we should only use this env var for `arm`, since `arm64` is fully
supported by etcd folks, let us drop this!
(ex - https://github.com/etcd-io/etcd/releases/tag/v3.5.6)
ppc64le comment should be dropped as well
Signed-off-by: Davanum Srinivas <davanum@gmail.com>
The check for "resources available on a node" must treat nodes that are not
listed as "no resources available". The previous logic only worked because all
nodes were listed during E2E testing. The upcoming integration testing is
covering additional scenarios and triggered this broken case.
It provides more readable output and has additional APIs for using it inside a
unit test. goleak.IgnoreCurrent is needed to filter out the goroutine that gets
started when importing go.opencensus.io/stats/view.
In order to handle background goroutines that get created on demand and cannot
be stopped (like the one for LogzHealth), a helper function ensures that those
are running before calling goleak.IgnoreCurrent. Keeping those goroutines
running is not a problem and thus not worth the effort of adding new APIs to
stop them.
Other goroutines are genuine leaks for which no fix is available. Those get
suppressed via IgnoreTopFunction, which works as long as that function
is unique enough.
Example output for the leak fixed in https://github.com/kubernetes/kubernetes/pull/115423:
E0202 09:30:51.641841 74789 etcd.go:205] "EtcdMain goroutine check" err=<
found unexpected goroutines:
[Goroutine 4889 in state chan receive, with k8s.io/apimachinery/pkg/watch.(*Broadcaster).loop on top of the stack:
goroutine 4889 [chan receive]:
k8s.io/apimachinery/pkg/watch.(*Broadcaster).loop(0xc0076183c0)
/nvme/gopath/src/k8s.io/kubernetes/vendor/k8s.io/apimachinery/pkg/watch/mux.go:268 +0x65
created by k8s.io/apimachinery/pkg/watch.NewBroadcaster
/nvme/gopath/src/k8s.io/kubernetes/vendor/k8s.io/apimachinery/pkg/watch/mux.go:77 +0x116
>
- test/e2e/framework/*.go should have very minimal dependencies.
We can enforce that via import-boss.
- What each test/e2e/framework/* sub-package uses is less relevant,
although ideally it also should be as minimal as possible in each case.
Enforcing this via import-boss ensures that new dependencies get flagged as
problem and thus will get additional scrutiny. It might be okay to add them,
but it needs to be considered.
Since the pod names are reused across the test, searching the logs is
currently difficult.
Use a uuid for each pod name to make grepping the logs easier. Also,
always include the pod name and pod namespace in any logs or error
messages to make debugging easier.
Signed-off-by: David Porter <david@porter.me>
The previous approach was based on the observation that some Prow jobs use the
--report-dir parameter instead of the E2E_REPORT_DIR env variable. Parsing the
command line was necessary to use the --json-report and --junit-report
parameters.
But that is complex and can be avoided by triggering the creation of complete
reports in the E2E test suite. The paths are hard-coded and relative to the
report directory to keep the code simple.
There was a report that k8s-triage started processing more data after
6db4b741dd was merged. It's unclear whether
that was because of the new <report-dir>/ginkgo_report.xml file. To avoid
this potential problem, the reports are now in a "ginkgo" sub-directory.
While at it, error checking gets enhanced:
- Create directories at the start of
the suite and bail out early if that fails.
- *All* e2e suites using the framework do this, not just test/e2e.
- Added missing error checking of truncated JUnit report writing.
Currently `kubectl debug` only supports passing names in command line.
However, users might want to pass resources in files by passing `-f` flag like
in all other kubectl commands.
This PR adds this ability.
With the change of the CRI-O jobs to use butane, we now have a
verification for base64 data urls in place. This means that the
following URL is invalid:
```
data:text/plain;base64,GCE_SSH_PUBLIC_KEY_FILE_CONTENT
```
This means we have to pass valid base64 to the URL. To fix that, we now
allow to inject SSH key values with both, the
`GCE_SSH_PUBLIC_KEY_FILE_CONTENT` field and its base64 encoded variant.
Signed-off-by: Sascha Grunert <sgrunert@redhat.com>
The NPD test checks when NPD started to determine if it is needed to
check the kubelet start event. The current logic requires parsing the
journalctl logs which is quite fragile and is broken now because of
systemd changing the expected log format.
Newer versions of systemd do not print "end at" or "logs begin at" and
instead may print "No entries", which will result in the test panicking.
```
$ journalctl -u foo.service
-- No entries --
```
For units started, it will not print "end at" or "logs begin at":
```
root@ubuntu-jammy:~# journalctl -u foo.service
Feb 08 22:02:19 ubuntu-jammy systemd[1]: Started /usr/bin/sleep 1s.
Feb 08 22:02:20 ubuntu-jammy systemd[1]: foo.service: Deactivated successfully.
```
To avoid relying on log parsing which is fragile, let's instead directly
ask systemd when the NPD service started and parse the resulting
timestamp.
Signed-off-by: David Porter <david@porter.me>
There is a test in wait.sh integration suite which is checking the
given timeout value(passed by user) is equal to actual happened timeout value.
However, this test rarely gets `no matching resources found` error and
causes flakyiness. The reason is we are running wait command, immediately
after applying deployment. In reality, timeout test does not care about
deployment, since it is testing the timeout by passing invalid configurations.
But we need this deployment to not get `no matching resources found` error.
That's why, this PR adds deployment assertion before executing wait command.
The recently introduced failure handling in ExpectNoError depends on error
wrapping: if an error prefix gets added with `fmt.Errorf("foo: %v", err)`, then
ExpectNoError cannot detect that the root cause is an assertion failure and
then will add another useless "unexpected error" prefix and will not dump the
additional failure information (currently the backtrace inside the E2E
framework).
