[
https://issues.apache.org/jira/browse/HDDS-1046?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Shashikant Banerjee updated HDDS-1046:
--------------------------------------
Description:
{code:java}
java.lang.StackOverflowError
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.getSubject(Subject.java:297)
at
org.apache.hadoop.security.UserGroupInformation.getCurrentUser(UserGroupInformation.java:569)
at
org.apache.hadoop.hdds.scm.storage.ContainerProtocolCalls.getEncodedBlockToken(ContainerProtocolCalls.java:578)
at
org.apache.hadoop.hdds.scm.storage.ContainerProtocolCalls.writeChunkAsync(ContainerProtocolCalls.java:318)
at
org.apache.hadoop.hdds.scm.storage.BlockOutputStream.writeChunkToContainer(BlockOutputStream.java:602)
at
org.apache.hadoop.hdds.scm.storage.BlockOutputStream.writeChunk(BlockOutputStream.java:464)
at
org.apache.hadoop.hdds.scm.storage.BlockOutputStream.close(BlockOutputStream.java:480)
at
org.apache.hadoop.ozone.client.io.BlockOutputStreamEntry.close(BlockOutputStreamEntry.java:137)
at
org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:489)
at
org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:501)
at
org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:501)
{code}
The failure is happening because, ozone client receives a CONTAINER_NOT+OPEN
exception from daranode, and it allocates a new and retries to write. But every
allocate block call to SCM allocates a block on the same quasi closed container
and hence client retries indefinitely and ultimately runs out of stack space.
Logs below indicate 3 successive block allocations from SCM in quasi closed
container.
{code:java}
15:15:26.812 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189852894070 bcId: 0} |
ret=FAILURE
org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
Container 2 in QUASI_CLOSED state
15:15:26.818 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189853352823 bcId: 0} |
ret=FAILURE
org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
Container 2 in QUASI_CLOSED state
15:15:26.825 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189853746040 bcId: 0} |
ret=FAILURE
org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
Container 2 in QUASI_CLOSED state
{code}
was:
{code:java}
java.lang.StackOverflowError
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.getSubject(Subject.java:297)
at
org.apache.hadoop.security.UserGroupInformation.getCurrentUser(UserGroupInformation.java:569)
at
org.apache.hadoop.hdds.scm.storage.ContainerProtocolCalls.getEncodedBlockToken(ContainerProtocolCalls.java:578)
at
org.apache.hadoop.hdds.scm.storage.ContainerProtocolCalls.writeChunkAsync(ContainerProtocolCalls.java:318)
at
org.apache.hadoop.hdds.scm.storage.BlockOutputStream.writeChunkToContainer(BlockOutputStream.java:602)
at
org.apache.hadoop.hdds.scm.storage.BlockOutputStream.writeChunk(BlockOutputStream.java:464)
at
org.apache.hadoop.hdds.scm.storage.BlockOutputStream.close(BlockOutputStream.java:480)
at
org.apache.hadoop.ozone.client.io.BlockOutputStreamEntry.close(BlockOutputStreamEntry.java:137)
at
org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:489)
at
org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:501)
at
org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:501)
{code}
The failure is happening because, ozone client receives a CONTAINER_NOT+OPEN
exception from daranode, and it allocates a new and retries to write. But every
allocate block call to SCM allocates a block on the same quasi closed container
and hence client retries indefinitely and ultimately runs out of stack space.
Logs below indicate 3 successive block allocations from SCM in quasi closed
container.
{code:java}
15:15:26.812 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189852894070 bcId: 0} |
ret=FAILURE
org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
Container 2 in QUASI_CLOSED state
15:15:26.818 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189853352823 bcId: 0} |
ret=FAILURE
org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
Container 2 in QUASI_CLOSED state
15:15:26.825 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189853746040 bcId: 0} |
ret=FAILURE
org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
Container 2 in QUASI_CLOSED state
{code}
> TestCloseContainerByPipeline#testIfCloseContainerCommandHandlerIsInvoked
> fails intermittently
> ---------------------------------------------------------------------------------------------
>
> Key: HDDS-1046
> URL: https://issues.apache.org/jira/browse/HDDS-1046
> Project: Hadoop Distributed Data Store
> Issue Type: Improvement
> Components: SCM
> Affects Versions: 0.4.0
> Reporter: Shashikant Banerjee
> Priority: Major
> Fix For: 0.4.0
>
>
>
> {code:java}
> java.lang.StackOverflowError
> at java.security.AccessController.doPrivileged(Native Method)
> at javax.security.auth.Subject.getSubject(Subject.java:297)
> at
> org.apache.hadoop.security.UserGroupInformation.getCurrentUser(UserGroupInformation.java:569)
> at
> org.apache.hadoop.hdds.scm.storage.ContainerProtocolCalls.getEncodedBlockToken(ContainerProtocolCalls.java:578)
> at
> org.apache.hadoop.hdds.scm.storage.ContainerProtocolCalls.writeChunkAsync(ContainerProtocolCalls.java:318)
> at
> org.apache.hadoop.hdds.scm.storage.BlockOutputStream.writeChunkToContainer(BlockOutputStream.java:602)
> at
> org.apache.hadoop.hdds.scm.storage.BlockOutputStream.writeChunk(BlockOutputStream.java:464)
> at
> org.apache.hadoop.hdds.scm.storage.BlockOutputStream.close(BlockOutputStream.java:480)
> at
> org.apache.hadoop.ozone.client.io.BlockOutputStreamEntry.close(BlockOutputStreamEntry.java:137)
> at
> org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:489)
> at
> org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:501)
> at
> org.apache.hadoop.ozone.client.io.KeyOutputStream.handleFlushOrClose(KeyOutputStream.java:501)
> {code}
> The failure is happening because, ozone client receives a CONTAINER_NOT+OPEN
> exception from daranode, and it allocates a new and retries to write. But
> every allocate block call to SCM allocates a block on the same quasi closed
> container and hence client retries indefinitely and ultimately runs out of
> stack space.
> Logs below indicate 3 successive block allocations from SCM in quasi closed
> container.
> {code:java}
> 15:15:26.812 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
> op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189852894070 bcId: 0} |
> ret=FAILURE
> org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
> Container 2 in QUASI_CLOSED state
> 15:15:26.818 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
> op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189853352823 bcId: 0} |
> ret=FAILURE
> org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
> Container 2 in QUASI_CLOSED state
> 15:15:26.825 [grpc-default-executor-3] ERROR DNAudit - user=null | ip=null |
> op=WRITE_CHUNK {blockData=conID: 2 locID: 101533189853746040 bcId: 0} |
> ret=FAILURE
> org.apache.hadoop.hdds.scm.container.common.helpers.ContainerNotOpenException:
> Container 2 in QUASI_CLOSED state
> {code}
>
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
