[
https://issues.apache.org/jira/browse/SPARK-23888?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
wuyi updated SPARK-23888:
-------------------------
Description:
There's a bug in:
{code:java}
/** Check whether a task is currently running an attempt on a given host */
private def hasAttemptOnHost(taskIndex: Int, host: String): Boolean = {
taskAttempts(taskIndex).exists(_.host == host)
}
{code}
This will ignore hosts which have finished attempts, so we should check whether
the attempt is currently running on the given host.
And it is possible for a speculative task to run on a host where another
attempt failed here before.
Assume we have only two machines: host1, host2. We first run task0.0 on host1.
Then, due to a long time waiting for task0.0, we launch a speculative task0.1
on host2. And, task0.1 finally failed on host1, but it can not re-run since
there's already a copy running on host2. After another long time, we launch a
new speculative task0.2. And, now, we can run task0.2 on host1 again, since
there's no more running attempt on host1.
******
After discussion in the PR, we simply make the comment be consistent the
method's behavior. See details in PR#20998.
was:
There's a bug in:
{code:java}
/** Check whether a task is currently running an attempt on a given host */
private def hasAttemptOnHost(taskIndex: Int, host: String): Boolean = {
taskAttempts(taskIndex).exists(_.host == host)
}
{code}
This will ignore hosts which have finished attempts, so we should check whether
the attempt is currently running on the given host.
And it is possible for a speculative task to run on a host where another
attempt failed here before.
Assume we have only two machines: host1, host2. We first run task0.0 on host1.
Then, due to a long time waiting for task0.0, we launch a speculative task0.1
on host2. And, task0.1 finally failed on host1, but it can not re-run since
there's already a copy running on host2. After another long time, we launch a
new speculative task0.2. And, now, we can run task0.2 on host1 again, since
there's no more running attempt on host1.
> speculative task should not run on a given host where another attempt is
> already running on
> -------------------------------------------------------------------------------------------
>
> Key: SPARK-23888
> URL: https://issues.apache.org/jira/browse/SPARK-23888
> Project: Spark
> Issue Type: Bug
> Components: Scheduler, Spark Core
> Affects Versions: 2.3.0
> Reporter: wuyi
> Priority: Major
> Labels: speculation
> Fix For: 2.3.0
>
>
>
> There's a bug in:
> {code:java}
> /** Check whether a task is currently running an attempt on a given host */
> private def hasAttemptOnHost(taskIndex: Int, host: String): Boolean = {
> taskAttempts(taskIndex).exists(_.host == host)
> }
> {code}
> This will ignore hosts which have finished attempts, so we should check
> whether the attempt is currently running on the given host.
> And it is possible for a speculative task to run on a host where another
> attempt failed here before.
> Assume we have only two machines: host1, host2. We first run task0.0 on
> host1. Then, due to a long time waiting for task0.0, we launch a speculative
> task0.1 on host2. And, task0.1 finally failed on host1, but it can not re-run
> since there's already a copy running on host2. After another long time, we
> launch a new speculative task0.2. And, now, we can run task0.2 on host1
> again, since there's no more running attempt on host1.
> ******
> After discussion in the PR, we simply make the comment be consistent the
> method's behavior. See details in PR#20998.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
