hadoop-mapreduce-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brahma Reddy Battula <brahmareddy.batt...@huawei.com>
Subject RE: Huge disk IO on only one disk
Date Mon, 03 Mar 2014 06:51:30 GMT




What should be the standard around setting up the hadoop.tmp.dir parameter.



>>>>>>>> As I know hadoop.tmp.dir   will be used for follow properites,
If you are configuring following properties,then you no need to configure this one..





MapReduce:



mapreduce.cluster.local.dir     ${hadoop.tmp.dir}/mapred/local  The local directory where
MapReduce stores intermediate data files. May be a comma-separated list of directories on
different devices in order to spread disk i/o. Directories that do not exist are ignored.
mapreduce.jobtracker.system.dir ${hadoop.tmp.dir}/mapred/system The directory where MapReduce
stores control files.
mapreduce.jobtracker.staging.root.dir   ${hadoop.tmp.dir}/mapred/staging        The root of
the staging area for users' job files In practice, this should be the directory where users'
home directories are located (usually /user)
mapreduce.cluster.temp.dir      ${hadoop.tmp.dir}/mapred/temp   A shared directory for temporary
files.



Yarn :



yarn.nodemanager.local-dirs     ${hadoop.tmp.dir}/nm-local-dir  List of directories to store
localized files in. An application's localized file directory will be found in: ${yarn.nodemanager.local-dirs}/usercache/${user}/appcache/application_${appid}.
Individual containers' work directories, called container_${contid}, will be subdirectories
of this.





HDFS :



dfs.namenode.name.dir   file://${hadoop.tmp.dir}/dfs/name       Determines where on the local
filesystem the DFS name node should store the name table(fsimage). If this is a comma-delimited
list of directories then the name table is replicated in all of the directories, for redundancy.



dfs.datanode.data.dir   file://${hadoop.tmp.dir}/dfs/data       Determines where on the local
filesystem an DFS data node should store its blocks. If this is a comma-delimited list of
directories, then data will be stored in all named directories, typically on different devices.
Directories that do not exist are ignored.



dfs.namenode.checkpoint.dir     file://${hadoop.tmp.dir}/dfs/namesecondary      Determines
where on the local filesystem the DFS secondary name node should store the temporary images
to merge. If this is a comma-delimited list of directories then the image is replicated in
all of the directories for redundancy.









Thanks & Regards



Brahma Reddy Battula



________________________________
From: Siddharth Tiwari [siddharth.tiwari@live.com]
Sent: Monday, March 03, 2014 11:20 AM
To: USers Hadoop
Subject: RE: Huge disk IO on only one disk

Hi Brahma,

No I havnt, I have put comma separated list of disks here dfs.datanode.data.dir . Have put
disk5 for hadoop.tmp.dir. My Q is, should we set up hadoop.tmp.dir or not ? if yes what should
be standards around.


*------------------------*
Cheers !!!
Siddharth Tiwari
Have a refreshing day !!!
"Every duty is holy, and devotion to duty is the highest form of worship of God.”
"Maybe other people will try to limit me but I don't limit myself"


________________________________
From: brahmareddy.battula@huawei.com
To: user@hadoop.apache.org
Subject: RE: Huge disk IO on only one disk
Date: Mon, 3 Mar 2014 05:14:34 +0000



Seems to be you had started cluster with default values for the following two properties and
configured for only hadoop.tmp.dir .

dfs.datanode.data.dir --->  file://${hadoop.tmp.dir}/dfs/data (Default value)

>>>>Determines where on the local filesystem an DFS data node should store its
blocks. If this is a comma-delimited list of directories, then data will be stored in all
named directories, typically on different devices

yarn.nodemanager.local-dirs -->  ${hadoop.tmp.dir}/nm-local-dir (Default value)

>>>>>>To store localized files, It's like inetermediate files


Please configure above two values as muliple dir's..



Thanks & Regards
Brahma Reddy Battula

________________________________
From: Siddharth Tiwari [siddharth.tiwari@live.com]
Sent: Monday, March 03, 2014 5:58 AM
To: USers Hadoop
Subject: Huge disk IO on only one disk

Hi Team,

I have 10 disks over which I am running my HDFS. Out of this on disk5 I have my hadoop.tmp.dir
configured. I see that on this disk I have huge IO when I run my jobs compared to other disks.
Can you guide my to the standards to follow so that this IO can be distributed across to other
disks as well.
What should be the standard around setting up the hadoop.tmp.dir parameter.
Any help would be highly appreciated. below is IO while I am running a huge job.



Device:            tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn

sda               2.11        37.65       226.20  313512628 1883809216

sdb               1.47        96.44       152.48  803144582 1269829840

sdc               1.45        93.03       153.10  774765734 1274979080

sdd               1.46        95.06       152.73  791690022 1271944848

sde               1.47        92.70       153.24  772025750 1276195288

sdf               1.55        95.77       153.06  797567654 1274657320

sdg              10.10       364.26      1951.79 3033537062 16254346480

sdi               1.46        94.82       152.98  789646630 1274014936

sdh               1.44        94.09       152.57  783547390 1270598232

sdj               1.44        91.94       153.37  765678470 1277220208

sdk               1.52        97.01       153.02  807928678 1274300360


*------------------------*
Cheers !!!
Siddharth Tiwari
Have a refreshing day !!!
"Every duty is holy, and devotion to duty is the highest form of worship of God.”
"Maybe other people will try to limit me but I don't limit myself"

Mime
View raw message