Hey Rita,
All software developed by Cloudera for CDH is Apache (v2) licensed and
freely available. See these docs [1,2] for more info.
We publish source packages (which includes the packaging source) and
source tarballs, you can find these at
http://archive.cloudera.com/cdh/3/. See the CHANGES.t
Adding cdh-user@, BCC common-user@
Hey Steve,
Sounds like you need to chmod 777 the staging dir. By default
mapreduce.jobtracker.staging.root.dir is
${hadoop.tmp.dir}/mapred/staging but per the mapred configuration
below setting this to /user is better and should mean you don't need
to do the abo
restrictive) in CDH3 beta 2 so that it's possible to use more
than one build in a cluster.
http://archive.cloudera.com/cdh/3/hadoop-0.20.2+320.releasenotes.html
Thanks,
Eli
On Wed, Sep 1, 2010 at 11:24 AM, Eli Collins wrote:
> Hey guys,
>
> In CDH3 you can pin your repo to a
Hey guys,
In CDH3 you can pin your repo to a particular release. Eg in the
following docs to use beta 1 specify "redhat/cdh/3b1" instead of
"redhat/cdh/3" in the repo file (for RH), or "-cdh3b1" instead
of "-cdh3" in the list file (for Debian). You'll need to do a
"yum clean metadata" or "apt-get
symlinks:
http://people.apache.org/~tomwhite/hadoop-0.21.0-candidate-0/
Thanks,
Eli
On Fri, Jul 16, 2010 at 7:20 AM, Yujun Wu wrote:
> Hello,
>
> I am new to hadoop. Recently, I installed Hadoop 0.20.2 and it works. I
> tried to patch it with the symbolic links patch by Eli (Mr.
Hey David,
This issued was fixed in CDH2 and will be in the next beta of CDH3.
Appreciate the feedback!
Thanks,
Eli
On Tue, May 18, 2010 at 12:59 PM, David Howell wrote:
> Are you using Cloudera's hadoop 0.20.2?
>
> There's some logic in bin/hadoop-config.sh that seems to be failing if
> JAVA_H
Hey Michael,
The script specified by dfs.network.script is passed both host names
and IPs. In most cases an IP is passed, however in some cases (eg when
using dfs.hosts files) a hostname is passed.
Thanks,
Eli
ps - useful pointers:
http://mail-archives.apache.org/mod_mbox/hadoop-common-user/200
The issue that required you changing ports is HDFS-961.
Thanks,
Eli
On Fri, Apr 23, 2010 at 6:30 AM, Christian Baun wrote:
> Brian,
>
> You got it!!! :-)
> It works (partly)!
>
> i switched to Port 9000. core-site.xml includes now:
>
>
> fs.default.name
>
> in shell commands, with one physical copy of a big data set, different ppl
> can easily create different subset of it to work on by using symlinks.
>
> Thanks,
>
> Michael
>
> --- On Wed, 3/10/10, Eli Collins wrote:
>
> From: Eli Collins
> Subject: Re: Sym
dered in the next release?
>
> Thanks,
>
> Michael
>
> --- On Tue, 3/9/10, Eli Collins wrote:
>
> From: Eli Collins
> Subject: Re: Symbolic link in Hadoop HDFS?
> To: common-user@hadoop.apache.org
> Date: Tuesday, March 9, 2010, 8:01 PM
>
> Hey Michael,
>
Hey Michael,
Symbolic links has been implemented [1] but are not yet available in a
Hadoop release. The implementation is only available to clients that
use the new FileContext API so clients like Hive need to be migrated
from using FileSystem to FileContext. This is currently being done in
Hadoop
> From what I read, I thought, that bookkeeper would be the ideal enhancement
> for the namenode, to make it distributed and therefor finaly highly available.
Being distributed doesn't imply high availability. Availability is
about minimizing downtime. For example, a primary that can fail over
to
On Wed, Jan 27, 2010 at 8:38 AM, Stuart Sierra
wrote:
> Hello,
>
> Does anyone have up-to-date instructions for installing hadoop-core in
> a local Maven repository? The instructions at
> http://wiki.apache.org/hadoop/HowToContribute do not work (the
> mvn-install target is not defined).
>
> Than
On Fri, Jan 22, 2010 at 3:57 AM, stephen mulcahy
wrote:
> Hi,
>
> I've been running some tests on some new hardware we have acquired.
>
> As a baseline, I ran the Hadoop sort[1] with 10GB and 100GB of data. As an
> experiment, I ran it on 4 systems (1 configured as master+slave and 3 as
> slaves)
On Thu, Jan 14, 2010 at 12:43 AM, Erik Forsberg wrote:
> Hi!
>
> I'm having trouble figuring out the numbers reported by 'hadoop dfs
> -dus' versus the numbers reported by the namenode web interface.
>
> I have a 4 node clusters, 4TB of disk on each node.
>
> hadoop dfs -dus /
> hdfs://hdp01-01:90
Have you verified this new DNs Hadoop configuration files are the same
as the others? Do you see any errors in the NN when restarting HDFS on
this new node?
Thanks,
Eli
On Sat, Jan 9, 2010 at 9:44 AM, Saptarshi Guha wrote:
> Hello,
> I'm using Hadoop 0.20.1. I just added a new node to a 5 node
>
> data.replication = 2
>
> A bit of topic - is it safe to have such number? About a year ago I heard
> only 3 way replication was fully tested, while 2 way had some issues - was
> it fixed in subsequent versions?
