This PR adds two features:
1. add support for isolating the emptyDir volume use. If user
sets a size limit for emptyDir volume, kubelet's eviction manager
monitors its usage
and evict the pod if the usage exceeds the limit.
2. add support for isolating the local storage for container overlay. If
the container's overly usage exceeds the limit defined in container
spec, eviction manager will evict the pod.
This PR adds the check for local storage request when admitting pods. If
the local storage request exceeds the available resource, pod will be
rejected.
Automatic merge from submit-queue
node e2e: improve the validate OOM score test for infra containers
The test blindly checked all "pause" processes on the node, assuming
they were all infra containers. This change takes a snapshot of all
existing "pause" processes on the node, and exclude them in the
validation. The test still relies on the fact that it runs exclusively
on the node. If that assumption changes, we will need other methods to
locate the PIDs of the infra containers.
This fixes#37580
Automatic merge from submit-queue
test/e2e_node: prepull images with CRI
Part of https://github.com/kubernetes/kubernetes/issues/40739
- This PR builds on top of #40525 (and contains one commit from #40525)
- The second commit contains a tiny change in the `Makefile`.
- Third commit is a patch to be able to prepull images using the CRI (as opposed to run `docker` to pull images which doesn't make sense if you're using CRI most of the times)
Marked WIP till #40525 makes its way into master
@Random-Liu @lucab @yujuhong @mrunalp @rhatdan
bazel update
added new files to reflect that only one method has changed between arch types.
forgot to add changes to a commit.
changes made and gfmt run.
changed node_problem_detector to node_problem_detector_linux and made it linux only.
updated bazel
The test blindly checked all "pause" processes on the node, assuming
they were all infra containers. This change takes a snapshot of all
existing "pause" processes on the node, and exclude them in the
validation. The test still relies on the fact that it runs exclusively
on the node. If that assumption changes, we will need other methods to
locate the PIDs of the infra containers.
Automatic merge from submit-queue
Eviction Manager Enforces Allocatable Thresholds
This PR modifies the eviction manager to enforce node allocatable thresholds for memory as described in kubernetes/community#348.
This PR should be merged after #41234.
cc @kubernetes/sig-node-pr-reviews @kubernetes/sig-node-feature-requests @vishh
** Why is this a bug/regression**
Kubelet uses `oom_score_adj` to enforce QoS policies. But the `oom_score_adj` is based on overall memory requested, which means that a Burstable pod that requested a lot of memory can lead to OOM kills for Guaranteed pods, which violates QoS. Even worse, we have observed system daemons like kubelet or kube-proxy being killed by the OOM killer.
Without this PR, v1.6 will have node stability issues and regressions in an existing GA feature `out of Resource` handling.
Automatic merge from submit-queue (batch tested with PRs 41984, 41682, 41924, 41928)
Move node problem detector test into node e2e.
Move current NPD e2e test into node e2e.
In fact, current NPD e2e test is only a functionality test for NPD. It creates test NPD pod, sets test configuration, generates test logs and verifies test result.
It doesn't actually test the NPD really deployed in the cluster.
So it doesn't actually need to run in cluster e2e. Running it in node e2e will:
1) Make it easier to run the test.
2) Make it more light weight to introduce this as a pre/post submit test in NPD repo in the future.
Except this, I'm working on a cluster e2e to run some basic functionality test and benchmark test against the real NPD deployed in the cluster. Will send the PR later.
/cc @dchen1107 @kubernetes/node-problem-detector-reviewers
Automatic merge from submit-queue
Build release tars using bazel
**What this PR does / why we need it**: builds equivalents of the various kubernetes release tarballs, solely using bazel.
For example, you can now do
```console
$ make bazel-release
$ hack/e2e.go -v -up -test -down
```
**Special notes for your reviewer**: this is currently dependent on 3b29803eb5, which I have yet to turn into a pull request, since I'm still trying to figure out if this is the best approach.
Basically, the issue comes up with the way we generate the various server docker image tarfiles and load them on nodes:
* we `md5sum` the binary being encapsulated (e.g. kube-proxy) and save that to `$binary.docker_tag` in the server tarball
* we then build the docker image and tag using that md5sum (e.g. `gcr.io/google_containers/kube-proxy:$MD5SUM`)
* we `docker save` this image, which embeds the full tag in the `$binary.tar` file.
* on cluster startup, we `docker load` these tarballs, which are loaded with the tag that we'd created at build time. the nodes then use the `$binary.docker_tag` file to find the right image.
With the current bazel `docker_build` rule, the tag isn't saved in the docker image tar, so the node is unable to find the image after `docker load`ing it.