Instead of manually deciding on a case-by-case basis where %w is needed, all
error wrapping was updated automatically with
sed -i "s/fmt.Errorf\(.*\): '*\(%s\|%v\)'*\",\(.* err)\)/fmt.Errorf\1: %w\",\3/" $(git grep -l 'fmt.Errorf' test/e2e*)
This may be unnecessary in some cases, but it's not wrong.
Instead of pod responses being printed to the log each time polling fails, we
get a consolidated failure message with all unexpected pod responses if (and
only if) the check times out or a progress report gets produced.
This renames PodsResponding to WaitForPodsResponding for the sake of
consistency and adds a timeout parameter. That is necessary because some other
users of NewProxyResponseChecker used a much lower timeout (2min vs. 15min).
Besides simplifying some code, it also makes it easier to rewrite
ProxyResponseChecker because it only gets used in WaitForPodsResponding.
WaitForPodToDisappear was always called such that it listed all pods, which
made it less efficient than trying to get just the one pod it was checking for.
Being able to customize the poll interval in practice wasn't useful, therefore
it can be replaced with WaitForPodNotFoundInNamespace.
WaitForPods is now a generic function which lists pods and then checks the pods
that it found against some provided condition. A parameter determines how many
pods must be found resp. match the condition for the check to succeed.
The code becomes simpler (78 insertions, 91 deletions), easier to read (all
code entirely inside WaitForPodsRunningReady, no need to declare and later
overwrite variables) and possibly more correct (if all API calls failed,
the resulting error was ignored when allowedNotReadyPods > 0).
None of the users of the functions passed anything other than nil or an empty
map and the implementation ignore the parameter - it seems like a candidate for
simplification.
When a Gomega failure is converted to an error, the stack at the time when the
failure occurs may be useful: error wrapping provides some bread crumbs that
can be followed to determine where the failure really occurred, but error
wrapping may be missing or ambiguous.
To provide the additional information, a FailureError now includes a full stack
backtrace. The backtrace intentionally makes no attempt to exclude framework
functions besides the gomega support itself because helpers like
e2e/framework/pod may be relevant.
That backtrace is not included in the failure message for the sake of
brevity. Instead, it gets logged as part of the test's output.
gomega.Eventually provides better progress reports: instead of filling up the
log with rather useless one-line messages that are not enough to to understand
the current state, it integrates with Gingko's progress reporting (SIGUSR1,
--poll-progress-after) and then dumps the same complete failure message as
after a timeout. That makes it possible to understand why progress isn't
getting made without having to wait for the timeout.
The other advantage is that the failure message for some unexpected pod state
becomes more readable: instead of encapsulating it as "observed object" inside
an error, it directly gets rendered by gomega.
Calling gomega.Expect/Eventually/Consistently deep inside a helper call chain
has several challenges:
- the stack offset must be tracked correctly, otherwise the callstack
for the failure starts at some helper code, which is often not informative
- augmenting the failure message with additional information from each
caller implies that each caller must pass down a string and/or format
string plus arguments
Both challenges can be solved by returning errors:
- the stacktrace is taken at that level where the error is
treated as a failure instead of passing back an error, i.e.
inside the It callback
- traditional error wrapping can add additional information, if
desirable
What was missing was some easy way to generate an error via a gomega
assertion. The new infrastructure achieves that by mirroring the
Gomega/Assertion/AsyncAssertion interfaces with errors as return values instead
of calling a fail handler.
It is intentionally less flexible than the gomega APIs:
- A context must be passed to Eventually/Consistently as first
parameter because that is needed for proper timeout handling.
- No additional text can be added to the failure through this
API because error wrapping is meant to be used for this.
- No need to adjust the callstack offset because no backtrace
is recorded when a failure occurs.
To avoid the useless "unexpected error" log message when passing back a gomega
failure, ExpectNoError gets extended to recognize such errors and then skips
the logging.
Calling WaitForPodTerminatedInNamespace after testFlexVolume is useless because
the client pod that it waits for always gets deleted by testVolumeClient:
0fcc3dbd55/test/e2e/framework/volume/fixtures.go (L541-L546)
Worse, because WaitForPodTerminatedInNamespace treats "not found" as "must keep
polling", these two tests always kept waiting for 5 minutes:
Kubernetes e2e suite: [It] [sig-storage] Flexvolumes should be mountable
when non-attachable 6m4s
The only reason why these tests passed is that WaitForPodTerminatedInNamespace
used to return the "not found" API error. That is not guaranteed and about to
change.
Unknown pods are pods which are unknown pods to the kubelet, but are still
running in the container runtime. If kubelet detects a pod which is not in
the config (i.e. not present in API-server or static pod), but running as
detected in container runtime, kubelet should aggressively terminate the pod.
This situation can be encountered if a pod is running, then kubelet is
stopped, and while stopped, the manifest is deleted (by force deleting the
API pod or deleting the static pod manifest), and then restarting the
kubelet. Upon restart, kubelet will see the pod as running via the container
runtime, but it will not be present in the config, thus making the pod a
"unknown pod". Kubelet should then proceed to terminate these unknown pods.
Add two tests that ensure that unknown pods will be terminated (1)
static pods and (2) API pods. The test will start a pod, stop the
kubelet, force delete the pod (by deleting the manifest or force
deleting the pod), and then restarting the kubelet. The container
runtime is then queried to ensure the containers are terminated by
kubelet.
Signed-off-by: David Porter <david@porter.me>
This is used across multiple tests, so let's move into the util file.
Also, refactor it a bit to provide a better error message in case of a
failure.
Signed-off-by: David Porter <david@porter.me>
* Add integration tests for StatefulSetStartOrdinal feature
* Move expensive test setup (apiserver and running controller) to be run once in StatefulSetStartOrdinal parameterized tests
Image versions are already explicitly set in our manifests
configuration, so tests should not be setting these values
to ensure we're using the same versions across the board.
