[
https://issues.apache.org/jira/browse/HADOOP-13702?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15593805#comment-15593805
]
Jingcheng Du commented on HADOOP-13702:
---------------------------------------
Thanks a lot [~chris.douglas].
bq. Shouldn't the ThreadLocal<Long> readLockHeldTimeStamp field be final, not
transient?
Sorry for the careless, that was an old code for extending ReadLock which is
serializable, I only removed the readObject method but forgot this, so sorry.
bq. InstrumentedReadWriteLock could accept an instance of ReadWriteLock instead
of creating one. Similarly, the read/write locks could accept a ReadWriteLock
in their constructor. IRWL could also have the convenience constructor, as
written.
The instrumented read lock uses methods in ReentrantReadWriteLock for
instrumentation. If we accept a ReadWriteLock in constructors i am afraid we
cannot support the instrumentation in the read lock.
bq. Is InstrumentedLock unused in HDFS? I expected to see package changes in
the HDFS project for the imports to resolve.
Only FsDatasetImpl uses this class, and this change is included in the patch:)
> Add a new instrumented read-write lock
> --------------------------------------
>
> Key: HADOOP-13702
> URL: https://issues.apache.org/jira/browse/HADOOP-13702
> Project: Hadoop Common
> Issue Type: Improvement
> Components: common
> Reporter: Jingcheng Du
> Assignee: Jingcheng Du
> Attachments: HADOOP-13702-V6.patch, HADOOP-13702-V7.patch,
> HADOOP-13702-V8.patch, HDFS-10924-2.patch, HDFS-10924-3.patch,
> HDFS-10924-4.patch, HDFS-10924-5.patch, HDFS-10924.patch
>
>
> Add a new instrumented read-write lock in hadoop common, so that the
> HDFS-9668 can use this to improve the locking in FsDatasetImpl
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
