: %v", podInfraContainer.ID, err)
continue
}
}
}
}

// Check RestartPolicy for container
Expand Down Expand Up @@ -1167,6 +1159,20 @@ func (kl *Kubelet) syncPod(pod *api.BoundPod, dockerContainers dockertools.Docke
containersToKeep[containerID] = empty{}
}

if podChanged {
// Also kill associated pod infra container if the pod changed.
if podInfraContainer, found, _ := dockerContainers.FindPodContainer(podFullName, uid, dockertools.PodInfraContainerName); found {
if err := kl.killContainer(podInfraContainer); err != nil {
glog.V(1).Infof("Failed to kill pod infra container %q: %v", podInfraContainer.ID, err)
}
}
podInfraContainerID, err = kl.createPodInfraContainer(pod)
if err != nil {
glog.Errorf("Failed to recreate pod infra container: %v for pod %q", err, podFullName)
}
containersToKeep[podInfraContainerID] = empty{}
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this necessary? At the beginning of Sync Loop, we should already put PodInfraContainerID to containersToKeep, right?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

podInfraContainerID is returned by createPodInfraContainer. It is called at line 1169.

I mean they are the same variable name but might have different IDs. (I assume docker will not generate duplicate IDs for containers)

}

// Kill any containers in this pod which were not identified above (guards against duplicates).
for id, container := range dockerContainers {
curPodFullName, curUUID, _, _ := dockertools.ParseDockerName(container.Names[0])
Expand Down
2 changes: 1 addition & 1 deletion pkg/kubelet/kubelet_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -931,7 +931,7 @@ func TestSyncPodBadHash(t *testing.T) {
t.Errorf("unexpected error: %v", err)
}

verifyCalls(t, fakeDocker, []string{"list", "stop", "stop", "list", "create", "start"})
verifyCalls(t, fakeDocker, []string{"list", "stop", "list", "create", "start", "stop", "create", "start"})

// A map interation is used to delete containers, so must not depend on
// order here.
Expand Down