Automatic merge from submit-queue (batch tested with PRs 37228, 40146, 40075, 38789, 40189)
kubelet: storage: teardown terminated pod volumes
This is a continuation of the work done in https://github.com/kubernetes/kubernetes/pull/36779
There really is no reason to keep volumes for terminated pods attached on the node. This PR extends the removal of volumes on the node from memory-backed (the current policy) to all volumes.
@pmorie raised a concern an impact debugging volume related issues if terminated pod volumes are removed. To address this issue, the PR adds a `--keep-terminated-pod-volumes` flag the kubelet and sets it for `hack/local-up-cluster.sh`.
For consideration in 1.6.
Fixes#35406
@derekwaynecarr @vishh @dashpole
```release-note
kubelet tears down pod volumes on pod termination rather than pod deletion
```
Automatic merge from submit-queue
kubelet bootstrap: start hostNetwork pods before we have PodCIDR
Network readiness was checked in the pod admission phase, but pods that
fail admission are not retried. Move the check to the pod start phase.
Issue #35409
Issue #35521
We had another bug where we confused the hostname with the NodeName.
To avoid this happening again, and to make the code more
self-documenting, we use types.NodeName (a typedef alias for string)
whenever we are referring to the Node.Name.
A tedious but mechanical commit therefore, to change all uses of the
node name to use types.NodeName
Also clean up some of the (many) places where the NodeName is referred
to as a hostname (not true on AWS), or an instanceID (not true on GCE),
etc.
MTU selection is difficult, and if there is a transport such as IPSEC in
use may be impossible. So we allow specification of the MTU with the
network-plugin-mtu flag, and we pass this down into the network
provider.
Currently implemented by kubenet.
Currently kubelet volume management works on the concept of desired
and actual world of states. The volume manager periodically compares the
two worlds and perform volume mount/unmount and/or attach/detach
operations. When kubelet restarts, the cache of those two worlds are
gone. Although desired world can be recovered through apiserver, actual
world can not be recovered which may cause some volumes cannot be cleaned
up if their information is deleted by apiserver. This change adds the
reconstruction of the actual world by reading the pod directories from
disk. The reconstructed volume information is added to both desired
world and actual world if it cannot be found in either world. The rest
logic would be as same as before, desired world populator may clean up
the volume entry if it is no longer in apiserver, and then volume
manager should invoke unmount to clean it up.
This commit adds a new volume manager in kubelet that synchronizes
volume mount/unmount (and attach/detach, if attach/detach controller
is not enabled).
This eliminates the race conditions between the pod creation loop
and the orphaned volumes loops. It also removes the unmount/detach
from the `syncPod()` path so volume clean up never blocks the
`syncPod` loop.
DONE:
1. refactor all predicates: predicates return fitOrNot(bool) and error(Error) in which the latter is of type
PredicateFailureError or InsufficientResourceError. (For violation of either MaxEBSVolumeCount or
MaxGCEPDVolumeCount, returns one same error type as ErrMaxVolumeCountExceeded)
2. GeneralPredicates() is a predicate function, which includes serveral other predicate functions (PodFitsResource,
PodFitsHost, PodFitsHostPort). It is registered as one of the predicates in DefaultAlgorithmProvider, and
is also called in canAdmitPod() in Kubelet and should be called by other components (like rescheduler, etc)
if necessary. See discussion in issue #12744
3. remove podNumber check from GeneralPredicates
4. HostName is now verified in Kubelet's canAdminPod(). add TestHostNameConflicts in kubelet_test.go
5. add getNodeAnyWay() method in Kubelet to get node information in standaloneMode
TODO:
1. determine which predicates should be included in GeneralPredicates()
2. separate GeneralPredicates() into:
a. GeneralPredicatesEvictPod() and
b. GeneralPredicatesNotEvictPod()
3. DaemonSet should use GeneralPredicates()
Kubelet doesn't perform checkpointing and loses all its internal states after
restarts. It'd then mistaken pods from the api server as new pods and attempt
to go through the admission process. This may result in pods being rejected
even though they are running on the node (e.g., out of disk situation). This
change adds a condition to check whether the pod was seen before and categorize
such pods as updates. The change also removes freeze/unfreeze mechanism used to
work around such cases, since it is no longer needed and it stopped working
correctly ever since we switched to incremental updates.
- status.Manager always deals with the local (static) pod, but gets the
mirror pod when syncing
- This lets components like the probe workers ignore mirror pods
Each container with a readiness has an individual go-routine which
handles periodic probing for that container. The results are cached, and
written to the status.Manager in the pod sync path.
This refactor is in preparation for moving more state handling to the
status manager. It will become the canonical cache for the latest
information on running containers and probe status, as part of the
prober refactoring.
Currently, whenever there is any update, kubelet would force all pod workers to
sync again, causing resource contention and hence performance degradation.
This commit flips kubelet to use incremental updates (as opposed to snapshots).
This allows us to know what pods have changed and send updates to those pod
workers only. The `SyncPods` function has been replaced with individual
handlers, each handling an operation (ADD, REMOVE, UPDATE). Pod workers are
still triggered periodically, and kubelet performs periodic cleanup as well.
This commit also spawns a new goroutine solely responsible for killing pods.
This is necessary because pod killing could hold up the sync loop for
indefinitely long amount of time now user can define the graceful termination
period in the container spec.
This generalizes the handling of containers in the
ContainerManager.
Also introduces the ability to determine how much
resources are reserved for those system containers.
This change instructs kubelet to switch to using the Runtime interface. In order
to do it, the change moves the Prober instantiation to DockerManager.
Note that most of the tests in kubelet_test.go needs to be migrated to
dockertools. For now, we use type assertion to convert the Runtime interface to
DockerManager in most tests.
We want to stop leaking more docker details into kubelet, and we also want to
consolidate some of the existing docker interfaces/structs. This change creates
DockerManager as the new home of some functions in dockertools/docker.go. It
also absorbs containerRunner. In addition, GetDockerPodStatus is renamed to
GetPodStatus with the entire pod passed to it so that it is simialr to the what
is defined in the container Runtime interface.
Eventually, DockerManager should implement the container Runtime interface, and
integrate DockerCache with a flag to turn on/off caching. Code in kubelet.go
should not be using docker client directly.
Per-pod worker syncs the pod and container status, and write the pod status in
the pod status cache. Given that it already owns a copy of the pod, it can
bypass the additional pod lookup step completely. This change adds a new
generatePodStatusByPod() method to achieve this. In general, per-pod worker
should avoid accessing the internal pod array completely, as this would may
lead to high contention.
This change also changes the return type of GetPodByFullName to reflect the
name, and consolidates GetPodByFullName() and GetPodByName().
