[
https://issues.apache.org/jira/browse/HBASE-8314?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13630174#comment-13630174
]
Nicolas Liochon commented on HBASE-8314:
----------------------------------------
I would expect that if we have the lease, we will be able to get the file
length. Maybe a subset of unfortunate synchronisations and caches can make this
false. I would expect this to be impossible on recent hdfs versions with
HBASE-7878. This said, the patch won't break anything (it's just a retry), and
I have seen strange bugs on HDFS around this. May be we should add a warning in
the code we don't expect it to happen?
> HLogSplitter can retry to open a 0-length hlog file
> ---------------------------------------------------
>
> Key: HBASE-8314
> URL: https://issues.apache.org/jira/browse/HBASE-8314
> Project: HBase
> Issue Type: Bug
> Reporter: Jimmy Xiang
> Assignee: Jimmy Xiang
> Attachments: region-server.log, trunk-8314.patch
>
>
> In case a HLog file is of size 0, and it is under recovery, HLogSplitter will
> fail to open it since it can get the file length, therefore, master can't
> start.
> {noformat}
> java.io.IOException: Cannot obtain block length for LocatedBlock{...;
> getBlockSize()=0; corrupt=false; offset=0; locs=[...]}
> at
> org.apache.hadoop.hdfs.DFSInputStream.readBlockLength(DFSInputStream.java:238)
> at
> org.apache.hadoop.hdfs.DFSInputStream.fetchLocatedBlocksAndGetLastBlockLength(DFSInputStream.java:182)
> at org.apache.hadoop.hdfs.DFSInputStream.openInfo(DFSInputStream.java:124)
> at org.apache.hadoop.hdfs.DFSInputStream.<init>(DFSInputStream.java:117)
> at org.apache.hadoop.hdfs.DFSClient.open(DFSClient.java:1080)
> {noformat}
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
