xintongsong commented on a change in pull request #11353: [FLINK-16438][yarn]
Make YarnResourceManager starts workers using WorkerResourceSpec requested by
SlotManager
URL: https://github.com/apache/flink/pull/11353#discussion_r403831354
##########
File path:
flink-kubernetes/src/main/java/org/apache/flink/kubernetes/KubernetesResourceManager.java
##########
@@ -232,57 +229,75 @@ private void recoverWorkerNodesFromPreviousAttempts()
throws ResourceManagerExce
++currentMaxAttemptId);
}
- private void requestKubernetesPod() {
- numPendingPodRequests++;
+ private void requestKubernetesPod(WorkerResourceSpec
workerResourceSpec) {
+ final KubernetesTaskManagerParameters parameters =
+
createKubernetesTaskManagerParameters(workerResourceSpec);
+
+ podWorkerResources.put(parameters.getPodName(),
workerResourceSpec);
+ final int pendingWorkerNum =
pendingWorkerCounter.increaseAndGet(workerResourceSpec);
log.info("Requesting new TaskManager pod with <{},{}>. Number
pending requests {}.",
- defaultMemoryMB,
- defaultCpus,
- numPendingPodRequests);
+ parameters.getTaskManagerMemoryMB(),
+ parameters.getTaskManagerCPU(),
+ pendingWorkerNum);
+ log.info("TaskManager {} will be started with {}.",
parameters.getPodName(), workerResourceSpec);
+
+ final KubernetesPod taskManagerPod =
+
KubernetesTaskManagerFactory.createTaskManagerComponent(parameters);
+ kubeClient.createTaskManagerPod(taskManagerPod);
+ }
+
+ private KubernetesTaskManagerParameters
createKubernetesTaskManagerParameters(WorkerResourceSpec workerResourceSpec) {
+ // TODO: need to unset process/flink memory size from
configuration if dynamic worker resource is activated
Review comment:
The purpose of this `TODO` is to avoid conflict between `workerResourceSpec`
and `process.size`/`flink.size` in `flinkConfig`. It is not about forbidden
workers with different sizes. Actually, one of the main purpose of this PR is
to make the RMs not assuming workers have the same size.
I was thinking about change the `Configuration` at creating the
`SlotManager`, where we known which implementation of `SlotManager` is used and
unset flink/process size if the plugin supports dynamic worker sizes.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
