Migrating from v5.6 to v5.7
Fabric8 Kubernetes Client upgraded to v8.0.0
v5.7 uses version 8.0.0 of the Fabric8 Kubernetes Client. Please consult the client’s own release notes and migration guide for the complete list of changes. The ones that are most likely to affect operators built with JOSDK are described below.
Jackson 3
The client moved from Jackson 2 to Jackson 3. The packages changed from com.fasterxml.jackson to
tools.jackson (the annotations from jackson-annotations are the exception and remain in
com.fasterxml.jackson.annotation). If your operator uses Jackson types directly, for example
JsonNode obtained from the client’s KubernetesSerialization, update the imports and the
dependencies accordingly:
// before
import com.fasterxml.jackson.databind.JsonNode;
// after
import tools.jackson.databind.JsonNode;
Other notable Jackson 3 differences:
JsonProcessingExceptionwas replaced byJacksonException, which is unchecked.JsonNode.fields()was replaced byproperties(),fieldNames()bypropertyNames()andasText()byasString().ObjectMapperinstances are created using builders, for exampleYAMLMapper.builder().build().
Ambiguous withShardSelector(null)
Filterable now has an additional withShardSelector(ShardSelector) overload next to
withShardSelector(String). Calls passing a bare null no longer compile, cast the argument:
client.pods().withShardSelector((String) null);
Default HTTP client is now Vert.x 5
The operator-framework artifact now brings the Vert.x 5 based HTTP client of the Fabric8 client
(kubernetes-httpclient-vertx-5), which is also the default of the Fabric8 client itself, instead of
the JDK one (kubernetes-httpclient-jdk). If your operator doesn’t declare an HTTP client
implementation of its own there is nothing to do, it gets the new default. To keep using the JDK
client, or to use another implementation, exclude the default from the operator-framework
dependency and add the one you want:
<dependency>
<groupId>io.javaoperatorsdk</groupId>
<artifactId>operator-framework</artifactId>
<exclusions>
<exclusion>
<groupId>io.fabric8</groupId>
<artifactId>kubernetes-httpclient-vertx-5</artifactId>
</exclusion>
</exclusions>
</dependency>
<dependency>
<groupId>io.fabric8</groupId>
<artifactId>kubernetes-httpclient-jdk</artifactId>
<version>8.0.0</version>
</dependency>
The implementations provided as part of the Fabric8 client all follow the
kubernetes-httpclient-<implementation name> pattern for their artifact identifier.
If your operator already replaced the default HTTP client, update the exclusion: excluding
kubernetes-httpclient-jdk no longer has any effect and kubernetes-httpclient-vertx-5 would end up
on the classpath next to your implementation. This is especially important if you use the Vert.x 4
based client (kubernetes-httpclient-vertx), since Vert.x 4 and Vert.x 5 cannot be used together.
If you want to keep the Vert.x 4 based client, use it in place of kubernetes-httpclient-jdk in the
snippet above.
Executors of the polling and timer event sources
The polling event sources and the internal TimerEventSource no longer use a java.util.Timer, and
PerResourcePollingEventSource no longer creates its executor when its configuration is built. They
run on scheduled executors, see virtual threads
and Threading of the polling event sources.
This is transparent unless your operator provides an executor to a polling event source, or relies
on one of the following:
- A
ScheduledExecutorServiceprovided withwithExecutorService(...)to aPerResourcePollingConfigurationBuilderwas shut down withshutdownNow()when the event source was stopped. It is not anymore: an executor you provide is yours to shut down, like it is forPollingConfigurationBuilder. If you relied on the event source for this, shut it down yourself after stopping the operator, otherwise its threads can keep the JVM from exiting if they are not daemon threads. PerResourcePollingConfiguration.executorService()returnsnullwhen no executor was provided, it used to be aScheduledThreadPoolExecutorwith a single non-daemon thread created by the configuration. The event source now creates a single daemon thread executor, or one running on virtual threads ifuseVirtualThreadsis enabled, when it starts, shuts it down when it stops, and creates a new one if it is started again.PerResourcePollingConfiguration.DEFAULT_EXECUTOR_THREAD_NUMBERis deprecated, it isn’t used anymore.- Each polling event source has an executor of its own: they don’t share threads, and the number of threads of a source is not configurable other than by providing an executor.
new PollingEventSource<>(Class, PollingConfiguration)is deprecated. Use the constructor that takes theEventSourceContextas well,new PollingEventSource<>(Class, EventSourceContext, PollingConfiguration), otherwise the event source cannot know about theuseVirtualThreadssetting and polls on a platform thread.TimerEventSource(String, boolean)is deprecated, use the constructor that also takes an executor supplier,nullto have the event source create an executor of its own.ExecutorServiceManager.scheduledExecutorService()is deprecated. The framework doesn’t use it anymore, it creates the executor on first use, and shuts it down when the manager is stopped.
Event recorder: events about any object
The event recorder can now record events about objects other than the primary resource, with
context.eventRecorder(otherObject), see
Recording Kubernetes Events. Reconcilers that only use
context.eventRecorder() are not affected. The low level API, which is experimental, changed to
make this possible:
- A custom
EventRecorderimplementsrecord(HasMetadata regarding, EventRecord event, Context<?> context)instead ofrecord(EventRecord, Context<?>)andforContext(Context<?>). Those two are now default methods that delegate to the new one, as is the newforResource(HasMetadata, Context<?>). - A subclass of
DefaultEventRecorderoverrides or callstoEvent(HasMetadata regarding, EventRecord record, Context<?> context)instead oftoEvent(Context<?>, EventRecord). An override of the old signature without@Overridestill compiles but is no longer called.
Feedback
Was this page helpful?
Glad to hear it! Please tell us how we can improve.
Sorry to hear that. Please tell us how we can improve.