The two issues both implement DFSClient to directly open data blocks that
happen to be on the same machine function. What are advantage of HDFS-347?
Thanks,
LiuLei
Have you looked at graph processing for Hadoop? Like Hama (
http://hama.apache.org/) or Giraph (http://incubator.apache.org/giraph/).
I can't say for sure it would help you but it seems to be in the same
problem domain.
With regard to the chaining reducer issue this is indeed a general
implementat
Isn't also of some help using Cascading (http://www.cascading.org/) ?
*Fabio Pitzolu*
Consultant - BI & Infrastructure
Mob. +39 3356033776
Telefono 02 87157239
Fax. 02 93664786
*Gruppo Consulenza Innovazione - http://www.gr-ci.com*
2012/10/8 Bertrand Dechoux
> Have you looked at graph proce
The question is not how to sequence all. Cascading could indeed help in
that case.
But how to skip the map phase and do the split/local sort directly at the
end of the reduce so that the next reduce need only to do a merge on the
sorted files obtained from the previous reduce. This is basically a
> call context.write() in my mapper class)? If not, are there any other
> MR platforms that can do this? I've been searching around and couldn't
You can use Hama BSP[1] instead of Map/Reduce.
No stable release yet but I confirmed that large graph with billions
of nodes and edges can be crunched i
I don't believe that Hama would suffice.
In terms of M/R where you want to chain reducers...
Can you chain combiners? (I don't think so, but you never know)
If not, you end up with a series of M/R jobs and the Mappers are just identity
mappers.
Or you could use HBase, with a small caveat...
As HDFS-2246 itself states in one comment:
"""
HDFS-347 discusses ways to optimize reads for local clients. A clean
design is fairly involved. A shortcut has been proposed where the
client access the hdfs file blocks directly; this works if the client
is the same user/group as the DN daemon. This
I tried using FileUtil class for creating a symlink within hadoop actually
i want to create a symlink for my har directory so my code looks like
FileUtil.symLink("/user2/","har:///user/5oct2012.har")
but getting error like this
org.apache.hadoop.fs.FileUtil
- Command 'ln -s /user2/ har://user/5
Hi,
I know that DataNode and TaskTracker must restart to change topology.
Is there the method to execute the topology change without restart of
DataNode and TaskTracker?
In other words, can I change the topology by a command?
Thanks in advance!
Shinichi
Hi,
The FileUtil.symlink command does nothing more than call the unix "ln"
command, so it has no knowledge of how to work with Hadoop archive
files, only plain files and directories.
Is your archive on local disk, or in HDFS?
Cheers,
Dave
On 8 October 2012 13:43, Visioner Sadak wrote:
> I tried u
Hi-
Is there a simple way for error output / debug information generated by
a Mapper to be collected in one place for a given M/R job run? I guess
what I'm hoping for is sort of the reverse of a Distributed Cache function.
Can anyone suggest an approach?
Thanks,
Terry
thanks dave its in hdfs onlyany other methods of creating a symlink
On Mon, Oct 8, 2012 at 7:00 PM, Dave Beech wrote:
> Hi,
> The FileUtil.symlink command does nothing more than call the unix "ln"
> command, so it has no knowledge of how to work with Hadoop archive
> files, only plain files
Hello-
I know that it is contrary to normal Hadoop operation, but how can I
configure my M/R job to send one complete file to each mapper task? This
is intended to be used on many files in the 1.5 MB range as the first
step in a chain of processes.
thanks.
act i have to access my archived hadoop har
files thru http webhdfs normal files i am able to read thru http but once
my files are HAR archived i m not able to read it ...thts why creating a
symlink so tht the url remains same
On Mon, Oct 8, 2012 at 7:50 PM, Visioner Sadak wrote:
> thanks dave
Hi Terry
If you are having files smaller than hdfs block size and if you are using
Default TextInputFormat with the default properties for split sizes there
would be just one file per mapper.
If you are having larger file sizes, greater than the size of a hdfs block.
Please take a look at a sampl
I get those too. It's probably this issue:
https://issues.apache.org/jira/browse/MAPREDUCE-2374
If you go look in the task tracker log on the node that had the failed
task you'll probably see that "Text File Busy" message in there.
Looks like this is fixed in newer releases of CDH.
HTH,
DR
Hi Dave
Can you post the task logs corresponding to this. You can browse the web Ui
till the failed task log, it'll conatin more information to help you
analyze task failure reasons.
On Mon, Oct 8, 2012 at 5:58 PM, Dave Shine <
dave.sh...@channelintelligence.com> wrote:
> I’m starting to see th
If it is only to get a parse overview of the run you can use counters. But
you shouldn't overuse them.
You can for example count exceptions by type (but by message is not a good
approach unless you are sure that the message is constant.)