Adds integration tests for the following scenarios with
MultiCIDRRangeAllocator enabled:
- ClusterCIDR is released when an associated node is deleted.
- ClusterCIDR delete when a node is associated, validate the finalizer
behavior, make sure that deleted ClusterCIDR is cleaned up after the
associated node is deleted.
- ClusterCIDR marked as terminating due to deletion must not be used for
allocating PodCIDRs to new nodes.
- Tie break behavior when multiple ClusterCIDRs are eligible to
allocate PodCIDRs to a node.
The device plugin test expects that no other pods are running prior to
the test starting. However, it has been observed that in some cases
some resources may still be around from previous tests. This is because
the deletion of resources from other tests is handled by deleting that
test's framework's namespace which is done asynchronously without
waiting for the other test's namespace to be deleted.
As a result, when the node e2e device plugin starts, there may still be
other pods in process of termination. To work around this, add a retry
to the device plugin test to account for the time it takes to delete the
resources from the prior test.
Signed-off-by: David Porter <david@porter.me>
In the `NodeSupportsPreconfiguredRuntimeClassHandler`, update the check
for the runtime handler to return a failure if the
`/etc/containerd/config.toml` or `/etc/crio/crio.conf` config files do
not exist. If an error is returned, then the underlying test will be
skipped.
Test manually with starting a kind cluster and moving the containerd
config file and verifying that the test is skipped:
```
$ docker exec -it kind-worker /bin/bash
root@kind-worker:/# mv /etc/containerd/config.toml /etc/containerd/config.toml.bak
```
```
make WHAT="test/e2e/e2e.test"
$ ./_output/bin/e2e.test -kubeconfig /tmp/kubeconfig_kind -ginkgo.focus=".*should run a Pod requesting a RuntimeClass with a configured handler.*" --num-nodes=1 2>&1 -ginkgo.v=1 | tee -i "/tmp/build-log.txt"
[sig-node] RuntimeClass [It] should run a Pod requesting a RuntimeClass with a configured handler [NodeFeature:RuntimeHandler]
test/e2e/common/node/runtimeclass.go:85
[SKIPPED] Skipping test as node does not have E2E runtime class handler preconfigured in container runtime config: command terminated with exit code 1
```
Signed-off-by: David Porter <david@porter.me>
The recently introduced failure handling in ExpectNoError depends on error
wrapping: if an error prefix gets added with `fmt.Errorf("foo: %v", err)`, then
ExpectNoError cannot detect that the root cause is an assertion failure and
then will add another useless "unexpected error" prefix and will not dump the
additional failure information (currently the backtrace inside the E2E
framework).
Instead of manually deciding on a case-by-case basis where %w is needed, all
error wrapping was updated automatically with
sed -i "s/fmt.Errorf\(.*\): '*\(%s\|%v\)'*\",\(.* err)\)/fmt.Errorf\1: %w\",\3/" $(git grep -l 'fmt.Errorf' test/e2e*)
This may be unnecessary in some cases, but it's not wrong.
Instead of pod responses being printed to the log each time polling fails, we
get a consolidated failure message with all unexpected pod responses if (and
only if) the check times out or a progress report gets produced.
This renames PodsResponding to WaitForPodsResponding for the sake of
consistency and adds a timeout parameter. That is necessary because some other
users of NewProxyResponseChecker used a much lower timeout (2min vs. 15min).
Besides simplifying some code, it also makes it easier to rewrite
ProxyResponseChecker because it only gets used in WaitForPodsResponding.
WaitForPodToDisappear was always called such that it listed all pods, which
made it less efficient than trying to get just the one pod it was checking for.
Being able to customize the poll interval in practice wasn't useful, therefore
it can be replaced with WaitForPodNotFoundInNamespace.
The test validates the following endpoints
- deleteApiregistrationV1CollectionAPIService
- patchApiregistrationV1APIServiceStatus
- replaceApiregistrationV1APIService
- replaceApiregistrationV1APIServiceStatus
WaitForPods is now a generic function which lists pods and then checks the pods
that it found against some provided condition. A parameter determines how many
pods must be found resp. match the condition for the check to succeed.
The code becomes simpler (78 insertions, 91 deletions), easier to read (all
code entirely inside WaitForPodsRunningReady, no need to declare and later
overwrite variables) and possibly more correct (if all API calls failed,
the resulting error was ignored when allowedNotReadyPods > 0).
None of the users of the functions passed anything other than nil or an empty
map and the implementation ignore the parameter - it seems like a candidate for
simplification.
When a Gomega failure is converted to an error, the stack at the time when the
failure occurs may be useful: error wrapping provides some bread crumbs that
can be followed to determine where the failure really occurred, but error
wrapping may be missing or ambiguous.
To provide the additional information, a FailureError now includes a full stack
backtrace. The backtrace intentionally makes no attempt to exclude framework
functions besides the gomega support itself because helpers like
e2e/framework/pod may be relevant.
That backtrace is not included in the failure message for the sake of
brevity. Instead, it gets logged as part of the test's output.
gomega.Eventually provides better progress reports: instead of filling up the
log with rather useless one-line messages that are not enough to to understand
the current state, it integrates with Gingko's progress reporting (SIGUSR1,
--poll-progress-after) and then dumps the same complete failure message as
after a timeout. That makes it possible to understand why progress isn't
getting made without having to wait for the timeout.
The other advantage is that the failure message for some unexpected pod state
becomes more readable: instead of encapsulating it as "observed object" inside
an error, it directly gets rendered by gomega.
Calling gomega.Expect/Eventually/Consistently deep inside a helper call chain
has several challenges:
- the stack offset must be tracked correctly, otherwise the callstack
for the failure starts at some helper code, which is often not informative
- augmenting the failure message with additional information from each
caller implies that each caller must pass down a string and/or format
string plus arguments
Both challenges can be solved by returning errors:
- the stacktrace is taken at that level where the error is
treated as a failure instead of passing back an error, i.e.
inside the It callback
- traditional error wrapping can add additional information, if
desirable
What was missing was some easy way to generate an error via a gomega
assertion. The new infrastructure achieves that by mirroring the
Gomega/Assertion/AsyncAssertion interfaces with errors as return values instead
of calling a fail handler.
