The kubeadm documentation instructs users to set the container
runtime driver to "systemd", since kubeadm manages a kubelet via
the systemd init system. The kubelet default however is "cgroupfs".
For new clusters set the driver to "systemd" unless the user
is explicit about it. The same defaulting would not happen
during "upgrade".
Errors from staticcheck:
cmd/preferredimports/preferredimports.go:38:2:
package golang.org/x/crypto/ssh/terminal is deprecated:
this package moved to golang.org/x/term. (SA1019)
vendor/k8s.io/client-go/plugin/pkg/client/auth/exec/exec.go:36:2:
package golang.org/x/crypto/ssh/terminal is deprecated:
this package moved to golang.org/x/term. (SA1019)
vendor/k8s.io/client-go/tools/clientcmd/auth_loaders.go:26:2:
package golang.org/x/crypto/ssh/terminal is deprecated:
this package moved to golang.org/x/term. (SA1019)
Please review the above warnings. You can test via:
hack/verify-staticcheck.sh <failing package>
If the above warnings do not make sense, you can exempt the line or
file. See:
https://staticcheck.io/docs/#ignoring-problems
generated:
- hack/update-internal-modules.sh
- hack/lint-dependencies.sh
- hack/update-vendor.sh
Signed-off-by: Stephen Augustus <foo@auggie.dev>
Pass the flag --pod-infra-container-image to the kubelet not only
for Docker but for all CRs.
This flag tells the kubelet to special case the image and not garbage
collect it.
This change updates the number of workers that the CSR signing
controllers use. If a large number of certificates (especially
short lived ones) are approved at the same time, it can take the
signing controllers a long time to process them serially. The
NewCSRSigningController logic is already go routine safe.
Signed-off-by: Monis Khan <mok@vmware.com>
Looks like there is a bit of an issue in the Bluderbuss (Prow plugin)
where it prefers to pick reviewers from a parent OWNERS files,
instead of using an approver from a current OWNERS file as
an additional reviewer.
The dependencycheck binary name was vendorcheck, which was the original
name of the tool. This updates it to dependencycheck.
Signed-off-by: hasheddan <georgedanielmangum@gmail.com>
As a part of cleaning up inactive members (those with no activity within
the past 18 months) from OWNERS files, this commit moves gmarek from an
approver to an emeritus_approver.
The new flag will parse the `--reserved-memory` flag straight forward
to the []kubeletconfig.MemoryReservation variable instead of parsing
it to the middle map representation.
It gives us possibility to get rid of a lot of unneeded code and use the single
presentation for the reserved-memory.
Signed-off-by: Artyom Lukianov <alukiano@redhat.com>
Updates kubeadm version resolution to use kubernetes community infra
bucket to fetch appropriate k8s ci versions. The images are already
being pulled from the kubernetes community infra bucket meaning that a
mismatch can occur when the ci version is fetched from the google infra
bucket and the image is not yet present on k8s infra.
Follow-up to kubernetes/kubernetes#97087
Signed-off-by: hasheddan <georgedanielmangum@gmail.com>
Originally raised as an issue with invalid versions to plan, but it has
been determined with air gapped environments and development versions it
is not possible to fully address that issue.
But one thing that was identified was that we can do a better job in how
we output the upgrade plan information. Kubeadm outputs the requested
version as "Latest stable version", though that may not actually be the
case. For this instance, we want to change this to "Target version" to
be a little more accurate.
Then in the component upgrade table that is emitted, the last column of
AVAILABLE isn't quite right either. Also changing this to TARGET to
reflect that this is the version we are targetting to upgrade to,
regardless of its availability.
There could be some improvements in checking available versions,
particularly in air gapped environments, to make sure we actually have
access to the requested version. But this at least clarifies some of the
output a bit.
Signed-off-by: Sean McGinnis <sean.mcginnis@gmail.com>
we disabled the /healthz check because our test blocks one post-start
hook from finishing. Instead we should check all the other /healthz/...
endpoints before running the tests
Add DefaultedStaticInitConfiguration() which can be
used instead of DefaultedInitConfiguration() during unit tests.
The later can be slow since it performs dynamic defaulting.
Apply the label:
"node.kubernetes.io/exclude-from-external-load-balancers"
To control plane nodes to preserve backwards compatibility
with the legacy mode where "master" nodes were excluded from
LBs.
This commit replaces the CSIMigrationXXXComplete flag
with InTreePluginXXUnregister flag. This new flag will
be a superset of the CSIMigrationXXXComplete. But this
decouple the plugin unregister from CSI migration. So
if a K8s distribution want to go directly with CSI and
do not support in-tree, they can use this flag directly.
Testing:
1. Enable the InTreePluginXXUnregister and not CSIMigrationXXX,
verify that the PVC using old plugin name will have error
saying cannot find the plugin
2. Enable both the InTreePluginXXUnregister and CSIMigrationXXX
verify that the PVC using old plugin name will start to use
the migrated CSI plugin
Migrate how resource lock and leader election config is generated to new way, hidding kubeClient. This also halfs kubeClient timeout, making it an useful value.
If timeout is equal to RenewDeadline and we hit client timeout on request, there will be no retry, as RenewDeadline part will cancel the context and lose leader election. So setting a timeout to value at least equal to RenewDeadline is pointless.
Setting it as half of RenewDeadline is a heuristic to resolve this missing retry problem without adding additional parameter.
Migrate how resource lock and leader election config is generated to new
way, hidding kubeClient. This also halfs kubeClient timeout, making it
an useful value.
During upgrade the coredns migration library seems to require
that the input version doesn't have the "v" prefix".
Fixes a bug where the user cannot run commands such as
"kubeadm upgrade plan" if they have `v1.8.0` installed.
Assuming this is caused by the fact that previously the image didn't
have a "v" prefix.
Fixes an issue where some kubeadm phases fail if a certificate file
contains a certificate chain with one or more intermediate CA
certificates. The validation algorithm has been changed from requiring
that a certificate was signed directly by the root CA to requiring that
there is a valid certificate chain back to the root CA.
