[
https://issues.apache.org/jira/browse/HDFS-17416?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17829446#comment-17829446
]
ASF GitHub Bot commented on HDFS-17416:
---------------------------------------
ZanderXu commented on code in PR #6647:
URL: https://github.com/apache/hadoop/pull/6647#discussion_r1533484754
##########
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java:
##########
@@ -5346,11 +5346,12 @@ NamenodeCommand startCheckpoint(NamenodeRegistration
backupNode,
public void processIncrementalBlockReport(final DatanodeID nodeID,
final StorageReceivedDeletedBlocks srdb)
throws IOException {
- writeLock();
+ // Needs the FSWriteLock since it may update quota and access storage
policyId and full path.
Review Comment:
Yes. `completeBlock` will updateQuota, so it needs BMWriteLock and
FSWriteLock.
`processExtraRedundancyBlock` chooses excess replica depends on storage
policyId, so it needs FSReadLock.
`isInSnapshot` depends on the full path, so it needs FSReadLock.
> [FGL] Monitor threads in BlockManager.class support fine-grained lock
> ---------------------------------------------------------------------
>
> Key: HDFS-17416
> URL: https://issues.apache.org/jira/browse/HDFS-17416
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: ZanderXu
> Assignee: ZanderXu
> Priority: Major
> Labels: pull-request-available
>
> There are some monitor threads in BlockManager.class.
>
> This ticket is used to make these threads supporting fine-grained locking.
> * BlockReportProcessingThread
> * MarkedDeleteBlockScrubber
> * RedundancyMonitor
> * Reconstruction Queue Initializer
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