It is intentionally less flexible than the gomega APIs:
- A context must be passed to Eventually/Consistently as first
parameter because that is needed for proper timeout handling.
- No additional text can be added to the failure through this
API because error wrapping is meant to be used for this.
- No need to adjust the callstack offset because no backtrace
is recorded when a failure occurs.
To avoid the useless "unexpected error" log message when passing back a gomega
failure, ExpectNoError gets extended to recognize such errors and then skips
the logging.
Calling WaitForPodTerminatedInNamespace after testFlexVolume is useless because
the client pod that it waits for always gets deleted by testVolumeClient:
0fcc3dbd55/test/e2e/framework/volume/fixtures.go (L541-L546)
Worse, because WaitForPodTerminatedInNamespace treats "not found" as "must keep
polling", these two tests always kept waiting for 5 minutes:
Kubernetes e2e suite: [It] [sig-storage] Flexvolumes should be mountable
when non-attachable 6m4s
The only reason why these tests passed is that WaitForPodTerminatedInNamespace
used to return the "not found" API error. That is not guaranteed and about to
change.
The `runPausePod` timeout was 1 minute previously which appears to be
too short and timing out in some tests.
Switch to `f.Timeouts.PodStartShort` which is the common timeout used to wait
for pods to start which defaults to 5min.
Also refactor to remove `runPausePodWithoutTimeout` and instead rely on
`runPausePod` since we do not make the timeout customizable directly
(it can be changed via the test framework if desired).
Signed-off-by: David Porter <david@porter.me>
Adds integration tests for the following scenarios with
MultiCIDRRangeAllocator enabled:
- ClusterCIDR is released when an associated node is deleted.
- ClusterCIDR delete when a node is associated, validate the finalizer
behavior, make sure that deleted ClusterCIDR is cleaned up after the
associated node is deleted.
- ClusterCIDR marked as terminating due to deletion must not be used for
allocating Pod CIDRs to new nodes.
- Tie break behavior when multiple ClusterCIDRs are eligible to
allocate Pod CIDRs to a node.
PVCs using the ReadWriteOncePod access mode can only be referenced by a
single pod. When a pod is scheduled that uses a ReadWriteOncePod PVC,
return "Unschedulable" if the PVC is already in-use in the cluster.
To support preemption, the "VolumeRestrictions" scheduler plugin
computes cycle state during the PreFilter phase. This cycle state
contains the number of references to the ReadWriteOncePod PVCs used by
the pod-to-be-scheduled.
During scheduler simulation (AddPod and RemovePod), we add and remove
reference counts from the cycle state if they use any of these
ReadWriteOncePod PVCs.
In the Filter phase, the scheduler checks if there are any PVC reference
conflicts, and returns "Unschedulable" if there is a conflict.
This is a required feature for the ReadWriteOncePod beta. See for more context:
https://github.com/kubernetes/enhancements/tree/master/keps/sig-storage/2485-read-write-once-pod-pv-access-mode#beta
Node E2E tests do not run a scheduler, so the host exec pod must have
the `spec.nodeName` set explicitly.
Signed-off-by: David Porter <david@porter.me>
The ClusterIP allocator tries to reserve on part of the ServiceCIDR
to allocate static IPs to the Services.
The heuristic of the allocator to obtain the offset was taking into
account the whole range size, not the IPs available in the range, the
subnet address and the broadcast address for IPv4 are not available.
This caused that for CIDRs with 16 hosts, /28 for IPv4 and /124 for
IPv6, the offset calculated was higher than the max number of available
addresses on the allocator, causing this to panic.
Change-Id: I6c6f527b0a600b3612be37769e405b8fb3dd33a8
There are two runtime class tests which required the container runtime
config to include explicit configuration for `test-handler`. The current
logic skips these tests in non GCE environments. This skip is too strict
since the test is skipped in node e2e environments and in other
environments such as kind, which support running the test and also
configure `test-handler`.
Instead of skipping based on provider, add a new function
`NodeSupportsPreconfiguredRuntimeClassHandler` which examines the
underlying container runtime config and checks if the config includes
`test-handler`. The check is a bit brittle since it assumes container
runtime config paths, but it is a net improvement over skipping the test
entirely on non GCE environments.
This results in the test working in the common test environments, namely
GCE kube-up, node e2e, and kind.
Signed-off-by: David Porter <david@porter.me>
When the e2e_node/checkpoint_container.go test was introduced no CRI
implementation supported the new CheckpointContainer RPC yet.
With the release of CRI-O 1.25 the CheckpointContainer is implemented
and the test has been extended to see if the content of the checkpoint
is as expected.
The test is skipped if the ContainerCheckpoint feature gate is disabled
or if the CRI implementation does not support the CheckpointContainer
RPC.
Signed-off-by: Adrian Reber <areber@redhat.com>
we are saving this information in an env variable `KUBE_INTEGRATION_ETCD_URL`
So just pick it up from there when needed. Currently when someone uses
framework.RunCustomEtcd directly, the global variable is *not* set and the
code that uses `GetEtcdURL` returns empty string.
Signed-off-by: Davanum Srinivas <davanum@gmail.com>
Since we need to gather kubelet metrics for CPU Manager and Topology
Manager, renaming this function to a more generic name.
Signed-off-by: Swati Sehgal <swsehgal@redhat.com>
The GlogSetter method is used by three components to change verbosity at
runtime through HTTP APIs. This used to work only for text output with klog
calls, but not for text output through the klog logger or for JSON output.