In kubeadm etcd join there is a a bug that exists where,
if a peer already exists in etcd, it attempts to mitigate
by continuing and generating the etcd manifest file. However,
this existing "member name" may actually be unset, causing
subsequent etcd consistency checks to fail.
This change checks if the member name is empty - if it is,
it sets the member name to the node name, and resumes.
The error messages when the user feeds an invalid discovery token CA
hash are vague. Make sure to:
- Print the list of supported hash formats (currently only "sha256").
- Wrap the error from pubKeyPins.Allow() with a descriptive message.
Implement pod resource metrics as described in KEP 1916. The new
`/metrics/resources` endpoint is exposed on the active scheduler
and reports kube_pod_resources* metrics that present the effective
requests and limits for all resources on the pods as calculated by
the scheduler and kubelet. This allows administrators using the
system to quickly perform resource consumption, reservation, and
pending utilization calculations when those metrics are read.
Because metrics calculation is on-demand, there is no additional
resource consumption incurred by the scheduler unless the endpoint
is scraped.
- Mark the "node-role.kubernetes.io/master" key for labels
and taints as deprecated.
- During "kubeadm init/join" apply the label
"node-role.kubernetes.io/control-plane" to new control-plane nodes,
next to the existing "node-role.kubernetes.io/master" label.
- During "kubeadm upgrade apply", find all Nodes with the "master"
label and also apply the "control-plane" label to them
(if they don't have it).
- During upgrade health-checks collect Nodes labeled both "master"
and "control-plane".
- Rename the constants.ControlPlane{Taint|Toleraton} to
constants.OldControlPlane{Taint|Toleraton} to manage the transition.
- Mark constants.OldControlPlane{{Taint|Toleraton} as deprecated.
- Use constants.OldControlPlane{{Taint|Toleraton} instead of
constants.ControlPlane{Taint|Toleraton} everywhere.
- Introduce constants.ControlPlane{Taint|Toleraton}.
- Add constants.ControlPlaneToleraton to the kube-dns / CoreDNS
Deployments to make them anticipate the introduction
of the "node-role.kubernetes.io/control-plane:NoSchedule"
taint (constants.ControlPlaneTaint) on kubeadm control-plane Nodes.
Aborted requests are the ones that were disrupted with http.ErrAbortHandler.
For example, the timeout handler will panic with http.ErrAbortHandler when a response to the client has been already sent
and the timeout elapsed.
Additionally, a new metric requestAbortsTotal was defined to count aborted requests. The new metric allows for aggregation for each group, version, verb, resource, subresource and scope.
without APIServerIdentity enabled, stale apiserver leases won't be GC'ed
and the same for stale storage version entries. In that case the storage
migrator won't operate correctly without manual intervention.
To make sure that the storage version filter can block certain requests until
the storage version updates are completed, and that the apiserver works
properly after the storage version updates are done.
StorageVersions are updated during apiserver bootstrap.
Also add a poststarthook to the aggregator which updates the
StorageVersions via the storageversion.Manager
Previously no timeout was set. Requests without explicit timeout might potentially hang forever and lead to starvation of the application.
When no timeout was specified a default one will be applied.
* api: structure change
* api: defaulting, conversion, and validation
* [FIX] validation: auto remove second ip/family when service changes to SingleStack
* [FIX] api: defaulting, conversion, and validation
* api-server: clusterIPs alloc, printers, storage and strategy
* [FIX] clusterIPs default on read
* alloc: auto remove second ip/family when service changes to SingleStack
* api-server: repair loop handling for clusterIPs
* api-server: force kubernetes default service into single stack
* api-server: tie dualstack feature flag with endpoint feature flag
* controller-manager: feature flag, endpoint, and endpointSlice controllers handling multi family service
* [FIX] controller-manager: feature flag, endpoint, and endpointSlicecontrollers handling multi family service
* kube-proxy: feature-flag, utils, proxier, and meta proxier
* [FIX] kubeproxy: call both proxier at the same time
* kubenet: remove forced pod IP sorting
* kubectl: modify describe to include ClusterIPs, IPFamilies, and IPFamilyPolicy
* e2e: fix tests that depends on IPFamily field AND add dual stack tests
* e2e: fix expected error message for ClusterIP immutability
* add integration tests for dualstack
the third phase of dual stack is a very complex change in the API,
basically it introduces Dual Stack services. Main changes are:
- It pluralizes the Service IPFamily field to IPFamilies,
and removes the singular field.
- It introduces a new field IPFamilyPolicyType that can take
3 values to express the "dual-stack(mad)ness" of the cluster:
SingleStack, PreferDualStack and RequireDualStack
- It pluralizes ClusterIP to ClusterIPs.
The goal is to add coverage to the services API operations,
taking into account the 6 different modes a cluster can have:
- single stack: IP4 or IPv6 (as of today)
- dual stack: IPv4 only, IPv6 only, IPv4 - IPv6, IPv6 - IPv4
* [FIX] add integration tests for dualstack
* generated data
* generated files
Co-authored-by: Antonio Ojea <aojea@redhat.com>
the controller manager should validate the podSubnet against the node-mask
because if they are incorrect can cause the controller-manager to fail.
We don't need to calculate the node-cidr-masks, because those should
be provided by the user, if they are wrong we fail in validation.
This PR adds trailing unit tests to check the service cluster IP range and
improves the code coverage of k8s.io/kubernetes/cmd/kube-apiserver/app from
5.7% to 6.2%.
1) Dual stack IPv4/IPv6
2) Invalid IPv4, IPv6 mask
3) missing IPv4, IPv6 mask
4) invalid IP address format
The tests 2, 3, 4 are suggsted by Antonio Ojea.
Currently the "generate-csr" command does not have any output.
Pass an io.Writer (bound to os.Stdout from /cmd) to the functions
responsible for generating the kubeconfig / certs keys and CSRs.
If nil is passed these functions don't output anything.
Discussion is ongoing about how to best handle dual-stack with clouds
and autodetected IPs, but there is at least agreement that people on
bare metal ought to be able to specify two explicit IPs on dual-stack
hosts, so allow that.
