One common frustration of end users running the e2e suite is that
they take a significant amount of time and it is difficult to
gauge progress.
Even if tailing the logs it can be difficult to see where one
test starts and another ends or understand the if there have been
failures in the past 1h of logs.
This change adds a new custom reporter which prints summary information
as tests complete. This includes the number of tests to run and how
many have been passed/failed/skipped along with which tests have failed.
A new flag can be set which pushes these values to an endpoint. This is
intended for integration with Sonobuoy but any endpoint could consume and
surface this data to the user so they can better understand the state
of the current test run.
e2e_node.test does not set default kubectlPath, which lead to test
errors as following:
[Fail] [sig-storage] EmptyDir volumes [It] pod should support
shared volumes between containers [Conformance]
When the test trying to read file in shared volume, it uses
"kubeclt exec namespace -c container_name -- cat file_name".
However, as variable framework.TestContext.KubectlPath not set,
kubectl binary can not be found in the test and the tast fails.
This patch move kubectlPath flag from RegisterClusterFlags to
RegisterCommonFlags, thus default value for
framework.TestContext.KubectlPath will be set,and
user can also use --kubectl-path flag to set kubectl path.
Signed-off-by: Howard Zhang <howard.zhang@arm.com>
Adds a new flag which allows users to specify a regexp
which will effectively whitelist certain taints and
allow the test framework to startup tests despite having
tainted nodes.
Fixes an issue where e2e tests were unable to be run
in tainted environments due to the framework waiting for
all the nodes to be schedulable without any tolerations.
HandleFlags() was used at e2e package and it depends on sub e2e
framework "config" in core e2e framework. That was invalid dependency.
So this moves HandleFlags() to e2e package for simple dependency.
Tests should never directly add to the global command line, because
some users of the tests might not want them there. For example,
options might only get set directly from a config file.
To achieve that, e2e/framework/config, e2e/framework/viperconfig, and
e2e/framework/test_context.go avoid using the global flag set and
instead expect to be told by the caller which flag set to use. Tests
that called flag directly either get updated or obsolete flags get
removed.
The exception is framework.HandleFlags, which as before directly
implements global command line handling.
This is a breaking change for test suites which do not use that
function (and only those): they now need to ensure that they copy
individual flags from tests. Because the RegisterCommonFlags prototype
has changed, test suite authors will notice due to the resulting
compilation errors.
Moved all flag code from `staging/src/k8s.io/apiserver/pkg/util/[flag|globalflag]` to `component-base/cli/[flag|globalflag]` except for the term function because of unwanted dependencies.
The feature is gated behind a newly introduced 'dump-systemd-journal' flag.
We want to dump the full systemd journal in our scalability performance tests.
Not accepting --provider= (i.e. setting an empty provider name) broke
some test jobs. As suggested in
https://github.com/kubernetes/kubernetes/pull/73402#issuecomment-459368230,
now --provider= and not passing --provider at all both trigger a
message and then continue as if --provider=skeleton had been used.
The empty string was the default and then triggered a special
warning. There's no good reason for that behavior, so now the special
handling for "unset provider" is gone and "skeleton" is the non-empty
default for the value.
This finishes the work started for 1.13: instead of merely warning
about an unknown value given to --profile, the test/e2e/e2e.test
binary will now print an error and refuse to run.
Fixes: #70200
While debugging issues I found myself having to change the constant in the code
for a cluster > 20 nodes, and then on a very small cluster I found myself passing
0 to avoid the mostly useless output (it's useful in specific scenarios but generates
a *lot* of output that doesn't help debugging the rest of the time).
- Move from the old github.com/golang/glog to k8s.io/klog
- klog as explicit InitFlags() so we add them as necessary
- we update the other repositories that we vendor that made a similar
change from glog to klog
* github.com/kubernetes/repo-infra
* k8s.io/gengo/
* k8s.io/kube-openapi/
* github.com/google/cadvisor
- Entirely remove all references to glog
- Fix some tests by explicit InitFlags in their init() methods
Change-Id: I92db545ff36fcec83afe98f550c9e630098b3135
The E2E refactoring tightened the sanity checking of the --provider
parameter such that it only allowed known providers. That seemed to
make sense because it catches typos, but it turned out that various
callers depended on the "accept arbitrary provider value" behavior,
therefore it gets restored.
