[jira] [Commented] (SPARK-25081) Nested spill in ShuffleExternalSorter may access a released memory page
[ https://issues.apache.org/jira/browse/SPARK-25081?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16576724#comment-16576724 ] Apache Spark commented on SPARK-25081: -- User 'zsxwing' has created a pull request for this issue: https://github.com/apache/spark/pull/22072 > Nested spill in ShuffleExternalSorter may access a released memory page > > > Key: SPARK-25081 > URL: https://issues.apache.org/jira/browse/SPARK-25081 > Project: Spark > Issue Type: Bug > Components: Spark Core >Affects Versions: 1.6.0, 1.6.1, 1.6.2, 1.6.3, 2.0.0, 2.0.1, 2.0.2, 2.1.0, > 2.1.1, 2.1.2, 2.1.3, 2.2.0, 2.2.1, 2.2.2, 2.3.0, 2.3.1, 2.3.2 >Reporter: Shixiong Zhu >Assignee: Shixiong Zhu >Priority: Blocker > Labels: correctness > Fix For: 2.4.0, 2.3.3 > > > This issue is pretty similar to SPARK-21907. > "allocateArray" in > [ShuffleInMemorySorter.reset|https://github.com/apache/spark/blob/9b8521e53e56a53b44c02366a99f8a8ee1307bbf/core/src/main/java/org/apache/spark/shuffle/sort/ShuffleInMemorySorter.java#L99] > may trigger a spill and cause ShuffleInMemorySorter access the released > `array`. Another task may get the same memory page from the pool. This will > cause two tasks access the same memory page. When a task reads memory written > by another task, many types of failures may happen. Here are some examples I > have seen: > - JVM crash. (This is easy to reproduce in a unit test as we fill newly > allocated and deallocated memory with 0xa5 and 0x5a bytes which usually > points to an invalid memory address) > - java.lang.IllegalArgumentException: Comparison method violates its general > contract! > - java.lang.NullPointerException at > org.apache.spark.memory.TaskMemoryManager.getPage(TaskMemoryManager.java:384) > - java.lang.UnsupportedOperationException: Cannot grow BufferHolder by size > -536870912 because the size after growing exceeds size limitation 2147483632 -- 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
[jira] [Commented] (SPARK-25081) Nested spill in ShuffleExternalSorter may access a released memory page
[ https://issues.apache.org/jira/browse/SPARK-25081?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16576700#comment-16576700 ] Thomas Graves commented on SPARK-25081: --- thanks, wanted to clarify since the description only mentioned exceptions/crashes. > Nested spill in ShuffleExternalSorter may access a released memory page > > > Key: SPARK-25081 > URL: https://issues.apache.org/jira/browse/SPARK-25081 > Project: Spark > Issue Type: Bug > Components: Spark Core >Affects Versions: 1.6.0, 1.6.1, 1.6.2, 1.6.3, 2.0.0, 2.0.1, 2.0.2, 2.1.0, > 2.1.1, 2.1.2, 2.1.3, 2.2.0, 2.2.1, 2.2.2, 2.3.0, 2.3.1, 2.3.2 >Reporter: Shixiong Zhu >Assignee: Shixiong Zhu >Priority: Blocker > Labels: correctness > Fix For: 2.4.0, 2.3.3 > > > This issue is pretty similar to SPARK-21907. > "allocateArray" in > [ShuffleInMemorySorter.reset|https://github.com/apache/spark/blob/9b8521e53e56a53b44c02366a99f8a8ee1307bbf/core/src/main/java/org/apache/spark/shuffle/sort/ShuffleInMemorySorter.java#L99] > may trigger a spill and cause ShuffleInMemorySorter access the released > `array`. Another task may get the same memory page from the pool. This will > cause two tasks access the same memory page. When a task reads memory written > by another task, many types of failures may happen. Here are some examples I > have seen: > - JVM crash. (This is easy to reproduce in a unit test as we fill newly > allocated and deallocated memory with 0xa5 and 0x5a bytes which usually > points to an invalid memory address) > - java.lang.IllegalArgumentException: Comparison method violates its general > contract! > - java.lang.NullPointerException at > org.apache.spark.memory.TaskMemoryManager.getPage(TaskMemoryManager.java:384) > - java.lang.UnsupportedOperationException: Cannot grow BufferHolder by size > -536870912 because the size after growing exceeds size limitation 2147483632 -- 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
[jira] [Commented] (SPARK-25081) Nested spill in ShuffleExternalSorter may access a released memory page
