[
https://issues.apache.org/jira/browse/FLINK-5747?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15872366#comment-15872366
]
ASF GitHub Bot commented on FLINK-5747:
---------------------------------------
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/3295#discussion_r101829659
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/ExecutionGraph.java
---
@@ -754,6 +759,139 @@ public void scheduleForExecution(SlotProvider
slotProvider) throws JobException
}
}
+ private void scheduleLazy(SlotProvider slotProvider) throws
NoResourceAvailableException {
+ // simply take the vertices without inputs.
+ for (ExecutionJobVertex ejv : this.tasks.values()) {
+ if (ejv.getJobVertex().isInputVertex()) {
+ ejv.scheduleAll(slotProvider,
allowQueuedScheduling);
+ }
+ }
+ }
+
+ /**
+ *
+ *
+ * @param slotProvider The resource provider from which the slots are
allocated
+ * @param timeout The maximum time that the deployment may take,
before a
+ * TimeoutException is thrown.
+ */
+ private void scheduleEager(SlotProvider slotProvider, final Time
timeout) {
+ checkState(state == JobStatus.RUNNING, "job is not running
currently");
+
+ // Important: reserve all the space we need up front.
+ // that way we do not have any operation that can fail between
allocating the slots
+ // and adding them to the list. If we had a failure in between
there, that would
+ // cause the slots to get lost
+ final ArrayList<ExecutionAndSlot[]> resources = new
ArrayList<>(getNumberOfExecutionJobVertices());
+ final boolean queued = allowQueuedScheduling;
+
+ // we use this flag to handle failures in a 'finally' clause
+ // that allows us to not go through clumsy cast-and-rethrow
logic
+ boolean successful = false;
+
+ try {
+ // collecting all the slots may resize and fail in that
operation without slots getting lost
+ final ArrayList<Future<SimpleSlot>> slotFutures = new
ArrayList<>(getNumberOfExecutionJobVertices());
+
+ // allocate the slots (obtain all their futures
+ for (ExecutionJobVertex ejv :
getVerticesTopologically()) {
+ // these calls are not blocking, they only
return futures
+ ExecutionAndSlot[] slots =
ejv.allocateResourcesForAll(slotProvider, queued);
+
+ // we need to first add the slots to this list,
to be safe on release
+ resources.add(slots);
+
+ for (ExecutionAndSlot ens : slots) {
+ slotFutures.add(ens.slotFuture);
+ }
+ }
+
+ // this future is complete once all slot futures are
complete.
+ // the future fails once one slot future fails.
+ final ConjunctFuture allAllocationsComplete =
FutureUtils.combineAll(slotFutures);
--- End diff --
True, it is not incorrect. But some tasks would be already deployed if we
start as soon as some futures are ready. They would need to be canceled again,
which gives these not so nice fast
deploy/out-of-resource/cancel/wait-for-cancellation/retry/etc loops.
> Eager Scheduling should deploy all Tasks together
> -------------------------------------------------
>
> Key: FLINK-5747
> URL: https://issues.apache.org/jira/browse/FLINK-5747
> Project: Flink
> Issue Type: Bug
> Components: JobManager
> Affects Versions: 1.2.0
> Reporter: Stephan Ewen
> Assignee: Stephan Ewen
> Fix For: 1.3.0
>
>
> Currently, eager scheduling immediately triggers the scheduling for all
> vertices and their subtasks in topological order.
> This has two problems:
> - This works only, as long as resource acquisition is "synchronous". With
> dynamic resource acquisition in FLIP-6, the resources are returned as Futures
> which may complete out of order. This results in out-of-order (not in
> topological order) scheduling of tasks which does not work for streaming.
> - Deploying some tasks that depend on other tasks before it is clear that
> the other tasks have resources as well leads to situations where many
> deploy/recovery cycles happen before enough resources are available to get
> the job running fully.
> For eager scheduling, we should allocate all resources in one chunk and then
> deploy once we know that all are available.
> As a follow-up, the same should be done per pipelined component in lazy batch
> scheduling as well. That way we get lazy scheduling across blocking
> boundaries, and bulk (gang) scheduling in pipelined subgroups.
> This also does not apply for efforts of fine grained recovery, where
> individual tasks request replacement resources.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