Now loggers can also provide a callback for changing their verbosity at
runtime. Implementing that implies that the Create factory method has to be
extended, which is an API break for the Go package, but not an API break for
the configuration file and command line flags, which is what matters for the
"api/v1" component API.
The condition methods will eventually all take a context. Since we
have been provided one, alter the accepted condition type and
change the four references in tree.
Collers of ExponentialBackoffWithContext should use a condition
aware function (ConditionWithContextFunc). If the context can be
ignored the helper ConditionFunc.WithContext can be used to convert
an existing function to the new type.
Using WaitTimeoutForPodRunningInNamespace followed by ExpectError was not very
precise (any error passed the check, not just the expected timeout) and
hard to read. Now the test's expectation is spelled out explicitly: the pod
must stay in pending.
These helper functions can be used in combination with
omega.Eventually/Consistently to implement polling of objects that is aware of
Kubernetes apiserver conventions:
- retry on certain errors instead of giving up,
with "not found" handling decided by the caller (may or may not
be fatal, depending on the test)
- sleep if requested by apiserver
The E2E framework contains several functions which only differ in how they get
name and namespace: from an API object (WaitForPodRunningInNamespace) or as
separate parameters (WaitTimeoutForPodRunningInNamespace).
NamespacedName and the NamedObject interface enable writing helper functions
that can be called with both an API object (like *v1.Pod, which implements
Object and thus NamedObject) and name+namespace string (via
NamespacedName).
The other advantage of NamespacedName is that the order of name and namespace
parameter cannot be mixed up.
NamespacedName was derived from k8s.io/apimachinery/pkg/types.NamespacedName. A
separate type in the framework package was chosen a) to avoid additional
imports in test code and b) because the interface might not be suitable for
k8s.io/apimachinery.
* Improving the output of tests in case of error
* Better error message
Also, the condition in the second case was reversed
* Fixing 2 tests whose condition was inverted
* Again I got the conditions wrong
* Sorry for the confusion
* Improved error messages on failures
This significantly reduces the surface area of the fieldmanager package
by hiding all the private "managers" objects, as well as the interface
that was made specifically for these. There is no reason to configure
these.
Primarily this protects against accidentally polling with the default interval
of 10ms. Setting these defaults may also make some tests simpler because they
don't need to override the defaults.
Various different tests all have their own poll intervals. As a start towards
consolidating that, the interval from test/e2e/framework/pod (as one of the
most common cases for polling) is moved into the framework.
Changing other helper packages and tests needs to follow.
This consolidates timeout handling. In the future, configuration of all
timeouts via a configuration file might get added. For now, the same three
legacy command line flags for the timeouts that get moved continue to be
supported.
All usage of builder pattern is convertible to cpuset.New()
with the same or fewer lines of code.
Migrate Builder.Add to a private method of CPUSet, with a comment
that it is only intended for internal use to preserve immutable
propoerty of the exported interface.
This also removes 'require' library dependency, which avoids
non-standard library usage.
In 'set', conversions to slice are done also, but with different names:
ToSliceNoSort() -> UnsortedList()
ToSlice() -> List()
Reimplement List() in terms of UnsortedList to save some duplication.
Removes exit/fatal from cpuset library.
Usage in podresources test was not necessary.
Library reference in cpu_manager_test was moved to a local function, and
converted to use e2e test framework error catching.
The terminationGracePeriodSeconds field was capitalized wrongly, leading
to unknown field warning when running deployment lifecycle e2e:
STEP: patching the Deployment @ 01/06/23 10:59:25.969
W0106 10:59:25.978292 433473 warnings.go:70] unknown field "spec.template.spec.TerminationGracePeriodSeconds"
The issue was undiscovered because the patched value didn't change, so
the test would succeed anyway.
Signed-off-by: Quan Tian <qtian@vmware.com>
Before, in RunPostFilterPlugins, we didn't distinguish between unschedulable and unresolvable
because we only have one postFilterPlugin by default, now, we have at least two, we should
make sure that once a postFilterPlugin returns unresolvable, we'll return directly
Signed-off-by: Kante Yin <kerthcet@gmail.com>
If we were to add new fields in TimeoutContext, the current users of
NewFrameworkWithCustomTimeouts might run into failures unless they get modified
to also set those new fields. This is error-prone.
A better approach is to let users of NewFrameworkWithCustomTimeouts override
fields by setting just those and use the normal defaults for the others.
Ginkgo relies on all workers defining all tests in exactly the same order. This
wasn't guaranteed for these tests, with the result that some tests might have
been executed more than once and others not at all when running in parallel.
This was noticed when some of these tests started to flake and then were
reported both as failure and success, as if they had been retried.
In v1.26.0, these tests only used the timeout context while waiting for a CSI
call. This restores that behavior, just in case that it is relevant. No test
flakes are known because of this.
The intend of timeout handling (for the entire "It" and not just a few calls)
becomes more obvious and simpler when using ginkgo.NodeTimeout as decorator.
It doesn't make sense for the E2E framework to have command line options that
don't do anything because then all test suites built with the framework inherit
those options.
For -list-images and -list-conformance-tests the solution is to move the
implementation into the framework (-list-images) respectively move the flag
into test/e2e (-list-conformance-tests).
The placement was decided based on the observation that image patching is
common functionality while conformance testing is specific to one test suite.
The "[sig-network] DNS HostNetwork should resolve DNS of partial qualified
names for services on hostNetwork pods with dnsPolicy:
ClusterFirstWithHostNet" test assumes that a service named "kube-dns"
exists in the "kube-system" namespace. This assumption is valid if the
cluster was configured using kubeadm, but the assumption may be invalid
otherwise.
As the test uses dnsPolicy: ClusterFirst (as opposed to dnsPolicy: None),
it does not need to specify the name server in dnsConfig. Omitting
dnsConfig.nameservers obviates the need to look up the service.
