* Add write-kubeconfig-group flag to server
* update kubectl unable to read config message for kubeconfig mode/group
Signed-off-by: Katherine Pata <me@kitty.sh>
(cherry picked from commit 7a0ea3c953)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Refactor agent supervisor listener startup and authn/authz to use upstream
auth delegators to perform for SubjectAccessReview for access to
metrics.
* Convert spegel and pprof handlers over to new structure.
* Promote bind-address to agent flag to allow setting supervisor bind
address for both agent and server.
* Promote enable-pprof to agent flag to allow profiling agents. Access
to the pprof endpoint now requires client cert auth, similar to the
spegel registry api endpoint.
* Add prometheus metrics handler.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit ff679fb3ab)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Disable cleanup of orphaned snapshots and patching of node annotations if running agentless
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit edb0440017)
Moving it into config.Agent so that we can use or modify it outside the context of containerd setup
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit 16d29398ad)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Render cri registry mirrors.x.endpoints and configs.x.tls into config_path; keep
using mirrors.x.rewrites and configs.x.auth those do not yet have an
equivalent in the new format.
The new config file format allows disabling containerd's fallback to the
default endpoint when using mirror endpoints; a new CLI flag is added to
control that behavior.
This also re-shares some code that was unnecessarily split into parallel
implementations for linux/windows versions. There is probably more work
to be done on this front but it's a good start.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit c45524e662)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Remove KubeletCredentialProviders and JobTrackingWithFinalizers feature-gates, both of which are GA and cannot be disabled.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit 231cb6ed20)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Create a generic helper function that finds extra containerd runtimes.
The code was originally inside of the nvidia container discovery file.
Signed-off-by: Flavio Castelli <fcastelli@suse.com>
Discover the containerd shims based on runwasi that are already
available on the node.
The runtimes could have been installed either by a package manager or by
the kwasm operator.
Signed-off-by: Flavio Castelli <fcastelli@suse.com>
The containerd configuration on a Linux system now handles the nvidia
and the WebAssembly runtimes.
Signed-off-by: Flavio Castelli <fcastelli@suse.com>
---------
Signed-off-by: Flavio Castelli <fcastelli@suse.com>
Added runtime classes for crun/wasm/nvidia
Signed-off-by: Vitor Savian <vitor.savian@suse.com>
Added default runtime flag
Signed-off-by: Vitor Savian <vitor.savian@suse.com>
Problem:
Configuring qos-class features in containerd requres a custom containerd configuration template.
Solution:
Look for configuration files in default locations and configure containerd to use them if they exist.
Signed-off-by: Oliver Larsson <larsson.e.oliver@gmail.com>
(cherry picked from commit 30c8ad926d)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Add --image-service-endpoint flag
Problem:
External container runtime can be set but image service endpoint is unchanged
and also is not exposed as a flag. This is useful for using containerd
snapshotters outside of the ones that have built-in support like
stargz-snapshotter.
Solution:
Add a flag --image-service-endpoint and also default image service endpoint to
container runtime endpoint if set.
Signed-off-by: Edgar Lee <edgarhinshunlee@gmail.com>
(cherry picked from commit fe18b1fce9)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Enable the feature-gate for both kubelet and cloud-controller-manager. Enabling it on only one side breaks RKE2, where feature-gates are not shared due to running in different processes.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Write the extra metadata both locally and to S3. These files are placed such that they will not be used by older versions of K3s that do not make use of them.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit 7464007037)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Consolidate NewCertCommands
* Add support for user defined new token
* Add E2E testlets
* Ensure agent token also changes
Signed-off-by: Derek Nola <derek.nola@suse.com>
* Update to v1.27.6
Signed-off-by: Pedro Tashima <pedro.tashima@suse.com>
* Bump containerd and stargz versions
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Print message on upgrade fail
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Send Bad Gateway instead of Service Unavailable when tunnel dial fails
Works around new handling for Service Unavailable by apiserver aggregation added in kubernetes/kubernetes#119870
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Add 60 seconds to server upgrade wait to account for delays in apiserver readiness
Also change cleanup helper to ensure upgrade test doesn't pollute the
images for the rest of the tests.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
---------
Signed-off-by: Pedro Tashima <pedro.tashima@suse.com>
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Co-authored-by: Pedro Tashima <pedro.tashima@suse.com>
Co-authored-by: Brad Davidson <brad.davidson@rancher.com>
Only configure enable-aggregator-routing and egress-selector-config-file
if required by egress-selector-mode.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Problem:
When support for etcd was added in 3957142, generation of certificates and keys for etcd was not gated behind use of managed etcd.
Keys are generated and distributed across servers even if managed etcd is not enabled.
Solution:
Allow generation of certificates and keys only if managed etc is enabled. Check config.DisableETCD flag.
Signed-off-by: Bartossh <lenartconsulting@gmail.com>
Allows nodes to join the cluster during a webhook outage. This also
enhances auditability by creating Kubernetes events for the deferred
verification.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Only actual admin actions should use the admin kubeconfig; everything done by the supervisor/deploy/helm controllers will now use a distinct account for audit purposes.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Add helper function for multiple arguments in stringslice
Signed-off-by: Derek Nola <derek.nola@suse.com>
* Cleanup server setup with util function
Signed-off-by: Derek Nola <derek.nola@suse.com>
Several places in the code used a 5-second retry loop to wait on
Runtime.Core to be set. This caused a race condition where OnChange
handlers could be added after the Wrangler shared informers were already
started. When this happened, the handlers were never called because the
shared informers they relied upon were not started.
Fix that by requiring anything that waits on Runtime.Core to run from a
cluster controller startup hook that is guaranteed to be called before
the shared informers are started, instead of just firing it off in a
goroutine that retries until it is set.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
We need to send the full chain in order for cross-signing to work
properly during switchover to a new root.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Turns out etcd-only nodes were never running **any** of the controllers,
so allowing multiple controllers didn't really fix things.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
