What does the job cleanup task do? My understanding was it just cleaned up
any intermediate/temporary files and moved the reducer output to the output
directory? Does it do more?
One of my jobs runs, all maps and reduces finish, but then the job cleanup
task never finishes. Instead it gets kill
://hadoop.apache.org/docs/stable/api/org/apache/hadoop/mapred/JobContext.html#getProgressible%28%29
On Sat, Apr 13, 2013 at 2:35 PM, Robert Dyer wrote:
> What does the job cleanup task do? My understanding was it just cleaned
> up any intermediate/temporary files and moved the reducer output to the
&g
It isn't GPL. OpenJDK[1] is GPLv2 with a Classpath Exception[2] (which is
important).
Read more here:
http://programmers.stackexchange.com/questions/52534/can-we-use-java-for-commercial-use
Also note that Hadoop[3] is licensed under Apache v2[4].
[1] http://openjdk.java.net/legal/
[2] http://op
You can, however note that unless you also run a TaskTracker on that node
(bad idea) then any blocks that are replicated to this node won't be
available as input to MapReduces and you are lowering the odds of having
data locality on those blocks.
On Tue, May 14, 2013 at 2:01 AM, Ramya S wrote:
>
I recently upgraded from 1.0.4 to 1.1.2. Now however my HDFS won't start
up. There appears to be something wrong in the edits file.
Obviously I can roll back to a previous checkpoint, however it appears
checkpointing has been failing for some time and my last check point is
over a month old.
Is
Actually, 1.2.1 is out (and marked stable). I see no reason not to upgrade.
http://hadoop.apache.org/docs/r1.2.1/releasenotes.html
As far as performance goes, when I upgraded our cluster from 1.0.4 to
1.1.2, our small jobs (that took about 1 min each) were taking about 20-30s
less time. So ther
I just noticed the job status for MR jobs tends to show 0's in the Map and
Reduce columns but actually shows the totals correctly.
I am not sure exactly when this started happening, but this cluster was
upgraded from Hadoop 1.0.4 to 1.1.2 and now to 1.2.1. It definitely worked
fine on 1.0.4, but
/browse/MAPREDUCE-5376) and attached a
> patch.
> But it is not fixed by the current release.
>
> Thanks,
> Shinichi
>
> (2013/09/03 11:20), Robert Dyer wrote:
>
> I just noticed the job status for MR jobs tends to show 0's in the Map
>
> and Reduce columns bu
I recently setup a 2.2.0 test cluster. For some reason, all of my MR jobs
are failing. The maps and reduces all run to completion, without any
errors. Yet the app is marked failed and there is no final output. Any
ideas?
Application Type: MAPREDUCE
State: FINISHED
FinalStatus: FAILED
Diagnosti
896 INFO [main] org.apache.hadoop.mapred.Task: Task
'attempt_1382415258498_0001_m_14_0' done.
On Tue, Oct 22, 2013 at 12:16 AM, Arun C Murthy wrote:
> If you follow the links on the web-ui to the logs of the map/reduce tasks,
> what do you see there?
>
> Arun
>
> On Oct 21, 2
So does anyone have any ideas how to track this down?
Is it perhaps an exception somewhere in an output committer that is being
swallowed and not showing up in the logs?
On Tue, Oct 22, 2013 at 2:19 AM, Robert Dyer wrote:
> The logs for the maps and reduces show nothing useful. There ar
Thanks Sandy! These seem helpful!
"MapReduce cluster configuration options have been split into YARN
configuration options, which go in yarn-site.xml; and MapReduce
configuration options, which go in mapred-site.xml. Many have been given
new names to reflect the shift. ... *We’ll follow up with a
Is there an easy way to get the uncompressed size of a sequence file that
is block compressed? I am using the Snappy compressor.
I realize I can obviously just decompress them to temporary files to get
the size, but I would assume there is an easier way. Perhaps an existing
tool that my search d
916)
On Sat, Nov 23, 2013 at 3:14 PM, Robert Dyer wrote:
> Is there an easy way to get the uncompressed size of a sequence file that
> is block compressed? I am using the Snappy compressor.
>
> I realize I can obviously just decompress them to temporary files to get
> the size, bu
I recently restarted my small cluster (1 namenode, 1 job tracker, 1
secondary nn, 6 compute/data nodes). However the namenode refuses to
start up due to a NPE.
After googling I saw some suggestions of doing a printf
"\xff\xff\xff\xee\xff" > edits, however this did not fix the problem.
Any ideas
I am currently using a SequenceFile as input to my MR job (on Hadoop
1.0.3). This works great, as my input is just a bunch of binary
blobs.
However it seems SequenceFile is only intended to append new data and
never update existing entries. Is that correct?
If so, would i be better off moving t
I have been reading up on HBase and my understanding is that the
physical files on the HDFS are split first by region and then by
column families.
Thus each column family has its own physical file (on a per-region basis).
If I run a MapReduce task that uses the HBase as input, wouldn't this
imply
one replica of the blocks on
> the datanode the same machine as the client (i.e. the regionserver from hdfs
> point of view).
>
> N.
>
>
>
> On Wed, Aug 29, 2012 at 6:20 AM, Robert Dyer wrote:
>>
>> I have been reading up on HBase and my understanding is that the
>
#x27;re right" :-).
> It's documented here:
> http://hbase.apache.org/book.html#regions.arch.locality
>
> On Wed, Aug 29, 2012 at 8:06 AM, Robert Dyer wrote:
>>
>> Ok but does that imply that only 1 of your compute nodes is promised
>> to have all of the data
_
> Disclaimer:This email and any attachments are sent in strictest confidence
> for the sole use of the addressee and may contain legally privileged,
> confidential, and proprietary data. If you are not the intended recipient,
> please advise the sender by replying promptly to this email and then delete
> and destroy this email and any attachments without any further use, copying
> or forwarding
>
>
--
Robert Dyer
rd...@iastate.edu
If the file is pre-sorted, why not just make multiple sequence files -
1 for each split?