Deprecate the experimental command "alpha self-hosting" and its
sub-command "pivot" that can be used to create a self-hosting
control-plane from static Pods.
The kubeconfig phase of "kubeadm init" detects external CA mode
and skips the generation of kubeconfig files. The kubeconfig
handling during control-plane join executes
CreateJoinControlPlaneKubeConfigFiles() which requires the presence
of ca.key when preparing the spec of a kubeconfig file and prevents
usage of external CA mode.
Modify CreateJoinControlPlaneKubeConfigFiles() to skip generating
the kubeconfig files if external CA mode is detected.
- Modify validateCACertAndKey() to print warnings for missing keys
instead of erroring out.
- Update unit tests.
This allows doing a CP node join in a case where the user has:
- copied shared certificates to the new CP node, but not copied
ca.key files, treating the cluster CAs as external
- signed other required certificates in advance
The provided DialContext wraps existing clients' DialContext in an attempt to
preserve any existing timeout configuration. In some cases, we may replace
infinite timeouts with golang defaults.
- scaleio: tcp connect/keepalive values changed from 0/15 to 30/30
- storageos: no change
The flag was deprecated as it is problematic since it allows
overrides of the kubelet configuration that is downloaded
from the cluster during upgrade.
Kubeadm node upgrades already download the KubeletConfiguration
and store it in the internal ClusterConfiguration type. It is then
only a matter of writing that KubeletConfiguration to disk.
For external CA users that have prepared the kubeconfig files
for components, they might wish to provide a custom API server URL.
When performing validation on these kubeconfig files, instead of
erroring out on such custom URLs, show a klog Warning.
This allows flexibility around topology setup, where users
wish to make the kubeconfigs point to the ControlPlaneEndpoint instead
of the LocalAPIEndpoint.
Fix validation in ValidateKubeconfigsForExternalCA expecting
all kubeconfig files to use the CPE. The kube-scheduler and
kube-controller-manager now use LAE.
This PR specifies minimum control plane version,
kubelet version and current K8s version for v1.20.
Signed-off-by: Kommireddy Akhilesh <akhileshkommireddy2412@gmail.com>
Client side period validation of certificates should not be
fatal, as local clock skews are not so uncommon. The validation
should be left to the running servers.
- Remove this validation from TryLoadCertFromDisk().
- Add a new function ValidateCertPeriod(), that can be used for this
purpose on demand.
- In phases/certs add a new function CheckCertificatePeriodValidity()
that will print warnings if a certificate does not pass period
validation, and caches certificates that were already checked.
- Use the function in a number of places where certificates
are loaded from disk.
CNI is no longer alpha and is widely used by almost every Kubernetes cluster, we should remove the alpha warnings that were originally added from the early days of CNI
Signed-off-by: Andrew Sy Kim <kim.andrewsy@gmail.com>
The isCoreDNSVersionSupported() check assumes that
there is a running kubelet, that manages the CoreDNS containers.
If the containers are being created it is not possible to fetch
their image digest. To workaround that, a poll can be used in
isCoreDNSVersionSupported() and wait for the CoreDNS Pods
are expected to be running. Depending on timing and CNI
yet to be installed this can cause problems related to
addon idempotency of "kubeadm init", because if the CoreDNS
Pods are waiting for another step they will never get running.
Remove the function isCoreDNSVersionSupported() and assume that
the version is always supported. Rely on the Corefile migration
library to error out if it must.
- Ensure the directory is created with 0700 via a new function
called CreateDataDirectory().
- Call this function in the init phases instead of the manual call
to MkdirAll.
- Call this function when joining control-plane nodes with local etcd.
If the directory creation is left to the kubelet via the
static Pod hostPath mounts, it will end up with 0755
which is not desired.
A bug was discovered in the `enforceRequirements` func for `upgrade plan`.
If a command line argument that specifies the target Kubernetes version is
supplied, the returned `ClusterConfiguration` by `enforceRequirements` will
have its `KubernetesVersion` field set to the new version.
If no version was specified, the returned `KubernetesVersion` points to the
currently installed one.
This remained undetected for a couple of reasons
- It's only `upgrade plan` that allows for the version command line argument to
be optional (in `upgrade plan` it's mandatory)
- Prior to 1.19, the implementation of `upgrade plan` did not make use of the
`KubernetesVersion` returned by `enforceRequirements`.
`upgrade plan` supports this optional command line argument to enable
air-gapped setups (as not specifying a version on the command line will end up
looking for the latest version over the Interned).
Hence, the only option is to make `enforceRequirements` consistent in the
`upgrade plan` case and always return the currently installed version in the
`KubernetesVersion` field.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Pinning the kube-controller-manager and kube-scheduler kubeconfig files
to point to the control-plane-endpoint can be problematic during
immutable upgrades if one of these components ends up contacting an N-1
kube-apiserver:
https://kubernetes.io/docs/setup/release/version-skew-policy/#kube-controller-manager-kube-scheduler-and-cloud-controller-manager
For example, the components can send a request for a non-existing API
version.
Instead of using the CPE for these components, use the LocalAPIEndpoint.
This guarantees that the components would talk to the local
kube-apiserver, which should be the same version, unless the user
explicitly patched manifests.
A check that verifies that kubeadm does not "upgrade" to an older release was
overly optimized by skipping upgrade if the new version is the same as the old
one. This somewhat makes sense, but that way changes in any of the etcd fields
in the ClusterConfiguration won't be applied if the etcd version is not
changed.
Hence, this simple change ensures that the upgrade is done even when no version
change takes place.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
As a part of cleaning up inactive members (who haven't been active since
beginning of 2019) from OWNERS files, this commit removes euank and
CaoShuFeng from the list of reviewers.
dependencycheck verifies that no violating depdendencies exist in pkgs
passed via a JSON file generated from go list.