Follow-up to commit add4652352.
* test/e2e/network/dns.go: Don't look up or use the kube-dns cluster IP
address as it might not exist on clusters that were not configured using
kubeadm.
Bring back the number of test spec which was dropped earlier.
It's now available in the reporting node of `ReportBeforeSuite` by extracting
the number from report.PreRunStats.SpecsThatWillBeRun.
Signed-off-by: Dave Chen <dave.chen@arm.com>
The old tests were no longer passing with Ginkgo v2.5.0. Instead of keeping the
old approach of checking recorded spec results, now the tests actually cover
what we care about most: the results recorded in JUnit.
This also gets rid of having to repeat the stack backtrace twice (once as part
of the output, once for the separate backtrace field).
All information that we want will be written into the failure XML element's
data. We don't need the message tag and don't want it because our
tools (kettle, testgrid, spyglass) would then just concatenate the two strings.
This gets implemented for us by Ginkgo. However, truncating the failure message
is not supported there at the moment. It's unclear how important that is,
therefore this (recently added feature) gets removed.
The NodePort functionality can be tested within the cluster.
Testing from outside the cluster assumes that there is connectivity
between the e2e.test binary and the cluster under test, that is not
always true, and in some cases is exposed to external factors or
misconfigurations like wrong routes or firewall rules that impact
on the test.
Change-Id: Ie2fc8929723e80273c0933dbaeb6a42729c819d0
* Wire generic context to better handle timeout
* Add integration test for wait timeout
* kubectl wait: Fix integration test always passing issue
Currently, `kubectl wait` integration test always passes even if
it gets an error. Problem is object check is done after errexit is
turned off.
This PR redirects error to output and correctly assures that
object is expected status and if it is not, test should fail.
The background goroutine was started with the context from ginkgo.BeforeEach,
which then led to "context canceled" errors. While at it, the entire goroutine
start/stop gets moved into the BeforeEach and simplified.
These tests create their own pods, however, they were inside a block
that was always creating a pod, that was not used later on the tests,
but can influence on the conditions asserted to succeed the test.
Change-Id: I3bb9a0f123fb0766d75934ef8e197f92e3f5f3b8
Using the ctx of the ginkgo.BeforeEach in callbacks that are invoked after the
BeforeEach is done causes "context canceled" errors. Previously, this code used
context.TODO(). The best solution is to create a new context and cancel it
during test cleanup, then that context can be used for the API calls and as
stop channel.
After adding error checking in df5d84ae81, the "[sig-cli] Kubectl client
Simple pod should return command exit codes [Slow] running a failing command
without --restart=Never, but with --rm" test was found to time out.
Doubling the timeout might help. Alternatively, the entire
WaitForPodToDisappear could get removed, which would make this scenario similar
to the others which also don't wait.
the test uses a BeforeEach block to create a pod with the name defined
but the simplePodName variables, however, some test use the value of
the variable directly.
To avoid future problems use the variable name instead of the value on
all tests.
Change-Id: I21a01019d91fe5ae7e35566184420001978ce355
All code must use the context from Ginkgo when doing API calls or polling for a
change, otherwise the code would not return immediately when the test gets
aborted.
* Add tracker types and tests
* Modify ResourceEventHandler interface's OnAdd member
* Add additional ResourceEventHandlerDetailedFuncs struct
* Fix SharedInformer to let users track HasSynced for their handlers
* Fix in-tree controllers which weren't computing HasSynced correctly
* Deprecate the cache.Pop function
The Feature:SCTPConnectivity tests cannot run at the same time as the
"X doesn't cause sctp.ko to be loaded" tests, since they may cause
sctp.ko to be loaded. We had dealt with this in the past by marking
them [Disruptive], but this isn't really fair; the problem is more
with the sctp.ko-checking tests than it is with the SCTPConnectivity
tests. So make them not [Disruptive] and instead make the
sctp.ko-checking tests be [Serial].
There were two SCTP tests grouped together in
test/e2e/network/service.go, but one of them wasn't a service test...
so move the SCTP service test to be grouped with the other service
tests, and the SCTP hostport tests to be grouped with other
non-service tests.
The SCTP HostPort test was checking that creating a pod with an SCTP
HostPort would create a certain iptables rule, but the handling of
HostPorts is now up to CRI, not kubelet, so kubernetes e2e cannot
assume it will implement the feature in any specific way.
(The test still ensures that (a) the apiserver accepts SCTP HostPorts,
and (b) neither kubelet nor the runtime causes the SCTP kernel module
to be loaded as part of creating a pod with an SCTP HostPort.)
We had a test that creating a Service with an SCTP port would create
an iptables rule with "-p sctp" in it, which let us test that
kube-proxy was doing vaguely the right thing with SCTP even if the e2e
environment didn't have SCTP support. But this would really make much
more sense as a unit test.
Windows ComputerNames cannot exceed 15 characters. This causes a few tests to fail
when the node names exceed that limit. Additionally, the checks should be case
insensitive.
The Gingo v2 time suffix is hh:mm:ss without the .xyz sub-second details if the
time stamp happens to land exactly on a second.
This change fixes test flakes like the following:
-STEP: Building a namespace api object, basename test-namespace
+STEP: Building a namespace api object, basename test-namespace 12/13/22 11:43:53
--- FAIL: TestCleanup (36.79s)
ginkgo.DeferCleanup has multiple advantages:
- The cleanup operation can get registered if and only if needed.
- No need to return a cleanup function that the caller must invoke.
- Automatically determines whether a context is needed, which will
simplify the introduction of context parameters.
- Ginkgo's timeline shows when it executes the cleanup operation.
- use `ginkgo.DeferCleanup` instead of clean up in the AfterEach block
- encourage use of ginkgo by not extending expect.go
Signed-off-by: Swati Sehgal <swsehgal@redhat.com>
Add a test case with a DaemonSet behind a simple load balancer whose
address is being constantly hit via HTTP requests.