I think that's still a relatively untested configuration, though I'm
not aware of any
> I actually tested it with a simple Java test loader I quickly put together,
> which ran on each machine and continuously has written random data to DFS. I
> tuned the writing rate until I got ~77Mb/s - above it the iowait loads on
> each disk (measured by iostat) became above 50% - 60%, which is
Hey Stas,
Can you provide more information about your workload and the
environment? eg are you running t.o.a.h.h.BenchmarkThroughput,
TestDFSIO, or timing hadoop fs -put/get to transfer data to hdfs from
another machine, looking at metrics, etc. What else is running on the
cluster? Have you profil
> Could the communication between blade server and disk array be the bottleneck?
Yes, depending on the number of blades, the network into the array
will bottleneck because it doesn't scale with the number of data
nodes.
On Thu, Dec 17, 2009 at 10:13 AM, Owen O'Malley wrote:
> For a while there has been a jira about removing all of the cases where we
> currently fork a subprocess and replacing it with a jni library. It would be
> lovely if someone did that. *smile*
Just posted a patch to HADOOP-4998.
Heading ou
; Call to org.apache.hadoop.fs.FileSystem::exists failed!
>>>>> unique: 7, error: -2 (No such file or directory), outsize: 16
>>>>> unique: 8, opcode: LOOKUP (1), nodeid: 1, insize: 46
>>>>> LOOKUP /hdfs:
>>>>> getattr /hdfs:
>&
> fuse_dfs TRACE - readdir /
> unique: 4, success, outsize: 200
> unique: 5, opcode: RELEASEDIR (29), nodeid: 1, insize: 64
> unique: 5, success, outsize: 16
>
> Does It seem OK?
Hm, seems like it's not finding any directory entries. Mind putting a
printf in dfs_readdir after hdfsListDirectori
Thanks for the info. Please uncomment "//#define DOTRACE" in
fuse_dfs.h, recompile, and ls /mnt/dfs again and post the trace. That
will help identify the particular error that's causing the failure
below.
> getdents(3, 0x61fec8, 512) = -1 EIO (Input/output error)
> write(2, "ls: ", 4
1
>>>>> max_readahead=0x0002
>>>>> max_write=0x0002
>>>>> unique: 1, success, outsize: 40
>>>>> unique: 2, opcode: GETATTR (3), nodeid: 1, insize: 56
>>>>> getattr /
>>>>> unique: 3, opcode: GET
> For a while there has been a jira about removing all of the cases where we
> currently fork a subprocess and replacing it with a jni library. It would be
> lovely if someone did that. *smile*
I wrote such a native class for the local implementation of symlinks
(eg to make a jni callout for readl
The "fuse-dfs didn't recognize " and "fuse-dfs ignoring
option -d" are expected, they get passed along from fuse-dfs to fuse
(via fuse_main).
Does it work if you pass -o private? Still nothing reported by dmesg?
What does jps indicate is running?What linux distribution and kernel
are you using?
See HADOOP-5059.
On Fri, Dec 11, 2009 at 7:57 AM, pavel kolodin wrote:
> On Fri, 11 Dec 2009 11:52:57 -, Sean Owen wrote:
>
> "-Xmx900" means "give the entire JVM only 900 bytes of heap space"
>> which can't possibly work.
>> You do not say what problem you are trying to solve here. What
>
> Zhang
>
>
> -----Original Message-
> From: Eli Collins [mailto:e...@cloudera.com]
> Sent: Friday, December 04, 2009 2:03 PM
> To: common-user@hadoop.apache.org
> Subject: Re: Namenode crashes while rolling edit log from secondary
> namenode
>
> Hey Zhang,
>
Hey Zhang,
> org.apache.hadoop.hdfs.server.namenode.FSNamesystem: Fatal Error : All
> storage directories are inaccessible.
Are the directories specified by dfs.namenode.[name|edits].dir
accessible? Perhaps they're NFS mounts that are flaking out?
Thanks,
Eli
Hey Siddu,
You the testcase flag,
eg ant -Dtestcase=TestHDFSCLI test
to run TestHDFSCLI.java
Thanks,
Eli
On Tue, Dec 1, 2009 at 10:36 AM, Siddu wrote:
> Hi all ,
>
> I am interested in the exploring the test folder . which is present in
> src/test/org/apache/hadoop/hdfs/*
>
> Please ca
Hey Chris,
Forgot to mention the patch for HADOOP-5611 is in CDH2 and the patch
for HDFS-790 will be in the next update. If you want to see how the
src/c++ is built on CDH checkout the cloudera/do-release-build script.
Thanks,
Eli
On Sun, Nov 29, 2009 at 10:21 PM, Eli Collins wrote:
>
Hey Chris,
Thanks for reporting. I filed HDFS-790 and uploaded a patch (which you
will need to apply after applying HADOOP-5611) and verified it
compiles on karmic 64-bit. In the mean time if you just need to build
libhdfs (which doesn't depend on c++-utils) you can do that
with ant -Dcompile.c++=
Hey Martin,
It would be an interesting experiment but I'm not sure it would
improve things as the host (and hardware to some extent) are already
reading ahead. A useful exercise would be to evaluate whether the new
default host parameters for on-demand readahead are suitable for
hadoop.
http://lw
Hey Yang Jie,
The following works for me on hadoop-0.20.1 on ubuntu 9.04 (amd64)
ant -Dcompile.c++=true -Dlibhdfs=true compile-c++-libhdfs
You can see how libhdfs (and the rest of hadoop) is built in CDH by
looking at the file cloudera/do-release-build in the source:
http://archive.cloudera.com
> These values determine how much HDFS is *not* allowed to use. There is no
> limit on how much MR can take. This is exactly the opposite of what he and
> pretty much every other admin wants. [Negative math is fun! Or something.]
Hey Allen -- is there a JIRA for this? A quick search didn't tur
36 matches
Mail list logo