I m trying to use Flink Kubernetes Operator with Flink 1 18 Apache Flink #random

I'm trying to use Flink Kubernetes Operator with F...

Vitor Leal

10/26/2023, 12:46 PM

I'm trying to use Flink Kubernetes Operator with Flink 1.18 but when submitting a deployment file I get

Copy code

The FlinkDeployment "app-deployment" is invalid: spec.flinkVersion: Unsupported value: "v1_18": supported values: "v1_13", "v1_14", "v1_15", "v1_16", "v1_17

I am running version 1.6.0 of Flink K8 operator:

Copy code

[INFO] Starting Flink Kubernetes Operator (Version: 1.6.0, Flink Version: 1.17.1, Rev:ebb1fed, Date:2023-08-10T14:33:52+02:00)
[INFO] OS current user: flink
[INFO] JVM: OpenJDK 64-Bit Server VM - Eclipse Adoptium - 11/11.0.20+8
[INFO] Arch: amd64
[INFO] Maximum heap size: 247 MiBytes
[INFO] JAVA_HOME: /opt/java/openjdk
[INFO] JVM Options:
[INFO]    -Dlog4j.configurationFile=/opt/flink/conf/log4j-operator.properties
[INFO] Program Arguments: (none)
[INFO] Classpath: ./flink-kubernetes-standalone-1.6.0.jar:./flink-kubernetes-operator-1.6.0-shaded.jar:./flink-kubernetes-operator-autoscaler-1.6.0.jar:/opt/flink/operator-lib/*

am i missing something?

Martijn Visser

10/26/2023, 12:47 PM

This component is compatible with Apache Flink version(s):

1.17.1

1.16.2

1.15.4

1.14.6

1.13.6

It’s not compatible with Flink 1.18 https://flink.apache.org/downloads/

Vitor Leal

10/26/2023, 12:49 PM

I'm confused. I am actually able to deploy this https://github.com/apache/flink-kubernetes-operator/blob/main/examples/autoscaling/autoscaling.yaml locally. it is also running 1.6.0:

Copy code

2023-10-26 12:31:06,002 o.a.f.k.o.FlinkOperator        [INFO ] --------------------------------------------------------------------------------
2023-10-26 12:31:06,025 o.a.f.k.o.FlinkOperator        [INFO ]  Starting Flink Kubernetes Operator (Version: 1.6.0, Flink Version: 1.17.1, Rev:ebb1fed, Date:2023-08-10T14:33:52+02:00)
2023-10-26 12:31:06,026 o.a.f.k.o.FlinkOperator        [INFO ]  OS current user: flink
2023-10-26 12:31:06,026 o.a.f.k.o.FlinkOperator        [INFO ]  JVM: OpenJDK 64-Bit Server VM - Eclipse Adoptium - 11/11.0.20+8
2023-10-26 12:31:06,026 o.a.f.k.o.FlinkOperator        [INFO ]  Arch: aarch64
2023-10-26 12:31:06,033 o.a.f.k.o.FlinkOperator        [INFO ]  Maximum heap size: 7512 MiBytes
2023-10-26 12:31:06,034 o.a.f.k.o.FlinkOperator        [INFO ]  JAVA_HOME: /opt/java/openjdk
2023-10-26 12:31:06,034 o.a.f.k.o.FlinkOperator        [INFO ]  JVM Options:
2023-10-26 12:31:06,034 o.a.f.k.o.FlinkOperator        [INFO ]     -Dlog4j.configurationFile=/opt/flink/conf/log4j-operator.properties
2023-10-26 12:31:06,034 o.a.f.k.o.FlinkOperator        [INFO ]  Program Arguments: (none)
2023-10-26 12:31:06,034 o.a.f.k.o.FlinkOperator        [INFO ]  Classpath: ./flink-kubernetes-standalone-1.6.0.jar:./flink-kubernetes-operator-1.6.0-shaded.jar:./flink-kubernetes-operator-autoscaler-1.6.0.jar:/opt/flink/operator-lib/*

the only difference is the Arch

Vitor Leal

10/26/2023, 12:50 PM

but when trying to deploy this autoscaling.yaml example in our K8 Dev cluster, which has the operator installed with the same version, the error appears