The test passes if there are no errors when doing HTTP requests to the
load balancer address, during DaemonSet `RollingUpdate` operations.
Signed-off-by: Ionut Balutoiu <ibalutoiu@cloudbasesolutions.com>
Some node e2e tests check for expected number of pods running
on the node to verify the correct state of that node after running
test scenarios. An example of such a check is in the device plugin
end to end test here: [1].
If the node is not left in a clean state after an e2e test finishes
running, it can lead to flaky tests because the node might have
unexpected pods running on the node.
In order to avoid that, we make sure that the test pods are
cleaned up after the test runs.
[1]: https://github.com/kubernetes/kubernetes/blob/master/test/e2e_node/device_plugin_test.go#L189-L190
Signed-off-by: Swati Sehgal <swsehgal@redhat.com>
wait.Until catches panics and logs them, which leads to confusing
output. Besides, the test is written so that failures must get reported to the
main goroutine.
Looking up the expected nodes in the goroutine raced with the test making
changes to the configuration. When doing (unrelated?) changes, the test started
to fail:
Oct 23 15:47:03.092: INFO: Unexpected error:
<*errors.errorString | 0xc001154c70>: {
s: "no subset of available IP address found for the endpoint test-rolling-update-with-lb within timeout 2m0s",
}
Oct 23 15:47:03.092: FAIL: no subset of available IP address found for the endpoint test-rolling-update-with-lb within timeout 2m0s
Now that everything is connected to a per-test context, the gRPC server might
encounter an error before it gets shut down normally. We must not panic in that
case because it would kill the entire Ginkgo worker process. This is not even
an error, so just log it as info message.
The wrapper can be used in combination with ginkgo.DeferCleanup to ignore
harmless "not found" errors during delete operations.
Original code suggested by Onsi Fakhouri.
It is set in all of the test/e2e* suites, but not in the ginkgo output
tests. This check is needed before adding a test case there which would trigger
this nil pointer access.
Adding the "context" import in the previous commit must get compensated by
removing one of the blank lines in the output unit tests, otherwise the stack
backtrace don't match expectations.
Adding "ctx" as parameter in the previous commit led to some linter errors
about code that overwrites "ctx" without using it.
This gets fixed by replacing context.Background or context.TODO in those code
lines with the new ctx parameter.
Two context.WithCancel calls can get removed completely because the context
automatically gets cancelled by Ginkgo when the test returns.
Every ginkgo callback should return immediately when a timeout occurs or the
test run manually gets aborted with CTRL-C. To do that, they must take a ctx
parameter and pass it through to all code which might block.
This is a first automated step towards that: the additional parameter got added
with
sed -i 's/\(framework.ConformanceIt\|ginkgo.It\)\(.*\)func() {$/\1\2func(ctx context.Context) {/' \
$(git grep -l -e framework.ConformanceIt -e ginkgo.It )
$GOPATH/bin/goimports -w $(git status | grep modified: | sed -e 's/.* //')
log_test.go was left unchanged.
Endpoints generated by the endpoints controller are in the canonical
form, however, custom endpoints can not be in canonical format
(there was a time they were canonicalized in the apiserver, but this
caused performance issues because the endpoint controller kept
updating them since the created endpoint were different than the
stored one due to the canonicalization)
There are cases where a custom endpoint may generate multiple slices
due to the controller, per example, when the same address is present
in different subsets.
The endpointslice mirroring controller should canonicalize the
endpoints subsets before start processing them to be consistent
on the slices generated, there is no risk of hotlooping because
the endpoint is only used as input.
Change-Id: I2a8cd53c658a640aea559a88ce33e857fa98cc5c
This ensures that the daemonset controller updates daemonset statuses in
a best-effort manner even if syncDaemonSet fails.
In order to add an integration test, this also replaces
`cmd/kube-apiserver/app/testing.StartTestServer` with
`test/integration/framework.StartTestServer` and adds
`setupWithServerSetup` to configure the admission control of the
apiserver.
Currently, if user executes `kubectl scale --dry-run`, output has no
indicator showing that this is not applied in reality.
This PR adds dry run suffix to the output as well as more integration
tests to verify it.
`kubectl scale` calls visitor two times. Second call fails when
the piped input is passed by returning an
`error: no objects passed to scale` error.
This PR uses the result of first visitor and fixes that piped
input problem. In addition to that, this PR also adds new
scale test to verify.
`kubectl exec` command supports getting files as inputs. However,
if the file contains multiple resources, it returns unclear error message;
`cannot attach to *v1.List: selector for *v1.List not implemented`.
Since `exec` command does not support multi resources, this PR
handles that and returns descriptive error message earlier.
One of the cpumanager tests doesn't remove the pod
that got created during the test.
This causes pollution of other tests and failures
from time to time (depends on the test execution order).
In order to defalke the tests, we should delete the pod
and wait for it to be completely remove.
Signed-off-by: Talor Itzhak <titzhak@redhat.com>
This introduces `singularNameProvider`. This provider will be used
by core types to have their singular names are defined in discovery
endpoint. Thanks to that, core resources singular name always have
higher precedence than CRDs shortcuts or singular names.
This adds new integration tests to test shortnames and
singular names are expanding to correct resources.
In this case, core types have always higher precendence than
CRDs.
This change will leverage the new PreFilterResult
to reduce down the list of eligible nodes for pod
using Bound Local PVs during PreFilter stage so
that only the node(s) which local PV node affinity
matches will be cosnidered in subsequent scheduling
stages.
Today, the NodeAffinity check is done during Filter
which means all nodes will be considered even though
there may be a large number of nodes that are not
eligible due to not matching the pod's bound local
PV(s)' node affinity requirement. Here we can
reduce down the node list in PreFilter to ensure that
during Filter we are only considering the reduced
list and thus can provide a more clear message to
users when node(s) are not available for scheduling
since the list only contains relevant nodes.