Not all users of the E2E framework want to run cloud-provider specific
tests. By splitting out the code it becomes possible to decide in
a E2E test suite which providers are supported.
This is achieved in two ways:
- the framework calls certain functions through a provider
interface instead of calling specific cloud provider functions
directly
- tests that are cloud-provider specific directly import the
new provider packages
The ingress test utilities are only needed by a few tests. Splitting
them out into a separate package makes the framework simpler for test
suites not using those tests.
Fixes: #66649
Tests settings should be defined in the test source code itself
because conceptually the framework is a separate entity that not all
test authors can modify.
For the sake of backwards compatibility the name of the command line
flags are not changed.
Tests settings should be defined in the test source code itself
because conceptually the framework is a separate entity that not all
test authors can modify.
Using the new framework/config code also has several advantages:
- defaults can be set with less code
- no confusion around what's a duration
- the options can also be set via command line flags
While at it, a minor bug gets fixed:
- readConfig() returns only defaults when called while
registering Ginkgo tests because Viperize() gets called later,
so the scale in the logging soak test couldn't really be configured;
now the value is read when the test runs and thus can be changed
The options get moved into the "instrumentation.logging"
resp. "instrumentation.monitoring" group to make it more obvious where
they are used. This is a breaking change, but that was already
necessary to improve the duration setting from plain integer to a
proper time duration.
Tests shouldn't have to use the central context for their settings,
because conceptually tests and framework get developed independently.
This does not yet use the new framework/config utility code because
that code still needs to be reviewed.
Besides moving the flags, they also get renamed from the top-level
"--csiImage{Version|Registry}" to
"--storage.csi.image.{version|registry}". These flags were introduced
fairly recently and shouldn't be in use much, so now is a good time to
introduce a hierarchical naming for storage flags, in particular
because more flags will be added soon.
Storing settings in the framework's TestContext is not something that
out-of-tree test authors can do because for them the framework is a
read-only upstream component. Conceptually the same is true for
in-tree tests, so the recommended approach is to define configuration
settings in the code that uses them.
How to do that is a bit uncertain. Viper has several
drawbacks (maintenance status uncertain, cannot list supported
options, cannot validate the configuration file). How to handle
configuration files is currently getting discussed for kubeadm, with
similar concerns about
Viper (https://github.com/kubernetes/kubeadm/issues/1040).
Instead of making a choice now for E2E, the recommendation is that
test authors continue to define command line flags as before, except
that they should do it in their own code and with better flag names.
But the ability to read options also from a file is useful, so
several enhancements get added:
- all settings defined via flags can also be read from a
configuration file, without extra work for test authors
- framework/config makes it possible to populate a struct directly
and define flags with a single function call
- a path and file suffix can be given to --viper-config (as in
"--viper-config /tmp/e2e.json") instead of expecting the file in
the current directory; as before, just plain "--viper-config e2e"
still works
- if "--viper-config" is set, the file must exist; otherwise the
"e2e" config is optional (as before)
- errors from Viper are no longer silently ignored, so syntax errors
are detected early
- Viper support is optional: test suite authors who don't want
it are not forced to use it by the e2e/framework
Individual implementations are not yet being moved.
Fixed all dependencies which call the interface.
Fixed golint exceptions to reflect the move.
Added project info as per @dims and
https://github.com/kubernetes/kubernetes-template-project.
Added dims to the security contacts.
Fixed minor issues.
Added missing template files.
Copied ControllerClientBuilder interface to cp.
This allows us to break the only dependency on K8s/K8s.
Added TODO to ControllerClientBuilder.
Fixed GoDeps.
Factored in feedback from JustinSB.