My changes to the rule save the tag in the docker image tar, though I don't know if there are subtle issues with it. (Maybe we want to only tag when `--stamp` is given?)
Also, the docker images produced by bazel have the timestamp set to the unix epoch, which is not great for debugging. Might be another thing to change with a `--stamp`.
Long story short, we probably need to follow up with bazel folks on the best way to solve this problem.
**Release note**:
```release-note
NONE
```
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
Automatic merge from submit-queue
Node Conformance Test: Add system verification
For #30122 and #29081.
This PR introduces system verification test in node e2e and conformance test. It will run before the real test. Once the system verification fails, the test will just fail. The output of the system verification is like this:
```
I0909 23:33:20.622122 2717 validators.go:45] Validating os...
OS: Linux
I0909 23:33:20.623274 2717 validators.go:45] Validating kernel...
I0909 23:33:20.624037 2717 kernel_validator.go:79] Validating kernel version
KERNEL_VERSION: 3.16.0-4-amd64
I0909 23:33:20.624146 2717 kernel_validator.go:93] Validating kernel config
CONFIG_NAMESPACES: enabled
CONFIG_NET_NS: enabled
CONFIG_PID_NS: enabled
CONFIG_IPC_NS: enabled
CONFIG_UTS_NS: enabled
CONFIG_CGROUPS: enabled
CONFIG_CGROUP_CPUACCT: enabled
CONFIG_CGROUP_DEVICE: enabled
CONFIG_CGROUP_FREEZER: enabled
CONFIG_CGROUP_SCHED: enabled
CONFIG_CPUSETS: enabled
CONFIG_MEMCG: enabled
I0909 23:33:20.679328 2717 validators.go:45] Validating cgroups...
CGROUPS_CPU: enabled
CGROUPS_CPUACCT: enabled
CGROUPS_CPUSET: enabled
CGROUPS_DEVICES: enabled
CGROUPS_FREEZER: enabled
CGROUPS_MEMORY: enabled
I0909 23:33:20.679454 2717 validators.go:45] Validating docker...
DOCKER_GRAPH_DRIVER: aufs
```
It verifies the system following a predefined `SysSpec`:
``` go
// DefaultSysSpec is the default SysSpec.
var DefaultSysSpec = SysSpec{
OS: "Linux",
KernelVersion: []string{`3\.[1-9][0-9].*`, `4\..*`}, // Requires 3.10+ or 4+
// TODO(random-liu): Add more config
KernelConfig: KernelConfig{
Required: []string{
"NAMESPACES", "NET_NS", "PID_NS", "IPC_NS", "UTS_NS",
"CGROUPS", "CGROUP_CPUACCT", "CGROUP_DEVICE", "CGROUP_FREEZER",
"CGROUP_SCHED", "CPUSETS", "MEMCG",
},
Forbidden: []string{},
},
Cgroups: []string{"cpu", "cpuacct", "cpuset", "devices", "freezer", "memory"},
RuntimeSpec: RuntimeSpec{
DockerSpec: &DockerSpec{
Version: []string{`1\.(9|\d{2,})\..*`}, // Requires 1.9+
GraphDriver: []string{"aufs", "overlay", "devicemapper"},
},
},
}
```
Currently, it only supports:
- Kernel validation: version validation and kernel configuration validation
- Cgroup validation: validating whether required cgroups subsystems are enabled.
- Runtime Validation: currently, only validates docker graph driver.
The validating framework is ready. The specific validation items could be added over time.
@dchen1107
/cc @kubernetes/sig-node
Automatic merge from submit-queue
[AppArmor] Hold bad AppArmor pods in pending rather than rejecting
Fixes https://github.com/kubernetes/kubernetes/issues/32837
Overview of the fix:
If the Kubelet needs to reject a Pod for a reason that the control plane doesn't understand (e.g. which AppArmor profiles are installed on the node), then it might contiinuously try to run the pod on the same rejecting node. This change adds a concept of "soft rejection", in which the Pod is admitted, but not allowed to run (and therefore held in a pending state). This prevents the pod from being retried on other nodes, but also prevents the high churn. This is consistent with how other missing local resources (e.g. volumes) is handled.
A side effect of the change is that Pods which are not initially runnable will be retried. This is desired behavior since it avoids a race condition when a new node is brought up but the AppArmor profiles have not yet been loaded on it.
``` release-note
Pods with invalid AppArmor configurations will be held in a Pending state, rather than rejected (failed). Check the pod status message to find out why it is not running.
```
@kubernetes/sig-node @timothysc @rrati @davidopp
Jing Xu