[ https://issues.apache.org/jira/browse/SPARK-25081?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16576636#comment-16576636 ] Shixiong Zhu commented on SPARK-25081: -- That's possible. That's why I added the "corrnectness" label. > Nested spill in ShuffleExternalSorter may access a released memory page > > > Key: SPARK-25081 > URL: https://issues.apache.org/jira/browse/SPARK-25081 > Project: Spark > Issue Type: Bug > Components: Spark Core >Affects Versions: 2.3.1 >Reporter: Shixiong Zhu >Assignee: Shixiong Zhu >Priority: Major > Labels: correctness > > This issue is pretty similar to SPARK-21907. > "allocateArray" in > [ShuffleInMemorySorter.reset|https://github.com/apache/spark/blob/9b8521e53e56a53b44c02366a99f8a8ee1307bbf/core/src/main/java/org/apache/spark/shuffle/sort/ShuffleInMemorySorter.java#L99] > may trigger a spill and cause ShuffleInMemorySorter access the released > `array`. Another task may get the same memory page from the pool. This will > cause two tasks access the same memory page. When a task reads memory written > by another task, many types of failures may happen. Here are some examples I > have seen: > - JVM crash. (This is easy to reproduce in a unit test as we fill newly > allocated and deallocated memory with 0xa5 and 0x5a bytes which usually > points to an invalid memory address) > - java.lang.IllegalArgumentException: Comparison method violates its general > contract! > - java.lang.NullPointerException at > org.apache.spark.memory.TaskMemoryManager.getPage(TaskMemoryManager.java:384) > - java.lang.UnsupportedOperationException: Cannot grow BufferHolder by size > -536870912 because the size after growing exceeds size limitation 2147483632 -- 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
[jira] [Commented] (SPARK-25081) Nested spill in ShuffleExternalSorter may access a released memory page
[ https://issues.apache.org/jira/browse/SPARK-25081?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16576234#comment-16576234 ] Thomas Graves commented on SPARK-25081: --- Does this ever result in the task reading the wrong data and succeeding? so we essentially lost data or got the wrong results. > Nested spill in ShuffleExternalSorter may access a released memory page > > > Key: SPARK-25081 > URL: https://issues.apache.org/jira/browse/SPARK-25081 > Project: Spark > Issue Type: Bug > Components: Spark Core >Affects Versions: 2.3.1 >Reporter: Shixiong Zhu >Assignee: Shixiong Zhu >Priority: Major > Labels: correctness > > This issue is pretty similar to SPARK-21907. > "allocateArray" in > [ShuffleInMemorySorter.reset|https://github.com/apache/spark/blob/9b8521e53e56a53b44c02366a99f8a8ee1307bbf/core/src/main/java/org/apache/spark/shuffle/sort/ShuffleInMemorySorter.java#L99] > may trigger a spill and cause ShuffleInMemorySorter access the released > `array`. Another task may get the same memory page from the pool. This will > cause two tasks access the same memory page. When a task reads memory written > by another task, many types of failures may happen. Here are some examples I > have seen: > - JVM crash. (This is easy to reproduce in a unit test as we fill newly > allocated and deallocated memory with 0xa5 and 0x5a bytes which usually > points to an invalid memory address) > - java.lang.IllegalArgumentException: Comparison method violates its general > contract! > - java.lang.NullPointerException at > org.apache.spark.memory.TaskMemoryManager.getPage(TaskMemoryManager.java:384) > - java.lang.UnsupportedOperationException: Cannot grow BufferHolder by size > -536870912 because the size after growing exceeds size limitation 2147483632 -- 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
[jira] [Commented] (SPARK-25081) Nested spill in ShuffleExternalSorter may access a released memory page
[ https://issues.apache.org/jira/browse/SPARK-25081?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16575598#comment-16575598 ] Apache Spark commented on SPARK-25081: -- User 'zsxwing' has created a pull request for this issue: https://github.com/apache/spark/pull/22062 > Nested spill in ShuffleExternalSorter may access a released memory page > > > Key: SPARK-25081 > URL: https://issues.apache.org/jira/browse/SPARK-25081 > Project: Spark > Issue Type: Bug > Components: Spark Core >Affects Versions: 2.3.1 >Reporter: Shixiong Zhu >Assignee: Shixiong Zhu >Priority: Major > > This issue is pretty similar to SPARK-21907. > "allocateArray" in > [ShuffleInMemorySorter.reset|https://github.com/apache/spark/blob/9b8521e53e56a53b44c02366a99f8a8ee1307bbf/core/src/main/java/org/apache/spark/shuffle/sort/ShuffleInMemorySorter.java#L99] > may trigger a spill and cause ShuffleInMemorySorter access the released > `array`. Another task may get the same memory page from the pool. This will > cause two tasks access the same memory page. When a task reads memory written > by another task, many types of failures may happen. Here are some examples I > have seen: > - JVM crash. (This is easy to reproduce in a unit test as we fill newly > allocated and deallocated memory with 0xa5 and 0x5a bytes which usually > points to an invalid memory address) > - java.lang.IllegalArgumentException: Comparison method violates its general > contract! > - java.lang.NullPointerException at > org.apache.spark.memory.TaskMemoryManager.getPage(TaskMemoryManager.java:384) > - java.lang.UnsupportedOperationException: Cannot grow BufferHolder by size > -536870912 because the size after growing exceeds size limitation 2147483632 -- 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