Signed-off-by: hasheddan <georgedanielmangum@gmail.com>
The implementation consists of
- identifying all places where VolumeSource.PersistentVolumeClaim has
a special meaning and then ensuring that the same code path is taken
for an ephemeral volume, with the ownership check
- adding a controller that produces the PVCs for each embedded
VolumeSource.EphemeralVolume
- relaxing the PVC protection controller such that it removes
the finalizer already before the pod is deleted (only
if the GenericEphemeralVolume feature is enabled): this is
needed to break a cycle where foreground deletion of the pod
blocks on removing the PVC, which waits for deletion of the pod
The controller was derived from the endpointslices controller.
* Creates private keys and CSR files for all the control-plane certificates
* Helps with External CA mode of kubeadm
Signed-off-by: Richard Wall <richard.wall@jetstack.io>
Back in the v1alpha2 days the fuzzer test needed to be disabled. To ensure that
there were no config breaks and everything worked correctly extensive replacement
tests were put in place that functioned as unit tests for the kubeadm config utils
as well.
The fuzzer test has been reenabled for a long time now and there's no need for
these replacements. Hence, over time most of these were disabled, deleted and
refactored. The last remnants are part of the LoadJoinConfigurationFromFile test.
The test data for those old tests remains largely unused today, but it still receives
updates as it contains kubelet's and kube-proxy's component configs. Updates to these
configs are usually done because the maintainers of those need to add a new field.
Hence, to cleanup old code and reduce maintenance burden, the last test that depends
on this test data is finally refactored and cleaned up to represent a simple unit test
of `LoadJoinConfigurationFromFile`.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Over the course of recent development of the `componentconfigs` package,
it became evident that most of the tests in this package cannot be implemented without
using a component config. As all of the currently supported component configs are
external to the kubeadm project (kubelet and kube-proxy), practically all of the tests
in this package are now dependent on external code.
This is not desirable, because other component's configs may change frequently and
without much of a notice. In particular many configs add new fields without bumping their
versions. In addition to that, some components may be deprecated in the future and many
tests may use their configs as a place holder of a component config just to test some
common functionality.
To top that, there are many tests that test the same common functionality several times
(for each different component config).
Thus a kubeadm managed replacement and a fake test environment are introduced.
The new test environment uses kubeadm's very own `ClusterConfiguration`.
ClusterConfiguration is normally not managed by the `componentconfigs` package.
It's only used, because of the following:
- It's a versioned API that is under the control of kubeadm maintainers. This enables us to test
the componentconfigs package more thoroughly without having to have full and always up to date
knowledge about the config of another component.
- Other components often introduce new fields in their configs without bumping up the config version.
This, often times, requires that the PR that introduces such new fields to touch kubeadm test code.
Doing so, requires more work on the part of developers and reviewers. When kubeadm moves out of k/k
this would allow for more sporadic breaks in kubeadm tests as PRs that merge in k/k and introduce
new fields won't be able to fix the tests in kubeadm.
- If we implement tests for all common functionality using the config of another component and it gets
deprecated and/or we stop supporting it in production, we'll have to focus on a massive test refactoring
or just continue importing this config just for test use.
Thus, to reduce maintenance costs without sacrificing test coverage, we introduce this mini-framework
and set of tests here which replace the normal component configs with a single one (`ClusterConfiguration`)
and test the component config independent logic of this package.
As a result of this, many of the older test cases are refactored and greatly simplified to reflect
on the new change as well. The old tests that are strictly tied to specific component configs
(like the defaulting tests) are left unchanged.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Kubeadm setup of kube-controller-manager and kube-scheduler is
lacking the --port=0 option which caused the component to enable
the insecure port by default and serve insecurely on the default
node interface.
Add --port=0 by default to both components. Users are still allowed
the explicitly set the flag (via extraArgs), which allows them
to override this default kubeadm behavior and enable the insecure port.
NOTE: the flag is deprecated and should be removed from kubeadm manifests
once it's removed from core.
`kubeadm config upload` is a GA command that has been deprecated and scheduled
for removal since Kubernetes 1.15 (released 06/19/2019). This change will
finally removed it in Kubernetes 1.19 (planned for August 2020).
The original command has long since been replaced by a GA init phase:
`kubeadm init phase upload-config`
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Add PatchStaticPod() in staticpod/utils.go
Apply patches to static Pods in:
- phases/controlplane/CreateStaticPodFiles()
- phases/etcd/CreateLocalEtcdStaticPodManifestFile() and
CreateStackedEtcdStaticPodManifestFile()
Add unit tests and update Bazel.
Previously, separate interfaces were defined for Reserve and Unreserve
plugins. However, in nearly all cases, a plugin that allocates a
resource using Reserve will likely want to register itself for Unreserve
as well in order to free the allocated resource at the end of a failed
scheduling/binding cycle. Having separate plugins for Reserve and
Unreserve also adds unnecessary config toil. To that end, this patch
aims to merge the two plugins into a single interface called a
ReservePlugin that requires implementing both the Reserve and Unreserve
methods.
This change enables kubeadm upgrade plan to print a state table with
information regarding known component config API groups. Most importantly this
information includes current and preferred version for each group and an
indication if a manual user upgrade is required.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
`kubeadm upgrade plan` is using the external (currently `v1alpha1`) types of
the kubeadm output API to collect upgrade plans. This is counter intuitive
since code structure gets bound to the whatever version the output API is at.
In addition to that, the versioned API is used only in the very last stages of
a machine readable output (which is currently not implemented).
Hence, to increase flexibility and keep up with the standard Kubernetes
ecosystem practice, `kubeadm upgrade plan` is migrated to use the internal
types of the output API.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
UploadConfiguration() now always retries the underling API calls,
which can make TestUploadConfiguration run for a long time.
Remove the negative test cases, where errors are expected.
Negative test cases should be tested in app/util/apiclient,
where a short timeout / retry count should be possible for unit tests.
Currently, kubeadm would refuse to perfom an upgrade (or even planing for one)
if it detects a user supplied unsupported component config version. Hence,
users are required to manually upgrade their component configs and store them
in the config maps prior to executing `kubeadm upgrade plan` or
`kubeadm upgrade apply`.
This change introduces the ability to use the `--config` option of the
`kubeadm upgrade plan` and `kubeadm upgrade apply` commands to supply a YAML
file containing component configs to be used in place of the existing ones in
the cluster upon upgrade.
