Github user zhzhan commented on a diff in the pull request:
https://github.com/apache/spark/pull/15541#discussion_r84619879
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala ---
@@ -250,24 +251,24 @@ private[spark] class TaskSchedulerImpl(
private def resourceOfferSingleTaskSet(
taskSet: TaskSetManager,
maxLocality: TaskLocality,
- shuffledOffers: Seq[WorkerOffer],
- availableCpus: Array[Int],
- tasks: IndexedSeq[ArrayBuffer[TaskDescription]]) : Boolean = {
+ taskAssigner: TaskAssigner) : Boolean = {
var launchedTask = false
- for (i <- 0 until shuffledOffers.size) {
- val execId = shuffledOffers(i).executorId
- val host = shuffledOffers(i).host
- if (availableCpus(i) >= CPUS_PER_TASK) {
+ taskAssigner.init()
+ while (taskAssigner.hasNext) {
+ var isAccepted = false
+ val currentOffer = taskAssigner.next()
+ val execId = currentOffer.workOffer.executorId
+ val host = currentOffer.workOffer.host
+ if (currentOffer.coresAvailable >= CPUS_PER_TASK) {
try {
for (task <- taskSet.resourceOffer(execId, host, maxLocality)) {
- tasks(i) += task
+ currentOffer.assignTask(task, CPUS_PER_TASK)
val tid = task.taskId
taskIdToTaskSetManager(tid) = taskSet
taskIdToExecutorId(tid) = execId
executorIdToTaskCount(execId) += 1
- availableCpus(i) -= CPUS_PER_TASK
- assert(availableCpus(i) >= 0)
--- End diff --
Thanks @viirya for the comments. Actually I was thinking removing the check
although it is part of the legacy code. Now the check is moved into OfferState,
which makes more sense. IMHO, typically the assertion should never fail. But
from the OfferState's perspective, it should guarantee such restriction.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
