[
https://issues.apache.org/jira/browse/HADOOP-18546?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17641644#comment-17641644
]
ASF GitHub Bot commented on HADOOP-18546:
-----------------------------------------
pranavsaxena-microsoft opened a new pull request, #5176:
URL: https://github.com/apache/hadoop/pull/5176
JIRA: https://issues.apache.org/jira/browse/HADOOP-18546
**Details:**
AbfsInputStream.close() can trigger the return of buffers used for active
prefetch GET requests into the ReadBufferManager free buffer pool.
A subsequent prefetch by a different stream in the same process may acquire
this same buffer. This can lead to risk of corruption of its own prefetched
data, data which may then be returned to that other thread.
Parent JIRA: https://issues.apache.org/jira/browse/HADOOP-18521
In this PR, we are disabling the purging of the inprogressList. The
readBuffers in InProgressList will get to ReadBufferWorker and get processed
and finally get into completedList. After a thresholdAgeMilliseconds, the
readBuffer would be evicted
(https://github.com/apache/hadoop/blob/trunk/hadoop-tools/hadoop-azure/src/main/java/org/apache/hadoop/fs/azurebfs/services/ReadBufferManager.java#L280-L285)
**Testing:**
https://github.com/apache/hadoop/blob/trunk/hadoop-tools/hadoop-azure/src/main/java/org/apache/hadoop/fs/azurebfs/services/ReadBufferManager.java#L280-L285
> disable purging list of in progress reads in abfs stream closed
> ---------------------------------------------------------------
>
> Key: HADOOP-18546
> URL: https://issues.apache.org/jira/browse/HADOOP-18546
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/azure
> Affects Versions: 3.3.4
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Major
>
> turn off the prune of in progress reads in
> ReadBufferManager::purgeBuffersForStream
> this will ensure active prefetches for a closed stream complete. they wiill
> then get to the completed list and hang around until evicted by timeout, but
> at least prefetching will be safe.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
