6.3.0 (2022-12-12)

[manusa]

Bugs

• Fix refactor: extracted PortForwarderWebsocketListener and added tests #4159: ensure the token refresh obeys how the Config was created
• Fix KnativeClient.isSupported doesn't check all of the applicable API Groups #4447: isSupported doesn't check all of the applicable API Groups
• Fix fix #4365 correcting backoff interval #4473: correcting backoff interval regression introduced in feat: add possibility to add an exception handler to Watchers #4365 (6.2.0)
• Fix RejectedExecutionException when checking status of removed operator through BaseOperation.get #4491: added a more explicit shutdown exception for okhttp
• Fix Resource leak when creating private keys #4509: do not reuse KeyFactory instance after a failure
• Fix Stack overflow on recursive types (again) #4510: Fix StackOverflow on cyclic references involving collections
• Fix Null pointer exception executing java-generator-cli #4534: Java Generator CLI default handling of skipGeneratedAnnotations
• Fix File upload unusual behaviour on a ' character #4535: The shell command string will now have single quotes sanitized
• Fix Additional Attributes are generated to be ignored, because of wrong setting of JsonAnySetter annotation #4543: (Java Generator) additionalProperties JsonAny setter method generated as setAdditionalProperty
• Fix Deserializing Kube and custom resources from the same template #4590: only using a builder if there are visitors
• Fix Builder methods expecting HasMetadata/KubernetesResource are not correct #4540: treating GenericKubernetesResource and RawExtension as buildable
• Fix Timing issue with leader elector cancel (stop) #4547: preventing timing issues with leader election cancel
• Fix Fabric8 builds JdkHttpClientImpl with 0 as timeout, and that leads to "Invalid duration: PT0S" #4569: fixing jdk httpclient regression with 0 timeouts
• Fix [crd-generator] Fix float support #4581: "float" types in spec for CRD generator are supported
• Fix Inconsistent additionalPrinterColumns jsonPath #4610: inconsistent additionalPrinterColumns jsonPath
• Fix Integration Test Failures #4641: fixed regression with missing initial watch event

Improvements

• Fix Remove the need for the openshift-client dependency #4014: added support for OpenShift Build log version.
• Fix [refactor] Change JettyHttpClient#sendAsync method to not read full response in memory #4201: Removed sendAsync from the individual http client implementations
• Fix exec and other operations should support a default container #4355: for exec, attach, upload, and copy operations the container id/name will be validated or chosen prior to the remote call. You may also use the kubectl.kubernetes.io/default-container annotation to specify the default container.
• Fix Primitive default values in specifications break the object instantiation #4530: generalizing the Serialization logic to allow for primitive values and clarifying the type expectations.
• Fix Api methods related to patch generation #4363: exposed ResourceCompare.metadataChanged

New Features

• Fix Generally support fieldValidation #4136: added support for fieldValidation as a dsl method for POST/PUT/PATCH operations
• Fix CreateOrReplace not fully equivalent with apply #3896: added dsl support for server side apply
• Fix updated [client.secrets] createOrReplace #4582: updated [client.secrets] createOrReplace document
• Fix Replacement for deleteExisting behavior. #4516: added support for blocking delete operations using the withTimeout methods: op.withTimeout(1, TimeUnit.MINUTE).delete() - will wait for up to 1 minute for the resources to be fully deleted. This makes for a more concise replacement of the deletingExisting method.

Note: Breaking changes

• Fix Remove the need for KubernetesResourceMappingProvider #3923: removed KubernetesResourceMappingProvider - a META-INF/services/io.fabric8.kubernetes.api.model.KubernetesResource list of resources is used instead.
• Fix Remove model properties #4515: files located at the root of jars named model.properties, e.g. core.properties, have been removed
• Fix Removing internal usage of registerCustomKind #4579: the implicit registration of resource and list types that happens when using the resource(class) methods has been removed. This makes the behavior of the client more predictable as that was an undocumented side-effect. If you expect to see instances of a custom type from an untyped api call - typically KubernetesClient.load, KubernetesClient.resourceList, KubernetesClient.resource(InputStream|String), then you must either create a META-INF/services/io.fabric8.kubernetes.api.model.KubernetesResource file (see above Remove the need for KubernetesResourceMappingProvider #3923), or make calls to KubernetesDeserializer.registerCustomKind - however since KubernetesDeserializer is an internal class that mechanism is not preferred.
• Fix [crd-generator] Remove deprecated support for NotNull #4597: remove the deprecated support for javax.validation.constraints.NotNull in the crd-generator, to mark a property as required it needs to be annotated with io.fabric8.generator.annotation.Required
• Fix Api methods related to patch generation #4363: deprecated existing ResourceCompare methods such as compareKubernetesResource as they are not for general use

Full Changelog: v6.2.0...v6.3.0

---

This discussion was created from the release 6.3.0 (2022-12-12).