Then you don't have to compute InputSplits because the physical files
are already split.
On Tue, Sep 11, 2012 at 11:00 PM, Harsh J wrote:
> Hey Jason,
>
> Is the file pre-sorted? You could override the Outpu
Hi Manoj,
If the data is the same for both tests and the number of mappers is
fewer, then each mapper has more (uncompressed) data to process. Thus
each mapper should take longer and overall execution time should
increase.
As a simple example: if your data is 128MB uncompressed it may use 2
mapp
Has anyone experienced a TaskTracker/DataNode behaving like the attached
image?
This was during a MR job (which runs often). Note the extremely high
System CPU time. Upon investigating I saw that out of 64GB ram the system
had allocated almost 45GB to cache!
I did a sudo sh -c "sync ; echo 3 >
ch your
> MR job again.
> Can you share your logs in pastebin?
>
>
> On Sat 08 Dec 2012 07:09:02 PM CST, Robert Dyer wrote:
>
>> Has anyone experienced a TaskTracker/DataNode behaving like the
>> attached image?
>>
>> This was during a MR job (which
se and
> slowdown. drop_caches doesn't have any impact on correctness; it won't
> cause data loss (by dropping a dirty page or whatever). I've had sync
> calls take 10 minutes to complete, so the unnecessary impact can be
> significant.
>
> -andy
>
> On Sat, Dec
unnecessary impact can be
> significant.
>
> -andy
>
> On Sat, Dec 8, 2012 at 4:09 PM, Robert Dyer wrote:
> > Has anyone experienced a TaskTracker/DataNode behaving like the attached
> > image?
> >
> > This was during a MR job (which runs often). Note the extr
ly want paged
> out in favor of a larger filesystem cache.
> >
> > There is also a vm parameter that controls the minimum size of the free
> chain, might want to increase that a bit.
> >
> > Also, look into hosting your JVM heap on huge pages, they can't be p
I actually have this exact same error. After running my namenode for
awhile (with a snn), it gets to a point where the snn starts crashing and
if I try to restart the NN I will get this problem. I typically wind up
having to go with a much older copy of the image and edits files in order
to get i
You could create a CustomOutputCommitter and in the commitJob() method
simply read in the part-* files and write them out into a single aggregated
file.
This requires making a CustomOutputFormat class that uses the
CustomOutputCommittter and then setting that
via job.setOutputFormatClass(CustomOut
?
>
> --
> Jay Vyas
> http://jayunit100.blogspot.com
>
> --
>
> Robert Dyer
> rd...@iastate.edu
>
Forgot to mention: Hadoop 1.0.4
On Sat, Feb 16, 2013 at 2:38 PM, Robert Dyer wrote:
> I am at a bit of wits end here. Every single time I restart the namenode,
> I get this crash:
>
> 2013-02-16 14:32:42,616 INFO org.apache.hadoop.hdfs.server.common.Storage:
> Image file of siz
an easy
way to monitor (other than a script grep'ing the logs) the checkpoints to
see when this happens?
On Sat, Feb 16, 2013 at 2:39 PM, Robert Dyer wrote:
> Forgot to mention: Hadoop 1.0.4
>
>
> On Sat, Feb 16, 2013 at 2:38 PM, Robert Dyer wrote:
>
>> I am at a bi
ose
> the fsimage file.
>
> Warm Regards,
> Tariq
> https://mtariq.jux.com/
> cloudfront.blogspot.com
>
>
> On Mon, Feb 18, 2013 at 3:31 AM, Robert Dyer wrote:
>
>> It just happened again. This was after a fresh format of HDFS/HBase and
>> I am attemptin
/hdfs'.
> Warm Regards,
> Tariq
> https://mtariq.jux.com/
> cloudfront.blogspot.com
>
>
> On Mon, Feb 18, 2013 at 3:31 AM, Robert Dyer wrote:
>
>> It just happened again. This was after a fresh format of HDFS/HBase and
>> I am attempting to re-import the (backed
> P.s. How exactly are you shutting down the NN each time? A kill -9 or
> a regular SIGTERM shutdown?
>
I shut down the NN with 'bin/stop-dfs.sh'.
> On Mon, Feb 18, 2013 at 4:31 AM, Robert Dyer wrote:
> > On Sun, Feb 17, 2013 at 4:41 PM, Mohammad Tariq
> wrote:
&
I have a small 6 node dev cluster. I use a 1GB SequenceFile as input to a
MapReduce job, using a custom split size of 10MB (to increase the number of
maps). Each map call will read random entries out of a shared MapFile
(that is around 50GB).
I set replication to 6 on both of these files, so all
done over a local socket as
> well, and may appear in network traffic observing tools too (but do
> not mean they are over the network).
>
> On Mon, Feb 25, 2013 at 2:35 AM, Robert Dyer wrote:
> > I have a small 6 node dev cluster. I use a 1GB SequenceFile as input to
> a
> &
t
the short circuit.
Now I see no network utilization for this job and it runs *much* faster (13
mins instead of 2+ hours)! Problem solved! :-)
Thanks Harsh!
On Mon, Feb 25, 2013 at 1:41 AM, Robert Dyer wrote:
> I am using Ganglia.
>
> Note I have short circuit reads enabled (I
Generally speaking, static fields are not useful in Hadoop.
The issue you are seeing is that the reducer is running in a separate VM
(possibly on a different node!) and thus the static value you are reading
inside of Mid is actually a separate instantiation of that class and field.
If you have an
39 matches
Mail list logo