Vitor Leal

10/26/2023, 12:51 PM

also the flink k8 operator docs for 1.6.0 mention experimental support for 1.18? https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-release-1.6/docs/custom-resource/autoscaler/#flink-118-and-in-[…]e-scaling-support

Martijn Visser

10/26/2023, 12:52 PM

That’s kind of weird

Vitor Leal

10/26/2023, 1:04 PM

Yeah... i am really confused 😕

Martijn Visser

10/26/2023, 1:04 PM

Did you update the CRD?

Vitor Leal

10/26/2023, 1:12 PM

locally:

Copy code

❯ kubectl config use-context docker-desktop
❯ kubectl logs flink-kubernetes-operator-99d9b5698-dmnfk -n default | head -n 20

Starting Operator
WARNING: sun.reflect.Reflection.getCallerClass is not supported. This will impact performance.
2023-10-26 12:42:08,423 o.a.f.k.o.FlinkOperator        [INFO ] --------------------------------------------------------------------------------
2023-10-26 12:42:08,448 o.a.f.k.o.FlinkOperator        [INFO ]  Starting Flink Kubernetes Operator (Version: 1.5.0, Flink Version: 1.16.1, Rev:be07be7, Date:2023-05-12T16:56:02+02:00)
2023-10-26 12:42:08,448 o.a.f.k.o.FlinkOperator        [INFO ]  OS current user: flink
2023-10-26 12:42:08,448 o.a.f.k.o.FlinkOperator        [INFO ]  JVM: OpenJDK 64-Bit Server VM - Eclipse Adoptium - 11/11.0.19+7
2023-10-26 12:42:08,448 o.a.f.k.o.FlinkOperator        [INFO ]  Arch: aarch64
2023-10-26 12:42:08,451 o.a.f.k.o.FlinkOperator        [INFO ]  Maximum heap size: 7512 MiBytes
2023-10-26 12:42:08,451 o.a.f.k.o.FlinkOperator        [INFO ]  JAVA_HOME: /opt/java/openjdk
2023-10-26 12:42:08,452 o.a.f.k.o.FlinkOperator        [INFO ]  JVM Options:
2023-10-26 12:42:08,452 o.a.f.k.o.FlinkOperator        [INFO ]     -Dlog4j.configurationFile=/opt/flink/conf/log4j-operator.properties
2023-10-26 12:42:08,452 o.a.f.k.o.FlinkOperator        [INFO ]  Program Arguments: (none)
2023-10-26 12:42:08,452 o.a.f.k.o.FlinkOperator        [INFO ]  Classpath: ./flink-kubernetes-standalone-1.5.0-shaded.jar:./flink-kubernetes-operator-1.5.0-shaded.jar:./flink-kubernetes-operator-autoscaler-1.5.0.jar:/opt/flink/operator-lib/*
2023-10-26 12:42:08,452 o.a.f.k.o.FlinkOperator        [INFO ] --------------------------------------------------------------------------------
2023-10-26 12:42:08,643 o.a.f.k.o.c.FlinkConfigManager [INFO ] Setting default configuration to {parallelism.default=1, taskmanager.numberOfTaskSlots=1, kubernetes.operator.reconcile.interval=15 s, kubernetes.operator.metrics.reporter.slf4j.interval=5 MINUTE, kubernetes.operator.observer.progress-check.interval=5 s, kubernetes.operator.health.probe.enabled=true, kubernetes.operator.health.probe.port=8085, kubernetes.operator.metrics.reporter.slf4j.factory.class=org.apache.flink.metrics.slf4j.Slf4jReporterFactory}
2023-10-26 12:42:08,656 o.a.f.k.o.c.FlinkConfigManager [INFO ] Enabled dynamic config updates, checking config changes every PT5M
2023-10-26 12:42:08,658 o.a.f.k.o.m.OperatorMetricUtils [INFO ] Initializing operator metrics using conf: {metrics.reporter.slf4j.factory.class=org.apache.flink.metrics.slf4j.Slf4jReporterFactory, metrics.reporter.slf4j.interval=5 MINUTE, parallelism.default=1, taskmanager.numberOfTaskSlots=1, kubernetes.operator.reconcile.interval=15 s, kubernetes.operator.observer.progress-check.interval=5 s, kubernetes.operator.health.probe.enabled=true, kubernetes.operator.health.probe.port=8085}
2023-10-26 12:42:08,701 o.a.f.r.m.MetricRegistryImpl   [INFO ] Periodically reporting metrics in intervals of 5 min for reporter slf4j of type org.apache.flink.metrics.slf4j.Slf4jReporter.
2023-10-26 12:42:09,062 o.a.f.k.o.FlinkOperator        [INFO ] Configuring operator with 200 reconciliation threads.
2023-10-26 12:42:09,063 o.a.f.k.o.FlinkOperator        [INFO ] Operator leader election is disabled.