[scholzj]

@manusa After updating Strimzi to use 6.3.0, I'm getting a lot of warnings about OkHttp breponse bodies not being closed:

2022-12-12 20:10:17 WARN  OkHttpClient:134 - A connection to https://10.96.0.1/ was leaked. Did you forget to close a response body?
java.lang.Throwable: response.body().close()
	at okhttp3.internal.platform.Platform.getStackTraceForCloseable(Platform.java:148) ~[com.squareup.okhttp3.okhttp-3.12.12.jar:?]
	at okhttp3.RealCall.captureCallStackTrace(RealCall.java:116) ~[com.squareup.okhttp3.okhttp-3.12.12.jar:?]
	at okhttp3.RealCall.enqueue(RealCall.java:125) ~[com.squareup.okhttp3.okhttp-3.12.12.jar:?]
	at io.fabric8.kubernetes.client.okhttp.OkHttpClientImpl.sendAsync(OkHttpClientImpl.java:266) ~[io.fabric8.kubernetes-httpclient-okhttp-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.okhttp.OkHttpClientImpl.consumeBytesDirect(OkHttpClientImpl.java:322) ~[io.fabric8.kubernetes-httpclient-okhttp-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.http.StandardHttpClient.consumeBytes(StandardHttpClient.java:57) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.http.SendAsyncUtils.bytes(SendAsyncUtils.java:51) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.http.HttpResponse$SupportedResponses.sendAsync(HttpResponse.java:105) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.http.HttpClient.sendAsync(HttpClient.java:178) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.OperationSupport.retryWithExponentialBackoff(OperationSupport.java:618) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.OperationSupport.handleResponse(OperationSupport.java:595) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.OperationSupport.handleResponse(OperationSupport.java:570) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.OperationSupport.handleUpdate(OperationSupport.java:366) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.BaseOperation.handleUpdate(BaseOperation.java:709) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.HasMetadataOperation.lambda$replace$0(HasMetadataOperation.java:167) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.HasMetadataOperation.replace(HasMetadataOperation.java:172) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.HasMetadataOperation.replace(HasMetadataOperation.java:113) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.HasMetadataOperation.replace(HasMetadataOperation.java:41) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.BaseOperation.replace(BaseOperation.java:1077) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.dsl.internal.BaseOperation.replace(BaseOperation.java:93) ~[io.fabric8.kubernetes-client-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.extended.leaderelection.resourcelock.LeaseLock.update(LeaseLock.java:100) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.extended.leaderelection.LeaderElector.tryAcquireOrRenew(LeaderElector.java:214) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.extended.leaderelection.LeaderElector.lambda$renewWithTimeout$6(LeaderElector.java:175) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at io.fabric8.kubernetes.client.extended.leaderelection.LeaderElector.lambda$loop$8(LeaderElector.java:258) ~[io.fabric8.kubernetes-client-api-6.3.0.jar:?]
	at java.util.concurrent.CompletableFuture$AsyncRun.run(CompletableFuture.java:1804) ~[?:?]
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136) ~[?:?]
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635) ~[?:?]
	at java.lang.Thread.run(Thread.java:833) ~[?:?]

Is this a bug in the new Fabric8 release? As we do not really touch the OkHttp client, I would assume regardless of what Strimzi does, it would be Fabric8's responsibility to close the response bodies?

(I was not getting this issue with 6.2.0)

[scholzj]

PS: I guess I can open an issue for it, or?

[shawkins]

Right. But these seem to come from fairly regular calls which IMHO don't return any input streams.

Yes, I just mentioned that for completeness.

Yeah, it shows only after 5 minutes ... so it is quite tricky to notice.

Also this was a latent bug - it was already in 6.x releases prior to 6.3, it was just uncommon to use the okhttp httpclient in this way.

Out of curiosity ... is there now some recommended HTTP client to use? We use OkHttp more or less because it used to be the one used by Fabric8 out of the box. We do not have any other reason to use it.

If you can use JDK 16+, the JDK httpclient should be a better option. It also appears that we're aligning on supplying a vertx client with 6.4.

PS: I guess I can open an issue for it, or?

Yes, I have the fix ready.

[scholzj]

Ok, I opened #4666 for this. Thanks for looking into it.

If you can use JDK 16+, the JDK httpclient should be a better option. It also appears that we're aligning on supplying a vertx client with 6.4.

Actually, we are more or less moving away from Vert.x. But we just moved to Java 17. So the JDK client might be an option for us now. I will give it a try. Thanks.

Also this was a latent bug - it was already in 6.x releases prior to 6.3, it was just uncommon to use the okhttp httpclient in this way.

TBH, we were not getting the warnings with previous 6.x versions.

[shawkins]

Ok, I opened #4666 for this. Thanks for looking into it.

No problem, the gist of the fix is that we're missing this close call: https://github.com/fabric8io/kubernetes-client/pull/4665/files#diff-7779101a4a865555fc1bccf9250b2c814359527cfc2f37118c3d97a1d6a2ff8bR132

So the JDK client might be an option for us now. I will give it a try. Thanks.

That has the added benefit of not using 2 threads per watch / informer and won't have the possible max request issue mentioned on #4663 which you guys probably also hit.

TBH, we were not getting the warnings with previous 6.x versions.

It would have occurred only for very specific streaming cases (I can't recall exactly off of my head which ones in 6.2 were based asyncbody), and not for sendAsync calls that returned String, byte[], InputStream, or Readers - that's where you are seeing this now.

[manusa]

continues here: #4666 (comment)