[
https://issues.apache.org/jira/browse/SPARK-11293?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15765096#comment-15765096
]
Barry Becker commented on SPARK-11293:
--------------------------------------
Not sure if this is related, but I am running on spark 2.0.2 through spark
job-server and see tons of messages like this:
{code}
[2016-12-20 11:49:28,662] WARN he.spark.executor.Executor []
[akka://JobServer/user/context-supervisor/sql-context] - Managed memory leak
detected; size = 5762976 bytes, TID = 42621
[2016-12-20 11:49:28,662] WARN k.memory.TaskMemoryManager []
[akka://JobServer/user/context-supervisor/sql-context] - leak 5.5 MB memory
from org.apache.spark.util.collection.ExternalSorter@35f81493
[2016-12-20 11:49:28,662] WARN he.spark.executor.Executor []
[akka://JobServer/user/context-supervisor/sql-context] - Managed memory leak
detected; size = 5762976 bytes, TID = 42622
[2016-12-20 11:49:28,702] WARN k.memory.TaskMemoryManager []
[akka://JobServer/user/context-supervisor/sql-context] - leak 5.5 MB memory
from org.apache.spark.util.collection.ExternalSorter@16da7c1a
[2016-12-20 11:49:28,702] WARN he.spark.executor.Executor []
[akka://JobServer/user/context-supervisor/sql-context] - Managed memory leak
detected; size = 5762976 bytes, TID = 42623
[2016-12-20 11:49:28,702] WARN k.memory.TaskMemoryManager []
[akka://JobServer/user/context-supervisor/sql-context] - leak 5.5 MB memory
from org.apache.spark.util.collection.ExternalSorter@151060cf
[2016-12-20 11:49:28,702] WARN he.spark.executor.Executor []
[akka://JobServer/user/context-supervisor/sql-context] - Managed memory leak
detected; size = 5762976 bytes, TID = 42624
[Stage 5700:=========================> (44 + 4) /
92][2016-12-20 11:49:35,479] WARN k.memory.TaskMemoryManager []
[akka://JobServer/user/context-supervisor/sql-context] - l
{code}
Are managed memory leaks ever expected behavior? Or do they always indicate a
memory leak problem? I don't really see the memory going up much in jVisualVM.
> ExternalSorter and ExternalAppendOnlyMap should free shuffle memory in their
> stop() methods
> -------------------------------------------------------------------------------------------
>
> Key: SPARK-11293
> URL: https://issues.apache.org/jira/browse/SPARK-11293
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 1.3.1, 1.4.1, 1.5.1, 1.6.0
> Reporter: Josh Rosen
> Assignee: Josh Rosen
> Priority: Critical
> Fix For: 1.6.0
>
>
> I discovered multiple leaks of shuffle memory while working on my memory
> manager consolidation patch, which added the ability to do strict memory leak
> detection for the bookkeeping that used to be performed by the
> ShuffleMemoryManager. This uncovered a handful of places where tasks can
> acquire execution/shuffle memory but never release it, starving themselves of
> memory.
> Problems that I found:
> * {{ExternalSorter.stop()}} should release the sorter's shuffle/execution
> memory.
> * BlockStoreShuffleReader should call {{ExternalSorter.stop()}} using a
> {{CompletionIterator}}.
> * {{ExternalAppendOnlyMap}} exposes no equivalent of {{stop()}} for freeing
> its resources.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
