[
https://issues.apache.org/jira/browse/FLINK-9455?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16628951#comment-16628951
]
ASF GitHub Bot commented on FLINK-9455:
---------------------------------------
tillrohrmann commented on a change in pull request #6734: [FLINK-9455][RM] Add
support for multi task slot TaskExecutors
URL: https://github.com/apache/flink/pull/6734#discussion_r220614717
##########
File path:
flink-runtime/src/test/java/org/apache/flink/runtime/resourcemanager/slotmanager/SlotManagerTest.java
##########
@@ -1398,4 +1389,171 @@ private SlotManager
createSlotManager(ResourceManagerId resourceManagerId, Resou
return slotManager;
}
+
+ /**
+ * Tests that we only request new resources/containers once we have
assigned
+ * all pending task manager slots.
+ */
+ @Test
+ public void testRequestNewResources() throws Exception {
+ final int numberSlots = 2;
+ final AtomicInteger resourceRequests = new AtomicInteger(0);
+ final TestingResourceActions testingResourceActions = new
TestingResourceActionsBuilder()
+ .setAllocateResourceFunction(
+ convert(ignored -> {
+ resourceRequests.incrementAndGet();
+ return numberSlots;
+ }))
+ .build();
+
+ try (final SlotManager slotManager = createSlotManager(
+ ResourceManagerId.generate(),
+ testingResourceActions)) {
+
+ final JobID jobId = new JobID();
+
assertThat(slotManager.registerSlotRequest(createSlotRequest(jobId)), is(true));
+ assertThat(resourceRequests.get(), is(1));
+
+ // the second slot request should not try to allocate a
new resource because the
+ // previous resource was started with 2 slots.
+
assertThat(slotManager.registerSlotRequest(createSlotRequest(jobId)), is(true));
+ assertThat(resourceRequests.get(), is(1));
+
+
assertThat(slotManager.getNumberAssignedPendingTaskManagerSlots(), is(2));
+
+
assertThat(slotManager.registerSlotRequest(createSlotRequest(jobId)), is(true));
+ assertThat(resourceRequests.get(), is(2));
+ }
+ }
+
+ /**
+ * Tests that a failing allocation/slot request will return the pending
task manager slot.
+ */
+ @Test
+ public void testFailingAllocationReturnsPendingTaskManagerSlot() throws
Exception {
+ final int numberSlots = 2;
+ final TestingResourceActions resourceActions = new
TestingResourceActionsBuilder()
+ .setAllocateResourceFunction(convert(value ->
numberSlots))
+ .build();
+ try (final SlotManager slotManager =
createSlotManager(ResourceManagerId.generate(), resourceActions)) {
+ final JobID jobId = new JobID();
+
+ final SlotRequest slotRequest =
createSlotRequest(jobId);
+
assertThat(slotManager.registerSlotRequest(slotRequest), is(true));
+
+
assertThat(slotManager.getNumberPendingTaskManagerSlots(), is(numberSlots));
+
assertThat(slotManager.getNumberAssignedPendingTaskManagerSlots(), is(1));
+
+
slotManager.unregisterSlotRequest(slotRequest.getAllocationId());
+
+
assertThat(slotManager.getNumberPendingTaskManagerSlots(), is(numberSlots));
+
assertThat(slotManager.getNumberAssignedPendingTaskManagerSlots(), is(0));
+ }
+ }
+
+ /**
+ * Tests the completion of pending task manager slots by registering a
TaskExecutor.
+ */
+ @Test
+ public void testPendingTaskManagerSlotCompletion() throws Exception {
+ final int numberSlots = 3;
+ final TestingResourceActions resourceActions = new
TestingResourceActionsBuilder()
+ .setAllocateResourceFunction(convert(value ->
numberSlots))
+ .build();
+
+ try (final SlotManager slotManager =
createSlotManager(ResourceManagerId.generate(), resourceActions)) {
+ final JobID jobId = new JobID();
+
assertThat(slotManager.registerSlotRequest(createSlotRequest(jobId)), is(true));
+
+
assertThat(slotManager.getNumberPendingTaskManagerSlots(), is(numberSlots));
+
assertThat(slotManager.getNumberAssignedPendingTaskManagerSlots(), is(1));
+ assertThat(slotManager.getNumberRegisteredSlots(),
is(0));
+
+ final TaskExecutorConnection taskExecutorConnection =
createTaskExecutorConnection();
+ final SlotReport slotReport =
createSlotReport(taskExecutorConnection.getResourceID(), numberSlots - 1);
+
+ slotManager.registerTaskManager(taskExecutorConnection,
slotReport);
+
+ assertThat(slotManager.getNumberRegisteredSlots(),
is(numberSlots - 1));
+
assertThat(slotManager.getNumberPendingTaskManagerSlots(), is(1));
Review comment:
I think `getNumberPendingTaskManagerSlots == 1` is an even stricter
condition, because it requires that actually some of the pending task manager
slots have been completed.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Make SlotManager aware of multi slot TaskManagers
> -------------------------------------------------
>
> Key: FLINK-9455
> URL: https://issues.apache.org/jira/browse/FLINK-9455
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Coordination, ResourceManager
> Affects Versions: 1.5.0
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.7.0
>
>
> The {{SlotManager}} responsible for managing all available slots of a Flink
> cluster can request to start new {{TaskManagers}} if it cannot fulfill a slot
> request. The started {{TaskManager}} can be started with multiple slots
> configured but currently, the {{SlotManager}} thinks that it will be started
> with a single slot. As a consequence, it might issue multiple requests to
> start new TaskManagers even though a single one would be sufficient to
> fulfill all pending slot requests.
> In order to avoid requesting unnecessary resources which are freed after the
> idle timeout, I suggest to make the {{SlotManager}} aware of how many slots a
> {{TaskManager}} is started with. That way the SlotManager only needs to
> request a new {{TaskManager}} if all of the previously started slots
> (potentially not yet registered and, thus, future slots) are being assigned
> to slot requests.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
