Hello everyone,
I am going thru source code of MapReduce. In MergeQueue.merge, I can only
see the SEGMENTS are combined and sorted by length into a list for merge.
However, I could not find the procedure to sort those (key, value) in
segments by key...
here is the function:
1. RawKeyValueIter
Hello fellas,
problem here. I am trying to get some runtime info in FileInputFormat.java
and I try to use LOG.info(). However it never seems write anything into log
file. I tried to grep the "logs" directory and could not find the info
supposed to be printed out.
Any idea on this?
BTW, any other
Yu wrote:
> Please note this:
>
> private static class MergeQueue
> extends PriorityQueue> implements RawKeyValueIterator {
>
> priority queue is used to accomplish sorting.
>
> On Fri, Jun 18, 2010 at 8:14 PM, elton sky wrote:
>
> > Hello everyone,
&g
TY
> through snmp metrics by using snmp4j
>
> On Fri, Jun 18, 2010 at 8:15 PM, elton sky wrote:
>
> > Hello fellas,
> >
> > problem here. I am trying to get some runtime info in
> FileInputFormat.java
> > and I try to use LOG.info(). However it never seems wr
hello everyone,
I noticed there are 6 operations in HDFS:
OP_WRITE_BLOCK
OP_READ_BLOCK
OP_READ_METADATA
OP_REPLACE_BLOCK
OP_COPY_BLOCK
OP_BLOCK_CHECKSUM
and As I know there's no way to modify some arbitrary part in a existing
file in HDFS. So what if I create a say, 2 Petabytes, file and like to
Hi,
I am new to hadoop programming. I am trying to copy a local file to HDFS. My
code snippet is:
.
.
Configuration conf = new Configuration();
InputStream in=null;
OutputStream out = null;
try {
in = new BufferedInputStream(new FileInputStream(src));
Thanks for reply,
On Thu, Jun 24, 2010 at 7:35 PM, Steve Loughran wrote:
> elton sky wrote:
>
>> Hi,
>> I am new to hadoop programming.
>>
>
>
> OK. First hint of advice: don't spam the dev lists with user problems, you
> will only get ignored
Hello,
I am trying some simple code snippet to create a new file. And after create
and write to the file, I want to use "sync()" to synchronize all replicas.
However, I got "LeaseExpiredException" in FSNameSystem.checkLease():
my code:
.
.
InputStream in=null;
OutputStream out = null;
Hey Ted,
>The line numbers don't match those from hadoop 0.20.2
>What version are you using ?
I am using 0.20.2. I add some extra LOG or print lines in between which
makes the line number different from usual~ But my modification are just
print out for debugging this problem.
>I don't see how co
hello,
I am trying to separate my 6 nodes onto 2 different racks.
For test purpose, I wrote a bash file which smply returns "rack0" all the
time. And I add property "topology.script.file.name" in core-site.xml.
When I restart by start-dfs.sh, the namenode could not find any datanode at
all. All d
hello all,
I was running a test in hadoop source code,
test/org/apache/hadoop/mapred/NotificationTestCase.java. I ran it with
TestClusterMRNotification.
In my driver main() function:
"
TestClusterMRNotification t = new TestClusterMRNotification();
t.setUp();
t.testMR();
t.tearDown();
"
This pr
I am a bit confused of how this attribute is used.
My understanding is it's related with file read/write. And I can see, in
LineReader.java, it's used as the default buffer size for each line; in
BlockReader.newBlockReader(), it's used as the internal buffer size of the
BufferedInputStream. Also,
I think my question is ignored, so just post it again:
I am a bit confused of how this attribute is used.
My understanding is it's related with file read/write. And I can see, in
LineReader.java, it's used as the default buffer size for each line; in
BlockReader.newBlockReader(), it's used as the
Hello everyone,
I am doing some evaluation on my 6 nodes mini cluster. Each node has 4 core
Intel(R) Xeon(R) CPU 5130 @ 2.00GHz, 8GB memory, 500 GB disk, running Linux
version 2.6.18-164.11.1.el5(Red Hat 4.1.2-46).
