[
https://issues.apache.org/jira/browse/HDFS-12098?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16089283#comment-16089283
]
Weiwei Yang commented on HDFS-12098:
------------------------------------
Attached a test case patch to reproduce this issue. Please take a look at
[^HDFS-12098-HDFS-7240.testcase.patch]. This patch simulates the scenario
# Start mini ozone cluster without starting scm
# Datanode is unable to register to scm
# Start scm, waiting for datanode to register
# Wait a while but datanode is still unable to successfully register to scm
if you apply this patch, it's gonna fail. You might have noticed the patch
changes some more code than just adding a test, that is because the reason I
mentioned earlier. I also have added a method to check if a datanode is
registered to scm so that we can check datanode state even scm is not started.
I have a patch to fix this also, if applied that patch, this test will pass. I
am ready to share that as well.
Thanks
> Ozone: Datanode is unable to register with scm if scm starts later
> ------------------------------------------------------------------
>
> Key: HDFS-12098
> URL: https://issues.apache.org/jira/browse/HDFS-12098
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: datanode, ozone, scm
> Reporter: Weiwei Yang
> Assignee: Weiwei Yang
> Priority: Critical
> Attachments: disabled-scm-test.patch, HDFS-12098-HDFS-7240.001.patch,
> HDFS-12098-HDFS-7240.002.patch, HDFS-12098-HDFS-7240.testcase.patch, Screen
> Shot 2017-07-11 at 4.58.08 PM.png, thread_dump.log
>
>
> Reproducing steps
> 1. Start namenode
> {{./bin/hdfs --daemon start namenode}}
> 2. Start datanode
> {{./bin/hdfs datanode}}
> will see following connection issues
> {noformat}
> 17/07/13 21:16:48 INFO ipc.Client: Retrying connect to server:
> ozone1.fyre.ibm.com/172.16.165.133:9861. Already tried 0 time(s); retry
> policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1
> SECONDS)
> 17/07/13 21:16:49 INFO ipc.Client: Retrying connect to server:
> ozone1.fyre.ibm.com/172.16.165.133:9861. Already tried 1 time(s); retry
> policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1
> SECONDS)
> 17/07/13 21:16:50 INFO ipc.Client: Retrying connect to server:
> ozone1.fyre.ibm.com/172.16.165.133:9861. Already tried 2 time(s); retry
> policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1
> SECONDS)
> 17/07/13 21:16:51 INFO ipc.Client: Retrying connect to server:
> ozone1.fyre.ibm.com/172.16.165.133:9861. Already tried 3 time(s); retry
> policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1
> SECONDS)
> {noformat}
> this is expected because scm is not started yet
> 3. Start scm
> {{./bin/hdfs scm}}
> expecting datanode can register to this scm, expecting the log in scm
> {noformat}
> 17/07/13 21:22:30 INFO node.SCMNodeManager: Data node with ID:
> af22862d-aafa-4941-9073-53224ae43e2c Registered.
> {noformat}
> but did *NOT* see this log. (_I debugged into the code and found the datanode
> state was transited SHUTDOWN unexpectedly because the thread leaks, each of
> those threads counted to set to next state and they all set to SHUTDOWN
> state_)
> 4. Create a container from scm CLI
> {{./bin/hdfs scm -container -create -c 20170714c0}}
> this fails with following exception
> {noformat}
> Creating container : 20170714c0.
> Error executing
> command:org.apache.hadoop.ipc.RemoteException(org.apache.hadoop.ozone.scm.exceptions.SCMException):
> Unable to create container while in chill mode
> at
> org.apache.hadoop.ozone.scm.container.ContainerMapping.allocateContainer(ContainerMapping.java:241)
> at
> org.apache.hadoop.ozone.scm.StorageContainerManager.allocateContainer(StorageContainerManager.java:392)
> at
> org.apache.hadoop.ozone.protocolPB.StorageContainerLocationProtocolServerSideTranslatorPB.allocateContainer(StorageContainerLocationProtocolServerSideTranslatorPB.java:73)
> {noformat}
> datanode was not registered to scm, thus it's still in chill mode.
> *Note*, if we start scm first, there is no such issue, I can create container
> from CLI without any problem.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