in my dev cluster:

Copy code

❯ kubectl config use-context case-dev
❯ kubectl logs flink-kubernetes-operator-9c9bb864b-zqxx6 -n gspkafka | head -n 20

Starting Operator
WARNING: sun.reflect.Reflection.getCallerClass is not supported. This will impact performance.
2023-10-26 12:23:10,801 o.a.f.k.o.FlinkOperator        [INFO ] --------------------------------------------------------------------------------
2023-10-26 12:23:11,019 o.a.f.k.o.FlinkOperator        [INFO ]  Starting Flink Kubernetes Operator (Version: 1.6.0, Flink Version: 1.17.1, Rev:ebb1fed, Date:2023-08-10T14:33:52+02:00)
2023-10-26 12:23:11,021 o.a.f.k.o.FlinkOperator        [INFO ]  OS current user: flink
2023-10-26 12:23:11,096 o.a.f.k.o.FlinkOperator        [INFO ]  JVM: OpenJDK 64-Bit Server VM - Eclipse Adoptium - 11/11.0.20+8
2023-10-26 12:23:11,096 o.a.f.k.o.FlinkOperator        [INFO ]  Arch: amd64
2023-10-26 12:23:11,103 o.a.f.k.o.FlinkOperator        [INFO ]  Maximum heap size: 247 MiBytes
2023-10-26 12:23:11,107 o.a.f.k.o.FlinkOperator        [INFO ]  JAVA_HOME: /opt/java/openjdk
2023-10-26 12:23:11,108 o.a.f.k.o.FlinkOperator        [INFO ]  JVM Options:
2023-10-26 12:23:11,108 o.a.f.k.o.FlinkOperator        [INFO ]     -Dlog4j.configurationFile=/opt/flink/conf/log4j-operator.properties
2023-10-26 12:23:11,109 o.a.f.k.o.FlinkOperator        [INFO ]  Program Arguments: (none)
2023-10-26 12:23:11,109 o.a.f.k.o.FlinkOperator        [INFO ]  Classpath: ./flink-kubernetes-standalone-1.6.0.jar:./flink-kubernetes-operator-1.6.0-shaded.jar:./flink-kubernetes-operator-autoscaler-1.6.0.jar:/opt/flink/operator-lib/*
2023-10-26 12:23:11,109 o.a.f.k.o.FlinkOperator        [INFO ] --------------------------------------------------------------------------------
2023-10-26 12:23:11,922 o.a.f.k.o.c.FlinkConfigManager [INFO ] Setting default configuration to {job.autoscaler.restart.time=2m, job.autoscaler.enabled="true", job.autoscaler.stabilization.interval=1m, kubernetes.operator.watched.namespaces=gspkafka, kubernetes.operator.health.probe.enabled=true, job.autoscaler.catch-up.duration=5m, pipeline.max-parallelism="16", job.autoscaler.metrics.window=5m, job.autoscaler.target.utilization="0.6", parallelism.default=1, taskmanager.numberOfTaskSlots=1, fs.s3a.aws.credentials.provider=com.amazonaws.auth.WebIdentityTokenCredentialsProvider, kubernetes.operator.flink.client.timeout=61 s, kubernetes.operator.reconcile.interval=15 s, kubernetes.operator.metrics.reporter.slf4j.interval=5 MINUTE, kubernetes.operator.observer.progress-check.interval=5 s, job.autoscaler.target.utilization.boundary="0.2", kubernetes.operator.health.probe.port=8085, kubernetes.operator.metrics.reporter.slf4j.factory.class=org.apache.flink.metrics.slf4j.Slf4jReporterFactory}