I was trying to use different packet size (dfs.write.packet.size) and
bytePerChun
Hello experts,
I was benchmarking sequential write throughput of HDFS.
For testing affect of bytesPerChecksum (bpc) size to write performance, I am
using different bpc size: 2M, 256K, 32K, 4K, 512B.
My cluster has 1 name node and 5 data nodes. They are xen VMs and each of
them configured with 56
Hello,
I was benchmarking write/read of HDFS.
I changed the chunksize, i.e. bytesPerChecksum or bpc, and create a 1G file
with 128MB block size. The bpc I used: 512B, 32KB, 64KB, 256KB, 512KB, 2MB,
8MB.
The result surprised me. The performance for 512B, 32KB, 64KB are quite
similar, and then, as
I always have this question but couldn't find proper answer for this. For
system level applications, c/c++ is preferable. But why this one using java?
Hello,
In
hdfs.org.apache.hadoop.hdfs.DFSClient
.DFSOutputStream.writeChunk(byte[]
b, int offset, int len, byte[] checksum)
The second last line:
int psize = Math.min((int)(blockSize-bytesCurBlock), writePacketSize);
When I use blockSize bigger than 2GB, which is out of the boundary of
integer
>Why would you want to use a block size of > 2GB?
For keeping a maps input split in a single block~
On Tue, Oct 19, 2010 at 9:07 AM, Michael Segel wrote:
>
> Ok, I'll bite.
> Why would you want to use a block size of > 2GB?
>
>
>
> > Date: Mon, 18 Oct 2010 21:33:34 +1100
> > Subject: BUG: Anyone
I am curious, any specific reason to make it smaller than 2**31?
On Tue, Oct 19, 2010 at 10:27 AM, Owen O'Malley wrote:
> Block sizes larger than 2**31 are known to not work. I haven't ever tracked
> down the problem, just set my block size to be smaller than that.
>
> -- Owen
>
Hello,
I am trying to build mumak and have a bite, with "ant package" in mapred/
dir.
I got error:
[ivy:resolve] :: problems summary ::
[ivy:resolve] WARNINGS
[ivy:resolve] module not found: org.apache.hadoop#hadoop-common;0.21.0
[ivy:resolve] apache-snapshot: tried
[ivy:resolve]
https:
Milind,
You are right. But that only happens when your client is one of the data
nodes in HDFS. otherwise a random node will be picked up for the first
replica.
On Fri, Oct 22, 2010 at 3:37 PM, Milind A Bhandarkar
wrote:
> If a file of say, 12.5 GB were produced by a single task with replication
d come up again.
I ll try the blockdev as well, see if any improve though
On Thu, Nov 11, 2010 at 6:32 AM, Scott Carey wrote:
>
> On Oct 7, 2010, at 2:35 AM, elton sky wrote:
>
> > Hello experts,
> >
> > I was benchmarking sequential write throughput of
Hello forumers,
Before spill the data in kvbuffer to local disk in map task, k/v are
sorted using quick sort. The complexity of quick sort is O(nlogn) and
worst case is O(n^2).
Why using quick sort?
Regards
Well,
I think you are asking :
if you have 3 machines, and you want to start 3 maps, one for each
input file, will each maps reside on each different machine?
The answer is: not necessarily.
In responding heartbeat from a task tracker, task scheduler tries to
assign one local map task for each job
check the FAQ (
http://wiki.apache.org/hadoop/FAQ#What_does_.22file_could_only_be_replicated_to_0_nodes.2C_instead_of_1.22_mean.3F
)
On Tue, Apr 5, 2011 at 4:53 PM, Guy Doulberg wrote:
> Hey guys,
>
> We are trying to figure out why many of our Map/Reduce job on the cluster
> are failing.