The old behavior where `--config` is used to reconfigure a cluster is still
supported. kubeadm automatically detects which behavior to use based on the
presence (or absense) of kubeadm config types (API group
`kubeadm.kubernetes.io`).
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
We were detecting the IP family that kube-proxy should use
based on the bind address, however, this is not valid when
using an unspecified address, because on those cases
kube-proxy adopts the IP family of the address reported
in the Node API object.
The IP family will be determined by the nodeIP used by the proxier
The order of precedence is:
1. config.bindAddress if bindAddress is not 0.0.0.0 or ::
2. the primary IP from the Node object, if set
3. if no IP is found it defaults to 127.0.0.1 and IPv4
Signed-off-by: Antonio Ojea <antonio.ojea.garcia@gmail.com>
If an etcd member with the same address already exists, don't re-add it.
Instead, use the existing member list for creating the "initial cluster"
that is written for this etcd server instance static Pod.
Component configs are used by kubeadm upgrade plan at the moment. However, they
can prevent kubeadm upgrade plan from functioning if loading of an unsupported
version of a component config is attempted. For that matter it's best to just
stop loading component configs as part of the kubeadm config load process.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Under the feature gate DefaultPodTopologySpread, which will disable the legacy DefaultPodTopologySpread plugin from the default algorithm providers.
Signed-off-by: Aldo Culquicondor <acondor@google.com>
`getK8sVersionFromUserInput` would attempt to load the config from a user
specified YAML file (via the `--config` option of `kubeadm upgrade plan` or
`kubeadm upgrade apply`). This is done in order to fetch the `KubernetesVersion`
field of the `ClusterConfiguration`. The complete config is then immediately
discarded. The actual config that is used during the upgrade process is fetched
from within `enforceRequirements`.
This, along with the fact that `getK8sVersionFromUserInput` is always called
immediately after `enforceRequirements` makes it possible to merge the two.
Merging them would help us simplify things and avoid future problems in
component config related patches.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Until now, users were always asked to manually convert a component config to a
version supported by kubeadm, if kubeadm is not supporting its version.
This is true even for configs generated with older kubeadm versions, hence
getting users to make manual conversions on kubeadm generated configs.
This is not appropriate and user friendly, although, it tends to be the most
common case. Hence, we sign kubeadm generated component configs stored in
config maps with a SHA256 checksum. If a configs is loaded by kubeadm from a
config map and has a valid signature it's considered "kubeadm generated" and if
a version migration is required, this config is automatically discarded and a
new one is generated.
If there is no checksum or the checksum is not matching, the config is
considered as "user supplied" and, if a version migration is required, kubeadm
will bail out with an error, requiring manual config migration (as it's today).
The behavior when supplying component configs on the kubeadm command line
does not change. Kubeadm would still bail out with an error requiring migration
if it can recognize their groups but not versions.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
In case a malformed flag is passed to k8s components
such as "–foo", where "–" is not an ASCII dash character,
the components currently silently ignore the flag
and treat it as a positional argument.
Make k8s components/commands exit with an error if a positional argument
that is not empty is found. Include a custom error message for all
components except kubeadm, as cobra.NoArgs is used in a lot of
places already (can be fixed in a followup).
The kubelet already handles this properly - e.g.:
'unknown command: "–foo"'
This change affects:
- cloud-controller-manager
- kube-apiserver
- kube-controller-manager
- kube-proxy
- kubeadm {alpha|config|token|version}
- kubemark
Signed-off-by: Monis Khan <mok@vmware.com>
Signed-off-by: Lubomir I. Ivanov <lubomirivanov@vmware.com>
- Use a dummy nodename instead of OS hostname
- Inline toString() function
- Use backticks to wrap expected patch
- Remove redundant test name from error logs
kubelet.DownloadConfig is an old utility function which takes a client set and
a kubelet version, uses them to fetch the kubelet component config from a
config map, and places it in a local file. This function is simple to use, but
it is dangerous and unnecessary. Practically, in all cases the kubelet
configuration is present locally and does not need to be fetched from a config
map on the cluster (it just needs to be stored in a file).
Furthermore, kubelet.DownloadConfig does not use the kubeadm component configs
module in any way. Hence, a kubelet configuration fetched using it may not be
patched, validated, or otherwise, processed in any way by kubeadm other than
piping it to a file.
This patch replaces all but a single kubelet.DownloadConfig invocation with
equivalents that get the local copy of the kubelet component config and just
store it in a file. The sole remaining invocation covers the
`kubeadm upgrade node --kubelet-version` case.
In addition to that, a possible panic is fixed in kubelet.DownloadConfig and
it now takes the kubelet version parameter as string.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
Current calculations assume that -trimpath is not passed to go tool
compile, which is not the case for test binaries built with bazel. This
causes issues for integration tests right now but is generally not
correct.
The approach taken here is a bit of a hack but it works on the
assumption that if and only if trimpath is passed, we are running under
bazel. I didn't see a good spot for pkgPath(), so I just copied it
around.
We can remove all uses of `dockershim` from `cmd/kubelet`, by just
passing the docker options to the kubelet in their pure form, instead of
using them to create a `dockerClientConfig` (which is defined in
dockershim). We can then construct the `dockerClientConfig` only when we
actually need it.
Remove one of two uses of Dockershim in `cmd/kubelet`. The other is for
creating a docker client which we pass to the Kubelet... we will handle
that refactor in a separate diff.
I'm fairly confident, though need to double check, that no one is
actually using this experimental dockershim behavior. If they are, I
think we will want to find a new way to support it (that doesn't require
using the Kubelet only to launch Dockershim).
kubeadm is setting the IPv6DualStack feature gate in the command line of the kubelet.
However, the kubelet is gradually moving away from command line flags towards component config use.
Hence, we should set the IPv6DualStack feature gate in the component config instead.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
In slower setups it can take more time for the existing cluster
to be in a healthy state, so the existing backoff of ~50 seconds
is apparently not sufficient.
The client dial can also fail for similar reasons.
Improve kubeadm's join toleration of adding new etcd members.
