This should avoid the 30s delay caused by shell not responding to SIGTERM, and can only be killed by SIGKILL.
If the pod is deleted with the namespace during cleanup, this also makes cleanup faster, and frees up the resources for the next test cases faster.
Immediately exit on receiving an unexpected response, we currently
wait for all the attempts to be exhausted and then return an error.
Signed-off-by: Daman Arora <aroradaman@gmail.com>
We have "-kube-test-repo-list" command line flag to override the image registry. If we store it in global variable, then that overriding cannot take effect.
And this can cause puzzling bugs, e.g.: containerIsUnused() function will compare incorrect image address.
* fix pods tracking and internal error checking in statefulset tests
* fix stateful set pod recreation and event spam
- do not emit events when pod reaches terminal phase
- do not try to recreate pod until the old pod has been removed from
etcd storage
* fix conflict race in statefulset rest update
statefulset controller does less requests per sync now and thus can
reconcile status faster, thus resulting in a higher chance for conflicts
We want:
- To keep test annotations simple, using both WithFeatureGate
and WithFeature should only be necessary when a test really
has requirements that go beyond "feature gate needs to be enabled".
- To run tests which depend only on feature gates being enabled
in the ci-kubernetes-e2e-kind-alpha-features resp.
ci-kubernetes-e2e-kind-beta-features, because otherwise we
may have a proliferation of many bespoke jobs which only run
very few tests. This would make testing more expensive for
Kubernetes.
- To enable those tests only once in the ci-kubernetes-e2e-kind-alpha-features
and ci-kubernetes-e2e-kind-beta-features definition instead
of having to update those each time feature gates change.
This can be achieved by adding `Feature:Alpha` resp. `Feature:Beta` as Ginkgo
labels instead of just `Alpha` and `Beta`. Then jobs which are configured to
skip tests with feature dependencies via --label-filter=!/Feature:.+/ will skip
tests which are labeled with just WithFeatureGate. The ci-kubernetes jobs
can select to include such tests with a special regexp that mimicks
a negative lookahead (see k8s.io/community/contributors/devel/sig-testing/e2e-tests.md)
Note that removing WithFeature depends on first updating job definitions to use
--label-filter or to skip based on the inline `[Alpha]` or `[Beta]` text,
otherwise tests that were previously skipped because of WithFeature might
start to run in jobs which don't have the feature gate enabled.
Even if the textlogger which writes to Ginkgo is installed as the logger in
klog, klog still does some verbosity checks itself (for example,
klog.V().Enabled).
Therefore the framework has to keep the verbosity settings in the textlogger
and in klog consistent. This is done by wrapping the Set call instead of
replacing it.
The dead code was found with:
deadcode -test -filter=k8s.io/kubernetes/test/e2e/framework/... ./test/e2e ./test/e2e_node ./test/e2e_node ./test/e2e_kubeadm
See https://go.dev/blog/deadcode for an introduction.
This removes everything that was reported as dead code for
test/e2e/framework/providers.
When stopping polling, the provided messages becomes the complete failure
message. This means that the code which calls gomega.StopTrying must include
the pod in the message instead of just summarizing the phase. This makes the
failure more useful.
ginkgo.GinkgoHelper is a recent addition to ginkgo which allows functions to
mark themselves as helper. This then changes which callstack gets reported for
failures. It makes sense to support the same mechanism also for logging.
There's also no reason why framework.Logf should produce output that is in a
different format than klog log entries. Having time stamps formatted
differently makes it hard to read test output which uses a mixture of both.
Another user-visible advantage is that the error log entry from
framework.ExpectNoError now references the test source code.
With textlogger there is a simple replacement for klog that can be reconfigured
to let the caller handle stack unwinding. klog itself doesn't support that
and should be modified to support it (feature freeze).
Emitting printf-style output via that logger would work, but become less
readable because the message string would get quoted instead of printing it
verbatim as before. So instead, the traditional klog header gets reproduced
in the framework code. In this example, the first line is from klog, the second
from Logf:
I0111 11:00:54.088957 332873 factory.go:193] Registered Plugin "containerd"
...
I0111 11:00:54.987534 332873 util.go:506] >>> kubeConfig: /var/run/kubernetes/admin.kubeconfig
Indention is a bit different because the initial output is printed before
installing the logger which writes through ginkgo.GinkgoWriter.
One welcome side effect is that now "go vet" detects mismatched parameters for
framework.Logf because fmt.Sprintf is called without mangling the format
string. Some of the calls were incorrect.
A stand-alone binary shouldn't import the test/e2e/framework, which is targeted
towards usage in a Ginkgo test suite. This currently works, but will break once
test/e2e/framework becomes more opinionated about how to configure logging.
The simplest solution is to duplicate the one short function that the binary
was calling in the framework.
Now that we have it (8a89a1f5a5), let's also make sure that
the new WithFlaky is used everywhere instead if [Flaky]. This way it can be
used for filtering by label.