If error is encountered (e.g. PV cache read error) or
if node list reduction cannot be done (e.g. pod uses
no local PVs), then we will still proceed to consider
all nodes for the rest of scheduling stages.
Signed-off-by: Yibo Zhuang <yibzhuang@gmail.com>
In the Dynamic Resource allocation example specs, the claim
parameter name specified was inconsistent.
This commit fixes that with a better/more consistent name,
which is used to define the configmap and referenced in
the `ResourceClaimTemplate` spec.
Signed-off-by: Swati Sehgal <swsehgal@redhat.com>
A recent PR [1] updated the image versions we use for E2E tests. However, the ``windows-nanoserver`` image is meant to be in a private authenticated registry: ``gcr.io/authenticated-image-pulling/windows-nanoserver``, which requires credentials to pull images from it. This image is required by the ``[sig-node] Container Runtime blackbox test when running a container with a new image should be able to pull from private registry with secret [NodeConformance]`` test for Windows. The ``v3`` image does not exist, there's no automatic promotion process for that registry. Previously, it was built and pushed manually.
Because of this, the https://testgrid.k8s.io/sig-windows-signal#capz-windows-containerd-master jobs have started to fail.
Reverts the image version to ``v1``.
[1] https://github.com/kubernetes/kubernetes/pull/113900
Many clusters block direct requests from internal resources to the nodes
external IPs as best practice. All accesses from internal resources that
want to access resources running on nodes go through load balancers,
nodes being on private or public subnets. Let's prefer internal IPs
first, so the tests can work even when there are security group rules
present blocking requests to the external IPs.
We should not require ExternalIP for Conformance, but should keep
testing ExternalIPs in sig network.
Signed-off-by: Rafael Fonseca <r4f4rfs@gmail.com>
These instructions bring up a kind cluster with containerd 34d078e99, the
latest commit from the main branch. This version of containerd has
support for CDI.
The driver can be used manually against a cluster started with
local-up-cluster.sh and is also used for E2E testing. Because the tests proxy
connections from the nodes into the e2e.test binary and create/delete files via
the equivalent of "kubectl exec dd/rm", they can be run against arbitrary
clusters. Each test gets its own driver instance and resource class, therefore
they can run in parallel.
Add volumePath parameter to all disruptive checks, so subpath tests can use
"/test-volume" and disruptive tests can use "/mnt/volume1" for their
respective Pods.
This adds a new resource.k8s.io API group with v1alpha1 as version. It contains
four new types: resource.ResourceClaim, resource.ResourceClass, resource.ResourceClaimTemplate, and
resource.PodScheduling.
This removes WaitTimeoutForPodNoLongerRunningOrNotFoundInNamespace
introduced in f2b9479f8e and changes
the test to use goroutines to speed up the cleanups.
Most CI jobs run an OS that does not support SELinux, therefore tests that
need it should be skipped by default.
* [Feature:SELinux] marks tests that need SELinux (for any feature)
* [Feature:SELinuxMountReadWriteOncePod] marks tests that need
SELinuxMountReadWriteOncePod alpha gate enabled.
Currently, all SELinux tests have both, but it will change in the future.
Also make some design changes exposed in testing and review.
Do not remove the ambiguous old metric
`apiserver_flowcontrol_request_concurrency_limit` because reviewers
though it is too early. This creates a problem, that metric can not
keep both of its old meanings. I chose the configured concurrency
limit.
Testing has revealed a design flaw, which concerns the initialization
of the seat demand state tracking. The current design in the KEP is
as follows.
> Adjustment is also done on configuration change … For a newly
> introduced priority level, we set HighSeatDemand, AvgSeatDemand, and
> SmoothSeatDemand to NominalCL-LendableSD/2 and StDevSeatDemand to
> zero.
But this does not work out well at server startup. As part of its
construction, the APF controller does a configuration change with zero
objects read, to initialize its request-handling state. As always,
the two mandatory priority levels are implicitly added whenever they
are not read. So this initial reconfig has one non-exempt priority
level, the mandatory one called catch-all --- and it gets its
SmoothSeatDemand initialized to the whole server concurrency limit.
From there it decays slowly, as per the regular design. So for a
fairly long time, it appears to have a high demand and competes
strongly with the other priority levels. Its Target is higher than
all the others, once they start to show up. It properly gets a low
NominalCL once other levels show up, which actually makes it compete
harder for borrowing: it has an exceptionally high Target and a rather
low NominalCL.
I have considered the following fix. The idea is that the designed
initialization is not appropriate before all the default objects are
read. So the fix is to have a mode bit in the controller. In the
initial state, those seat demand tracking variables are set to zero.
Once the config-producing controller detects that all the default
objects are pre-existing, it flips the mode bit. In the later mode,
the seat demand tracking variables are initialized as originally
designed.
However, that still gives preferential treatment to the default
PriorityLevelConfiguration objects, over any that may be added later.
So I have made a universal and simpler fix: always initialize those
seat demand tracking variables to zero. Even if a lot of load shows
up quickly, remember that adjustments are frequent (every 10 sec) and
the very next one will fully respond to that load.
Also: revise logging logic, to log at numerically lower V level when
there is a change.
Also: bug fix in float64close.
Also, separate imports in some file
Co-authored-by: Han Kang <hankang@google.com>
This change enables hot reload of encryption config file when api server
flag --encryption-provider-config-automatic-reload is set to true. This
allows the user to change the encryption config file without restarting
kube-apiserver. The change is detected by polling the file and is done
by using fsnotify watcher. When file is updated it's process to generate
new set of transformers and close the old ones.
Signed-off-by: Nilekh Chaudhari <1626598+nilekhc@users.noreply.github.com>