Wrap both the client dial and member add in a longer backoff
(up to ~200 seconds).
This particular change should be backported to the support skew.
In a future change for master, all etcd client operations should be
make consistent so that the etcd logic is in a sane state.
specifically:
- cmd/kubeadm/.import-restrictions
- we don't need to explicitly allow k8s.io repos (external or published)
- rm pkg/controller/.import-restrictions
- pkg/client/unversioned was removed in 59042
- pkg/kubectl/.import-restrictions
- pkg/printers is no longer used
- pkg/api was masking all of the pkg/apis prefixes
- rm staging/src/k8s.io/code-generator/cmd/lister-gen/.import-restrictions
- noop / empty file
- test/e2e/framework/.import-restrictions
- we don't need to explicitly allow k8s.io repos (external or published)
yaml has comments, so we can explain why we have certain rules or
certain prefixes
for those files that weren't already commented yaml, I converted them to
yaml and took a best guess at comments based on the PRs that introduced
or updated them
Use an init container that performs the pre-pull of a component
and then start an instance of "pause" as a regular container to
get the DaemonSet Pod in a Running state.
More details on this change in the code comments.
pkg/util/rlimit/rlimit_linux.go:25:1: exported function RlimitNumFiles should have comment or be unexported
pkg/util/rlimit/rlimit_linux.go:25:6: func name will be used as rlimit.RlimitNumFiles by other packages, and that stutters; consider calling this NumFiles
pkg/util/rlimit/rlimit_unsupported.go:25:1: exported function RlimitNumFiles should have comment or be unexported
pkg/util/rlimit/rlimit_unsupported.go:25:6: func name will be used as rlimit.RlimitNumFiles by other packages, and that stutters; consider calling this NumFiles
Ref: https://github.com/kubernetes/kubernetes/issues/68026
The flag "--use-api" for "alpha certs renew" was deprecated in 1.18.
Remove the flag and related logic that executes certificate renewal
using "api/certificates/v1beta1". kubeadm continues to be able
to create CSR files and renew using the local CA on disk.
kubeadm init prints:
W0410 23:02:10.119723 13040 manifests.go:225] the default kube-apiserver
authorization-mode is "Node,RBAC"; using "Node,RBAC"
Add a new function compareAuthzModes() and a unit test for it.
Make sure the warning is printed only if the user modes don't match
the defaults.
Allow overriding the dry-run temporary directory with
an env. variable (KUBEADM_INIT_DRYRUN_DIR).
Use the same variable in test/cmd/init_test.go.
This allows running integration tests as non-root.
Make getKubeadmPath() fetch the KUBEADM_PATH env. variable.
Panic if it's missing. Don't handle the "--kubeadm-path"
flag. Remove the same flag from the BUILD bazel test rule.
Don't handle "--kubeadm-cmd-skip" usage of this flag is missing
from the code base.
Remove usage of "kubeadmCmdSkip" as the flag "--kubeadm-cmd-skip"
is never passed.
If the kube-proxy/dns ConfigMap are missing, show warnings and assume
that these addons were skipped during "kubeadm init",
and that their redeployment on upgrade is not desired.
TODO: remove this once "kubeadm upgrade apply" phases are supported:
https://github.com/kubernetes/kubeadm/issues/1318
The TLS bootstrapping timeout is increased to 5 minutes with a retry
once every 5 seconds. Failing fast if the kubelet is not healthy is also
preserved.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
kubeadm uses image tags (such as `v3.4.3-0`) to specify the version of
etcd. However, the upgrade code in kubeadm uses the etcd client API to
fetch the currently deployed version. The result contains only the etcd
version without the additional information (such as image revision) that
is normally found in the tag. As a result it would refuse an upgrade
where the etcd versions match and the only difference is the image
revision number (`v3.4.3-0` to `v3.4.3-1`).
To fix the above issue, the following changes are done:
- Replace the existing etcd version querying code, that uses the etcd
client library, with code that returns the etcd image tag from the
local static pod manifest file.
- If an etcd `imageTag` is specified in the ClusterConfiguration during
upgrade, use that tag instead. This is done regardless if the tag was
specified in the configuration stored in the cluster or with a new
configuration supplied by the `--config` command line parameter.
If no custom tag is specified, kubeadm will select one depending on
the desired Kubernetes version.
- `kubeadm upgrade plan` no longer prints upgrade information about
external etcd. It's the user's responsibility to manage it in that
case.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
If the user does not provide --config or --control-plane
but provides some other flags such as --certificate-key
kubeadm is supposed to print a warning.
The logic around printing the warning is bogus. Implement
proper checks of when to print the warning.
b117a928 added a new check during "join" whether a Node with
the same name exists in the cluster.
When upgrading from 1.17 to 1.18 make sure the required RBAC
by this check is added. Otherwise "kubeadm join" will complain that
it lacks permissions to GET a Node.
A narrow assumption of what is contained in the `imageID` fields for the
CoreDNS pods causes a panic upon upgrade.
Fix this by using a proper regex to match a trailing SHA256 image digest
in `imageID` or return an error if it cannot find it.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
So multiple instances of kube-apiserver can bind on the same address and
port, to provide seamless upgrades.
Signed-off-by: Mateusz Gozdek <mateusz@kinvolk.io>
This builds on previous work but only sets the sysctlConnReuse value
if the kernel is known to be above 4.19. To avoid calling GetKernelVersion
twice, I store the value from the CanUseIPVS method and then check the version
constraint at time of expected sysctl call.
Signed-off-by: Christopher M. Luciano <cmluciano@us.ibm.com>
This change removes support for basic authn in v1.19 via the
--basic-auth-file flag. This functionality was deprecated in v1.16
in response to ATR-K8S-002: Non-constant time password comparison.
Similar functionality is available via the --token-auth-file flag
for development purposes.
Signed-off-by: Monis Khan <mok@vmware.com>
Most of these could have been refactored automatically but it wouldn't
have been uglier. The unsophisticated tooling left lots of unnecessary
struct -> pointer -> struct transitions.
