[
https://issues.apache.org/jira/browse/HDFS-6527?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14042720#comment-14042720
]
Jing Zhao commented on HDFS-6527:
---------------------------------
Thanks for the report, [~l201514]! Actually the unit test failure can be
reproduced in branch-2.4.1.
I think the patch should not be included in branch-2.4.1. In branch-2.4.1, we
have the following code in the beginning of analyzeFileState:
{code}
final INodesInPath iip = dir.getINodesInPath4Write(src);
final INodeFile pendingFile
= checkLease(src, fileId, clientName, iip.getLastINode());
{code}
I.e., the inode is retrieved by resolving the path. Since we remove the inode
from the children list while holding the fsnamesystem lock, the race condition
can actually be avoided I guess.
HDFS-6294 changed this part to load the INode from the inode map, while in
FSNamesystem#deleteInternal, the inode is removed from the inode map without
holding fsnamesystem lock. I guess this is the cause of the issue, and the
issue is only in 2.5 and trunk.
> Edit log corruption due to defered INode removal
> ------------------------------------------------
>
> Key: HDFS-6527
> URL: https://issues.apache.org/jira/browse/HDFS-6527
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.4.0
> Reporter: Kihwal Lee
> Assignee: Kihwal Lee
> Priority: Blocker
> Fix For: 2.4.1
>
> Attachments: HDFS-6527.branch-2.4.patch, HDFS-6527.trunk.patch,
> HDFS-6527.v2.patch, HDFS-6527.v3.patch, HDFS-6527.v4.patch, HDFS-6527.v5.patch
>
>
> We have seen a SBN crashing with the following error:
> {panel}
> \[Edit log tailer\] ERROR namenode.FSEditLogLoader:
> Encountered exception on operation AddBlockOp
> [path=/xxx,
> penultimateBlock=NULL, lastBlock=blk_111_111, RpcClientId=,
> RpcCallId=-2]
> java.io.FileNotFoundException: File does not exist: /xxx
> {panel}
> This was caused by the deferred removal of deleted inodes from the inode map.
> Since getAdditionalBlock() acquires FSN read lock and then write lock, a
> deletion can happen in between. Because of deferred inode removal outside FSN
> write lock, getAdditionalBlock() can get the deleted inode from the inode map
> with FSN write lock held. This allow addition of a block to a deleted file.
> As a result, the edit log will contain OP_ADD, OP_DELETE, followed by
> OP_ADD_BLOCK. This cannot be replayed by NN, so NN doesn't start up or SBN
> crashes.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
