* Move all but mock provider out of tree
These have all been moved to repos under github.com/virtual-kubelet.
* Introduce a providers.Store
This essentially moves the the old register/ handling into a first class
object that can be controlled from the CLI rather than through build
tags deep in the code.
This actually would have made it a bit easier to build the provider
repos and makes the cmd/ code more re-usable.
We still use it internally, but this does not need to be part of the
public API. Instead just have callers pass us the relevent listers and
we create our own resource manager.
* Move tracing exporter registration
This doesn't belong in the library and should be configured by the
consumer of the opencensus package.
* Rename `vkublet` package to `node`
`vkubelet` does not convey any information to the consumers of the
package.
Really it would be nice to move this package to the root of the repo,
but then you wind up with... interesting... import semantics due to the
repo name... and after thinking about it some, a subpackage is really
not so bad as long as it has a name that convey's some information.
`node` was chosen since this package deals with all the semantics of
operating a node in Kubernetes.
Do not enable leases by default in the node constructor.
Simplify node constructor to not require a lease client when leases may
not even be enabled.
Updates node status update interval to the kubelet default of 10s (was
5s in vk).
If we don't handle the "pod not found" case then we end up with the pod
getting re-queued over and over until the max retries are hit. It also
blocks the queue for other pod status updates for that pod
namespace/name.
This changes the tracing package to accept an error on SetStatus, which
is really what we always want anyway.
This also decouples the trace package from opencensus.
This moves the logic for re-creating the a missing node up into the CLI.
We can make this optional, but for now I've just preserved existing
functionality.
Example:
env name: FOO
value: "foo"
name: BAR
value: "bar"
name: FOOBAR
value: "${FOO}${BAR}" <-- should expand to: "foobar"
Added testcase for the same as well.
Change is based on kubelet_pods.go.
Simplified some of the existing code.
* Use the VKUBELET API for fetching the pods during testing, which
makes it easier to run testing without wiring up an entire cluster
* Runningpods does not require each provider to implement the
optional stats interface for testing.
Don't use actual "mock" provider in vkubelet to reduce awkward import
cycles.
Really the "mock" provider is for e2e tests. Instead create our own mock
in the vkubelet package specifically for unit tests which can also be
more easily customized to test edge cases.
This makes the concept of a `Provider` wholely implemented in the cli
implementation in cmd/virtual-kubelet.
It allows us to slim down the interfaces used in vkubelet (and
vkubelet/api) to what is actually used there rather than a huge
interface that is only there to serve the CLI's needs.
* Light up UpdatePod
This PR updates the vkublet/pod.go createOrUpdate(..) method to actually handle
updates. It gets the pod from the provider as before, but now if it exists the method
checks the hash of the spec against the spec of the new pod. If they've changed, it
calls UpdatePod(..).
Also makes a small change to the Server struct to swap from kuberentes.Clientset to kubernetes.Interface
to better facilitate testing with fake ClientSet.
Co-Authored-By: Brian Goff <cpuguy83@gmail.com>
For some reason the kubelet tries to do things that it gets perm
denied. This changes the base.yml in the skaffold file to have
the permissions. In turn, this broke the node deletion test. So,
instead of looking for the deletion event (imperative), we wait
for the node UID to change (declarative).
We're in effect testing the K8s scheduler logic in the test
by setting taints, as opposed to the actual VK itself. If we
want to make sure the taint is set, we can just observe the node
object.
Instead, bind to the pod to the VK node explicitly.