The KCM is moving to means of only singing apiserver (kubelet) client
and kubelet serving certificates. See:
https://github.com/kubernetes/enhancements/blob/master/keps/sig-auth/20190607-certificates-api.md#signers
Up until now the experimental kubeadm functionality '--use-api'
under "kubeadm alpha certs renew" was using the KCM to sign *any*
certficate as long as the KCM has the root CA cert/key.
Post discussions with the kubeadm maintainers, it was decided that
this functionality should be removed from kubeadm due to the
requirement to have external signers for renewing the common
control-plane certificates that kubeadm manages.
- Uses correct pause image for Windows
- Omits systemd specific flags
- Common build flags function to be used by Linux and Windows
- Uses user configured image repository for Windows pause image
The old flag name doesn't make sense with the renamed API Priority and
Fairness feature, and it's still safe to change the flag since it hasn't done
anything useful in a released k8s version yet.
After the shift for init phases, GetStaticPodSpecs() from
app/phases/controlplane/manifests.go gets called on each control-plane
component sub-phase. This ends up calling the Printf from
AddExtraHostPathMounts() in app/phases/controlplane/volumes.go
multiple times printing the same volumes for different components.
- Remove the Printf call from AddExtraHostPathMounts().
- Print all volumes for a component in CreateStaticPodFiles() using klog
V(2).
Perhaps in the future a bigger refactor is needed here were a
single control-plane component spec can be requested instead of a
map[string]v1.Pod.
The selected key type is defined by kubeadm's --feature-gates option:
if it contains PublicKeysECDSA=true then ECDSA keys will be generated
and used.
By default RSA keys are used still.
Signed-off-by: Dmitry Rozhkov <dmitry.rozhkov@linux.intel.com>
When doing the very first upgrade from a cluster that contains the
source of truth in the ClusterStatus struct, the new kubeadm logic
will try to retrieve this information from annotations.
This changeset adds to both etcd and apiserver endpoint retrieval the
special case in which they won't retry if we are in such cases. The
logic will retry if we find any unknown error, but will not retry in
the following cases:
- etcd annotations do not contain etcd endpoints, but the overall list
of etcd pods is greater than 0. This means that we listed at least
one etcd pod, but they are missing the annotation.
- API server annotation is not found on the api server pod for a given
node name, but no errors aside from that one were found. This means
that the API server pod is present, but is missing the annotation.
In both cases there is no point in retrying, and so, this speeds up the
upgrade path when coming from a previous existing cluster.
While `ClusterStatus` will be maintained and uploaded, it won't be
used by the internal `kubeadm` logic in order to determine the etcd
endpoints anymore.
The only exception is during the first upgrade cycle (`kubeadm upgrade
apply`, `kubeadm upgrade node`), in which we will fallback to the
ClusterStatus to let the upgrade path add the required annotations to
the newly created static pods.
While kubeadm does not support CA rotation,
the users might still attempt to perform this manually.
For kubeconfig files, updating to a new CA is not reflected
and users need to embed new CA PEM manually.
On kubeconfig cert renewal, always keep the embedded CA
in sync with the one on disk.
Includes a couple of typo fixes.
- Add handlers for service account issuer metadata.
- Add option to manually override JWKS URI.
- Add unit and integration tests.
- Add a separate ServiceAccountIssuerDiscovery feature gate.
Additional notes:
- If not explicitly overridden, the JWKS URI will be based on
the API server's external address and port.
- The metadata server is configured with the validating key set rather
than the signing key set. This allows for key rotation because tokens
can still be validated by the keys exposed in the JWKs URL, even if the
signing key has been rotated (note this may still be a short window if
tokens have short lifetimes).
- The trust model of OIDC discovery requires that the relying party
fetch the issuer metadata via HTTPS; the trust of the issuer metadata
comes from the server presenting a TLS certificate with a trust chain
back to the from the relying party's root(s) of trust. For tests, we use
a local issuer (https://kubernetes.default.svc) for the certificate
so that workloads within the cluster can authenticate it when fetching
OIDC metadata. An API server cannot validly claim https://kubernetes.io,
but within the cluster, it is the authority for kubernetes.default.svc,
according to the in-cluster config.
Co-authored-by: Michael Taufen <mtaufen@google.com>
It turns out that the dual-stack feature enabled doesn't mean that
the cluster MUST be dual-stack, it only indicates that it MAY be
dual-stack but CAN be single-stack.
We should relax the validation to allow single-stack clusters
with dual-stack enabled.
If a Node name in the cluster is already taken and this Node is Ready,
prevent TLS bootsrap on "kubeadm join" and exit early.
This change requires that a new ClusterRole is granted to the
"system:bootstrappers:kubeadm:default-node-token" group to be
able get Nodes in the cluster. The same group already has access
to obtain objects such as the KubeletConfiguration and kubeadm's
ClusterConfiguration.
The motivation of this change is to prevent undefined behavior
and the potential control-plane breakdown if such a cluster
is racing to have two nodes with the same name for long periods
of time.
The following values are validated in the following precedence
from lower to higher:
- actual hostname
- NodeRegistration.Name (or "--node-name") from JoinConfiguration
- "--hostname-override" passed via kubeletExtraArgs
If the user decides to not let kubeadm know about a custom node name
and to instead override the hostname from a kubelet systemd unit file,
kubeadm will not be able to detect the problem.
- Extend the exponential backoff for add/remove/... retry to
11 steps ~=106 seconds. From experiments for 3 and more members
the race can take more that ~=26 seconds.
- Increase the dialTimeout for client creation to 40 seconds.
20 seconds seems racy for 3 and more members.
For the etcd client, amend AddMember() to handle a very
rare bug when multiple members can end up with the same
name. Match the member peer address and assign it the name of
the member we are adding. For the rest of the members with missing
names use their member IDs as name. The etcd node is not disrupted
by the unknown names.
The important aspects are:
- The number of members of the initial cluster must match
the members in the cluster.
- The member we are current adding is present in the initial cluster.
As part of #68522, Switching off the cAdvisor v1 Json API that we expose
directly. These include /stats/, /stats/container, /stats/{podName}/{containerName},
and /stats/{namespace}/{podName}/{uid}/{containerName}
The CoreDNS GA feature-gate in kubeadm was deprecated since 1.13.
