[
https://issues.apache.org/jira/browse/HADOOP-11802?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Colin Patrick McCabe updated HADOOP-11802:
------------------------------------------
Attachment: HADOOP-11802.001.patch
version 1 of the patch:
* {{DataXceiver.java}}: do not close the DomainSocket on an error. This is bad
because the socket might already be getting poll()ed by the thread. Instead,
call {{shutdown(RDWR)}} on the socket.
* Log all {{Throwables}} that terminate the {{DomainSocketWatcher}} thread. If
there is a follow-on error, we don't want it to obscure the true cause of the
problem.
* {{DomainSocketWatcher.c}}: look for both {{POLLIN}} and {{POLLHUP}} events
when calling {{poll()}}. Some UNIX variants (although not Linux) return
POLLHUP instead of POLLIN when shutdown is called on the socket.
* {{BlockReaderFactory.java}}: add a {{injectRequestShortCircuitShmFailure}}
method to the {{BlockReaderFactory#FailureInjector}} class.
* {{TestShortCircuitCache}}: add unit test
> DomainSocketWatcher thread terminates sometimes after there is an I/O error
> during requestShortCircuitShm
> ---------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-11802
> URL: https://issues.apache.org/jira/browse/HADOOP-11802
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.7.0
> Reporter: Eric Payne
> Assignee: Colin Patrick McCabe
> Attachments: HADOOP-11802.001.patch
>
>
> In {{DataXceiver#requestShortCircuitShm}}, we attempt to recover from some
> errors by closing the {{DomainSocket}}. However, this violates the invariant
> that the domain socket should never be closed when it is being managed by the
> {{DomainSocketWatcher}}. Instead, we should call {{shutdown}} on the
> {{DomainSocket}}. When this bug hits, it terminates the
> {{DomainSocketWatcher}} thread.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)