2023-10-26 12:23:11,933 o.a.f.k.o.c.FlinkConfigManager [INFO ] Enabled dynamic config updates, checking config changes every PT5M
2023-10-26 12:23:11,998 o.a.f.k.o.m.OperatorMetricUtils [INFO ] Initializing operator metrics using conf: {job.autoscaler.restart.time=2m, job.autoscaler.enabled="true", job.autoscaler.stabilization.interval=1m, metrics.reporter.slf4j.interval=5 MINUTE, kubernetes.operator.watched.namespaces=gspkafka, kubernetes.operator.health.probe.enabled=true, job.autoscaler.catch-up.duration=5m, pipeline.max-parallelism="16", job.autoscaler.metrics.window=5m, job.autoscaler.target.utilization="0.6", metrics.reporter.slf4j.factory.class=org.apache.flink.metrics.slf4j.Slf4jReporterFactory, parallelism.default=1, taskmanager.numberOfTaskSlots=1, fs.s3a.aws.credentials.provider=com.amazonaws.auth.WebIdentityTokenCredentialsProvider, kubernetes.operator.flink.client.timeout=61 s, kubernetes.operator.reconcile.interval=15 s, kubernetes.operator.observer.progress-check.interval=5 s, job.autoscaler.target.utilization.boundary="0.2", kubernetes.operator.health.probe.port=8085}
2023-10-26 12:23:12,133 o.a.f.c.p.DefaultPluginManager [INFO ] Plugin loader with ID not found, creating it: flink-metrics-prometheus
2023-10-26 12:23:12,153 o.a.f.c.p.DefaultPluginManager [INFO ] Plugin loader with ID not found, creating it: flink-metrics-datadog
2023-10-26 12:23:12,154 o.a.f.c.p.DefaultPluginManager [INFO ] Plugin loader with ID not found, creating it: flink-metrics-slf4j

Vitor Leal

10/26/2023, 1:12 PM

@Martijn Visser sorry which custom resource do i need to look for?

Vitor Leal

10/26/2023, 1:16 PM

Copy code

```
❯ kubectl config use-context docker-desktop
❯ kubectl apply -f autoscaling.yaml -n default
<http://flinkdeployment.flink.apache.org/autoscaling-example2|flinkdeployment.flink.apache.org/autoscaling-example2> created

dev:

Copy code

❯ kubectl config use-context case-dev
❯ kubectl apply -f autoscaling.yaml -n gspkafka
The FlinkDeployment "autoscaling-example2" is invalid: spec.flinkVersion: Unsupported value: "v1_18": supported values: "v1_13", "v1_14", "v1_15", "v1_16", "v1_17"
```

Vitor Leal

10/26/2023, 1:53 PM

and this is the image it uses: https://hub.docker.com/layers/apache/flink-kubernetes-operator/ebb1fed/images/sha256[…]57d35885952eb38e688c611f293112c59cb3b78a0504b?context=explore

sap1ens

10/26/2023, 3:04 PM

https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/docs/operations/upgrade/#1-upgrading-the-crd

2 Views

Open in Slack

Previous Next