The k8s policy is to remove the gate 2 releases after it transitions
to GA:
https://kubernetes.io/docs/reference/using-api/deprecation-policy/#deprecation
We kept it around for longer to prevent existing setups from breaking
as it caused minimal maintenance overhead.
This creates a new EndpointSliceProxying feature gate to cover EndpointSlice
consumption (kube-proxy) and allow the existing EndpointSlice feature gate to
focus on EndpointSlice production only. Along with that addition, this enables
the EndpointSlice feature gate by default, now only affecting the controller.
The rationale here is that it's really difficult to guarantee all EndpointSlices
are created in a cluster upgrade process before kube-proxy attempts to consume
them. Although masters are generally upgraded before nodes, and in most cases,
the controller would have enough time to create EndpointSlices before a new node
with kube-proxy spun up, there are plenty of edge cases where that might not be
the case. The primary limitation on EndpointSlice creation is the API rate limit
of 20QPS. In clusters with a lot of endpoints and/or with a lot of other API
requests, it could be difficult to create all the EndpointSlices before a new
node with kube-proxy targeting EndpointSlices spun up.
Separating this into 2 feature gates allows for a more gradual rollout with the
EndpointSlice controller being enabled by default in 1.18, and EndpointSlices
for kube-proxy being enabled by default in the next release.
The warning message
```
[config] WARNING: Ignored YAML document with GroupVersionKind ...
```
is printed for all GVKs that are not part of the kubeadm core types.
This is wrong as the component config types are supported and successfully
parsed and used despite the fact that the warning is printed for them too.
Hence this simple fix first checks if the group of the GVK is a supported
component config group and the warning is printed only if it's not.
Signed-off-by: Rostislav M. Georgiev <rostislavg@vmware.com>
kubeadm deploys the apiserver, controller-manager and the scheduler
using liveness probes.
The bind-address option is used to configure the probe address, in
case this is configured with an unspecified address, the probe
will fail. When using an unspecified address the probe host field is
left empty, otherwise the bind-address is used.
The function validateKubeConfig() can end up comparing
a user generated kubeconfig to a kubeconfig generated by kubeadm.
If a user kubeconfig has a CA that is base64 encoded with whitespace,
if said kubeconfig is loaded using clientcmd.LoadFromFile()
the CertificateAuthorityData bytes will be decoded from base64
and placed in the v1.Config raw. On the other hand a kubeconfig
generated by kubeadm will have the ca.crt parsed to a Certificate
object with whitespace ignored in the PEM input.
Make sure that validateKubeConfig() tolerates whitespace differences
when comparing CertificateAuthorityData.
kubeadm removed the deprecated "--address" flag for controller-manager
and scheduler in favor of "--bind-address"
We should use bind-address to configure the manifest probe addresses.
If the user has modified the kubelet.conf post TLS bootstrap
to become invalid, the function getNodeNameFromKubeletConfig() can
panic. This was observed to trigger in "kubeadm reset" use cases.
Add basic validation and unit tests around parsing the kubelet.conf
with the aforementioned function.
Currently if the controlplane fails to init, we print out a message
with some example commands that only show docker CLI.
This tries to improve that by printing the example commands for
docker, cri-o and containerd by checking the socket looking for
the default docker socket.
switch api helper functions to v1 CRD api
switch v1 CRD for apiserver internal
switch to v1 CRD for internal controllers
api storage/validation related changes
move local-defaulting utils private to prevent spreading
boilerplate
keep the subresource status/scale spec nil unless it's enabled
clean up empty space
CreateOrMutateConfigMap was not resilient when it was trying to Create
the ConfigMap. If this operation returned an unknown error the whole
operation would fail, because it was strict in what error it was
expecting right afterwards: if the error returned by the Create call
was a IsAlreadyExists error, it would work fine. However, if an
unexpected error (such as an EOF) happened, this call would fail.
We are seeing this error specially when running control plane node
joins in an automated fashion, where things happen at a relatively
high speed pace.
It was specially easy to reproduce with kind, with several control
plane instances. E.g.:
```
[upload-config] Storing the configuration used in ConfigMap "kubeadm-config" in the "kube-system" Namespace
I1130 11:43:42.788952 887 round_trippers.go:443] POST https://172.17.0.2:6443/api/v1/namespaces/kube-system/configmaps?timeout=10s in 1013 milliseconds
Post https://172.17.0.2:6443/api/v1/namespaces/kube-system/configmaps?timeout=10s: unexpected EOF
unable to create ConfigMap
k8s.io/kubernetes/cmd/kubeadm/app/util/apiclient.CreateOrMutateConfigMap
/go/src/k8s.io/kubernetes/_output/dockerized/go/src/k8s.io/kubernetes/cmd/kubeadm/app/util/apiclient/idempotency.go:65
```
This change makes this logic more resilient to unknown errors. It will
retry on the light of unknown errors until some of the expected error
happens: either `IsAlreadyExists`, in which case we will mutate the
ConfigMap, or no error, in which case the ConfigMap has been created.
kubeadm always use the IPv4 localhost address by defaultA for etcd
The probe hostname is obtained before the generation of the etcd
parameters, so it can't detect the right IP familiy for the
host of the probe.
This causes that with IPv6 clusters doesn't work because the probe
uses the IPv4 localhost address.
This patchs configures the right localhost address based on the used
AdvertiseAddress IP family.
- Use `schema.TypeMeta` instead of custom `struct` for VK
- More strict check on GVK after `Interpret` in `SplitYAMLDocuments`
- Adjust `Interpret` comment to include JSON
- Add retrieveValidatedConfigInfo to be able to better unit
test the function.
- Break some of the logic in RetrieveValidatedConfigInfo into
helper functions.
- Pass JoinConfiguration.Discovery to RetrieveValidatedConfigInfo
instead of JoinConfiguration.
- Use the discovery timeout per API call to fetch cluster-info
(optionally the user value can be slit in 2).
- Add detailed unit tests for retrieveValidatedConfigInfo.