Regards
Bertrand
On Mon, Oct 8, 2012 at 4:13 PM, Terry Hea
thanks Bejoy.
...Feeling a bit foolish as Tom White's book was 2 feet away
On 10/08/2012 10:28 AM, Bejoy Ks wrote:
> Hi Terry
>
> If you are having files smaller than hdfs block size and if you are
> using Default TextInputFormat with the default properties for split
> sizes there would be j
Visioner-
I hope it is not heresy to mention here, but I believe the MapR
implementation of Hadoop supports symlink via NFS.
-Terry
On 10/08/2012 10:20 AM, Visioner Sadak wrote:
> thanks dave its in hdfs onlyany other methods of creating a symlink
>
> On Mon, Oct 8, 2012 at 7:00 PM, Dave Be
You can create an HDFS symlink by using the FileContext#createSymlink
function. I don't think this can be done through the "hadoop fs"
command, so you're going to have to write some Java code to do this.
We should consider adding this functionality to the "hadoop fs"
command in the future.
Colin
Thank you for the comments. Some similar frameworks I looked at
include Haloop, Twister, Hama, Giraph and Cascading. I am also doing
large scale graph processing so I assumed one of them could serve the
purpose. Here is a summary of what I found out about them that is
relevant:
1) Haloop and Twist
Well I was thinking ...
Map -> Combiner -> Reducer -> Identity Mapper -> combiner -> reducer ->
Identity Mapper -> combiner -> reducer...
May make things easier.
HTH
0Mike
On Oct 8, 2012, at 2:09 PM, Jim Twensky wrote:
> Thank you for the comments. Some similar frameworks I looked at
> in
Hi Mike,
I'm already doing that but the output of the reduce goes straight back
to HDFS to be consumed by the next Identity Mapper. Combiners just
reduce the amount of data between map and reduce whereas I'm looking
for an optimization between reduce and map.
Jim
On Mon, Oct 8, 2012 at 2:19 PM,
It seems that MapReduce allows the input path for a job to be a string
containing regex patterns. Yet, Path.normalizePath() will replace \\ with /.
So how should one escape the regex special characters if needed when setting a
job's input paths?
Jane Chen
Lead Engineer
MarkLogic Corporation
With secure hadoop the user name is authenticated by the kerberos server.
But what about the groups that the user is a member of? Are these simple
the groups that the user is a member of on the namenode machine?
Is it viable to manage access to files on HDFS using groups on a secure
hadoop cluster?
> asking for. If anyone who used Hama can point a few articles about how
> the framework actually works and handles the messages passed between
> vertices, I'd really appreciate that.
Hama Architecture:
https://issues.apache.org/jira/secure/attachment/12528219/ApacheHamaDesign.pdf
Hama BSP progra
P.S., giraph is different in the sense that it runs as a map-only job.
On Tue, Oct 9, 2012 at 7:45 AM, Edward J. Yoon wrote:
>> asking for. If anyone who used Hama can point a few articles about how
>> the framework actually works and handles the messages passed between
>> vertices, I'd really ap
Jim,
You can use the combiner as a reducer albeit you won't get down to a single
reduce file output. But you don't need that.
As long as the output from the combiner matches the input to the next reducer
you should be ok.
Without knowing the specifics, all I can say is TANSTAAFL that is to sa
Mike, just FYI, it's my 08's approach[1].
1. https://blogs.apache.org/hama/entry/how_will_hama_bsp_different
On Tue, Oct 9, 2012 at 7:50 AM, Michael Segel wrote:
> Jim,
>
> You can use the combiner as a reducer albeit you won't get down to a single
> reduce file output. But you don't need that.
Koert,
If you use the org.apache.hadoop.security.ShellBasedUnixGroupsMapping
class (via hadoop.security.group.mapping), then yes the NameNode's
view of the local unix groups (and the primary group) of the user is
the final say on what groups the user belongs to. This can be relied
on - but note th
Hi Koert,
Another option is to use the LdapGroupsMapping which picks up the group
membership from a LDAP directory.
You can find more details on the JIRA issue:
https://issues.apache.org/jira/browse/HADOOP-8121
Up to now, it is available for ActiveDirectory and released in
hadoop-2.0.0-alpha and n
Hi Harsh, thank you for your replay.
In HDFS-2246, when Local DataNode is dead, the DFSClient still can read
data form local file. I think that may lead to DFSCient read wrong data.
HDFS-347 use domain socket to implement Local Read, when Local DataNode is
dead, can DFSClient read data from local
Thanks Colin I tried using FileContext but the class is showing as
depricated
On Tue, Oct 9, 2012 at 12:02 AM, Colin McCabe wrote:
> You can create an HDFS symlink by using the FileContext#createSymlink
> function. I don't think this can be done through the "hadoop fs"
> command, so you're going
34 matches
Mail list logo