hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Zhihua Deng (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-16212) Many connections to datanode are created when doing a large scan
Date Tue, 19 Jul 2016 03:09:20 GMT

    [ https://issues.apache.org/jira/browse/HBASE-16212?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15383514#comment-15383514

Zhihua Deng commented on HBASE-16212:

Thanks [~stack]. From the logging, it implies that different threads share the same DFSInputStream
instance, say 'defaultRpcServer.handler=7'(handler7) and 'defaultRpcServer.handler=4'(handler4),
for example. The original will prefect the next block header and cache the header into thread.
When defaultRpcServer.handler=4 comes,  it first checks that the cached header offset is equal
to the the block starting offset, unfortunately these two numbers are unequal(-1 != offset).
The handler4 knows nothing about the block header,  though the header has been prefected by
handler7.  The handler4 needs to seek the inputstream with the block starting offset for obtaining
the header,  while the inputstream has been over read by 33 bytes(the header size). So a new
connection to datanode should be recreated, the elder one will be closed. When the datanode
writes to a closed channel, an socket exception will be raised. When the same case happens
frequently, the datanode will be suffered from logging the message described as it is.

> Many connections to datanode are created when doing a large scan 
> -----------------------------------------------------------------
>                 Key: HBASE-16212
>                 URL: https://issues.apache.org/jira/browse/HBASE-16212
>             Project: HBase
>          Issue Type: Improvement
>    Affects Versions: 1.1.2
>            Reporter: Zhihua Deng
>         Attachments: HBASE-16212.patch, HBASE-16212.v2.patch, regionserver-dfsinputstream.log
> As described in https://issues.apache.org/jira/browse/HDFS-8659, the datanode is suffering
from logging the same repeatedly. Adding log to DFSInputStream, it outputs as follows:
> 2016-07-10 21:31:42,147 INFO  [B.defaultRpcServer.handler=22,queue=1,port=16020] hdfs.DFSClient:
DFSClient_NONMAPREDUCE_1984924661_1 seek DatanodeInfoWithStorage[,DS-086bc494-d862-470c-86e8-9cb7929985c6,DISK]
for BP-360285305- pos: 111506876, targetPos:
>  ...
> As the pos of this input stream is larger than targetPos(the pos trying to seek), A new
connection to the datanode will be created, the older one will be closed as a consequence.
When the wrong seeking ops are large, the datanode's block scanner info message is spamming
logs, as well as many connections to the same datanode will be created.
> hadoop version: 2.7.1

This message was sent by Atlassian JIRA

View raw message