> In lo
Agree with Harsh,
I think you need to write your own RecordRead.
On Tue, Apr 5, 2011 at 3:37 PM, Harsh Chouraria wrote:
> Hello Kevin,
>
> On Fri, Mar 25, 2011 at 12:52 AM, wrote:
> > -Dstream.map.output.field.separator= \
> > -Dstream.num.map.output.key.fields=13 \
> >
> > I have searched th
One of assumptions map reduce made, I think, is that size of map's output is
smaller than input. Although we can see many applications have the same size
of output with input, like, sort, merge,etc.
For my benchmark purpose, I am looking for some non-trivial, real life
applications which creates *b
kload for spill and merge on map's
local disk is heavy.
-Elton
On Sat, Apr 30, 2011 at 11:22 AM, Owen O'Malley wrote:
> On Fri, Apr 29, 2011 at 5:02 AM, elton sky wrote:
>
> > For my benchmark purpose, I am looking for some non-trivial, real life
> > applications which
I got 2 questions:
1. I am wondering how hadoop MR performs when it runs compute intensive
applications, e.g. Monte carlo method compute PI. There's a example in 0.21,
QuasiMonteCarlo, but that example doesn't use random number and it generates
psudo input upfront. If we use distributed random num
Ted,
MPI supports node-to-node communications in ways that map-reduce does not,
> however, which requires that you iterate map-reduce steps for many
> algorithms. With Hadoop's current implementation, this is horrendously
> slow (minimum 20-30 seconds per iteration).
>
> Sometimes you can avoid
hello all,
I am trying to keep the output and copied files for reduce tasks after a job
finishes.
I commented out many "remove", "delete" kind of code, from TaskTracker,
Task, etc, but still can not keep them.
Any idea?
intensive in map phase, for sampling all
possible combinations of items.
I am still looking for more applications, which creates bigger output and
not CPU bound.
Any further idea? I appreciate.
On Tue, May 3, 2011 at 3:10 AM, Steve Loughran wrote:
> On 30/04/2011 05:31, elton sky wr
"this"
> "this" "is"
> "is" "a"
> "a" "test"
> ...
>
>
> You may also be extracting all kinds of other features form the text, but
> the tokenization/n-gram is not that CPU intensive.
>
> --Bo
I am looking for applications can be implemented in MR, for which, in order
to get final result, only partial dataset is required to be processed. (i.e.
don't have to process the whole dataset).
The Only thing I can think about is search top K records (without ordering),
like in sql query.
Is the
hello everyone,
As I don't have experience with big scale cluster, I cannot figure out why
the inter-rack communication in a mapreduce job is "significantly" slower
than intra-rack.
I saw cisco catalyst 4900 series switch can reach upto 320Gbps forwarding
capacity. Connected with 48 nodes with 1Gb
28 PM, Steve Loughran wrote:
> On 06/06/11 08:22, elton sky wrote:
>
>> hello everyone,
>>
>> As I don't have experience with big scale cluster, I cannot figure out why
>> the inter-rack communication in a mapreduce job is "significantly" slower
>> tha
andwidth to push all of those ports at full tilt between
> racks. That's why Hadoop has the ability to take advantage of rack
> locality. It will try to schedule I/O local to a rack where it's less
> likely to block.
>
> -Joey
>
> On Mon, Jun 6, 2011 at 7:04 AM, elton sky
Hi John,
Because for map task, job tracker tries to assign them to local data nodes,
so there' not much n/w traffic.
Then the only potential issue will be, as you said, reducers, which copies
data from all maps.
So in other words, if the application only creates small intermediate
output, e.g. gre
text processing purpose. (i.e. few stages, low dependency,
> > highly parallel).
> >
> > Its when one tries to solve general purpose algorithms of modest
> > complexity that map/reduce gets into I/O churning problems.
> >
> > On Mon, 6 Jun 2011 23:58:53 +1000, elton
40 matches
Mail list logo