Hey all,
Just a note that you should avoid upgrading your clusters to 1.6.0u18.
We've seen a lot of segfaults or bus errors on the DN when running
with this JVM - Stack found the ame thing on one of his clusters as
well.
We've found 1.6.0u16 to be very stable.
-Todd
Hi Steve,
On Sun, Feb 14, 2010 at 12:11 PM, Steve Kuo kuosen...@gmail.com wrote:
I am running a hadoop job that combines daily results with results with
previous days. The reduce output is lzo compressed and growing daily in
size.
- DistributedLzoIndexer is used to index lzo files to
On a small cluster, I'm of the opinion that a value less than 3 would
actually be useful in reducing job startup time a little bit.
https://issues.apache.org/jira/browse/MAPREDUCE-1266
The issue got stalled a bit. If you want it, pipe up on the JIRA :)
Especially if you have hard data indicating
Hi Vasilis,
Two things:
1) You're missing a matching } in your hadoop.tmp.dir setting
2) When you use ${hadoop.tmp.dir}/dfs/data, it does a literal string
interpolation. Thus, it's not adding dfs/data to each of the
hadoop.tmp.dir directories, but rather just the last one.
I'd recommend setting
On Tue, Feb 9, 2010 at 11:37 AM, Edward Capriolo edlinuxg...@gmail.com wrote:
With the setting of 5 each tasktracker checks into the jobtracker
every 5 seconds. The concept is that with enough TaskTrackers , say a
1000 node cluster 1000/5= 200 will be checking in at and given second.
Actually
Yes, you're likely to see an error in the DN log. Do you see anything
about max number of xceivers?
-Todd
On Thu, Feb 4, 2010 at 11:42 PM, Meng Mao meng...@gmail.com wrote:
not sure what else I could be checking to see where the problem lies. Should
I be looking in the datanode logs? I looked
Why not try downloading the load gen source from 0.19 or 0.20,
compiling against 0.18, and seeing if there are any errors? If there
are, fix them. The APIs haven't changed much in a long time, it
shouldn't be a lot of work.
-Todd
On Fri, Feb 5, 2010 at 5:50 AM, Ashish Pareek pareek...@gmail.com
Hi Martin,
Not sure what you mean - why would it be faster to split it into two
threads? Keep in mind that there is a TCP send buffer so if the client
is reading faster than the disk, the server's sends won't block
anyway.
-Todd
On Fri, Feb 5, 2010 at 8:03 AM, Martin Mituzas
Hi Stas,
Hadoop doesn't daemonize itself. The shell scripts use nohup and a lot
of bash code to achieve a similar idea.
-Todd
On Thu, Feb 4, 2010 at 1:03 PM, Stas Oskin stas.os...@gmail.com wrote:
Hi.
Just wondering - does anyone know what framework Hadoop uses for
daemonizing?
Any chance
On Thu, Feb 4, 2010 at 1:21 PM, Stas Oskin stas.os...@gmail.com wrote:
Hi Todd.
Hadoop doesn't daemonize itself. The shell scripts use nohup and a lot
of bash code to achieve a similar idea.
Was there any design decision behind this approach?
It long predates my involvement in the
Hi Jim,
Sorry about the broken links. We just launched a new website a couple days
ago and a few of the pages are still in transition.
This link should help you get started:
http://archive.cloudera.com/docs/cdh2-pseudo-distributed.html
Thanks
-Todd
On Wed, Feb 3, 2010 at 11:07 AM, Jim Kusznir
Hi Vasilis,
Did you make sure to ant clean before rebuilding hadoop-lzo if you updated
the code? Also, can you paste your configuration for io.compression.codecs ?
Thanks
-Todd
On Tue, Feb 2, 2010 at 9:09 AM, Vasilis Liaskovitis vlias...@gmail.comwrote:
Hi,
I am trying to use hadoop-0.20.1
Hi Bradford,
My guess is that somewhere it's using hdfs://localhost:8020/ and elsewhere
using hdfs://localhost/ (and letting the default port get used). It's
certainly a bug that this causes a problem, but if you want to work around
it I'd just make sure you've used the fully specified
On Sun, Jan 17, 2010 at 6:19 PM, Tarandeep Singh tarand...@gmail.comwrote:
On Sun, Jan 17, 2010 at 1:57 PM, Vladimir Klimontovich
klimontov...@gmail.com wrote:
Maybe, hadoop running MR jobs using different user? For example, if you
followed installation instructions
from official site
...
[exec] configure: error: C compiler cannot create executables
On Fri, Jan 15, 2010 at 12:09 PM, Todd Lipcon t...@cloudera.com wrote:
Hi Ted,
Did you also install liblzo-devel?
Here are the packages I install for LZO:
lzo-2.02-2.el5.1
lzo-devel-2.02-2.el5.1
-Todd
On Tue
create executables
See `config.log' for more details.
On Fri, Jan 15, 2010 at 1:17 PM, Todd Lipcon t...@cloudera.com wrote:
Are you starting from a clean tarball of the lzo stuff?
Can you make sure your /tmp/ partition isn't full?
There should be a config.log file hanging around somewhere
Hi David,
The ability to administratively blacklist TaskTrackers wasn't added until
0.21, which is not yet released.
In general, there isn't usually a great reason to do this -- you can take
down a tasktracker by just stopping it. The purpose of the HDFS-side
operation is to decommission a node
returned: 2
Has anybody seen the above ?
Thanks
On Mon, Jan 11, 2010 at 3:34 PM, Todd Lipcon t...@cloudera.com wrote:
Hi Ted,
You need to install liblzo from EPEL:
http://fr.rpmfind.net/linux/RPM/Extras_Packages_for_Enterprise_Linux.html
-Todd
On Mon, Jan 11, 2010 at 3:21 PM
[exec] See `config.log' for more details.
BUILD FAILED
/home/rialto/kevinweil-hadoop-lzo-916aeae/build.xml:243: exec returned: 77
On Tue, Jan 12, 2010 at 10:25 AM, Todd Lipcon t...@cloudera.com wrote:
Hi Ted,
Please make sure you have version 2.02 of liblzo installed. There's
Hi Ted,
You need to install liblzo from EPEL:
http://fr.rpmfind.net/linux/RPM/Extras_Packages_for_Enterprise_Linux.html
-Todd
On Mon, Jan 11, 2010 at 3:21 PM, Ted Yu yuzhih...@gmail.com wrote:
Can someone tell me how I can install liblzo ?
[r...@tyu-linux lzo-2.03]# uname -a
Linux
http://catb.org/~esr/faqs/smart-questions.html#id383250
Please go through the above explanation of how to ask questions on a mailing
list, and repost your question. Thanks in advance.
On Sat, Jan 2, 2010 at 8:35 AM, Ravi ravindra.babu.rav...@gmail.com wrote:
import java.lang.Integer;
import
Hi Himanshu,
Sounds like your mapred.local.dir doesn't have enough space. My guess is
that you've configured it somewhere inside /tmp/. Instead you should spread
it across all of your local physical disks by comma-separating the
directories in the configuration. Something like:
property
Hi Thomas (and Debian developers),
My responses inline below:
On Wed, Dec 30, 2009 at 10:53 AM, Thomas Koch tho...@koch.ro wrote:
Hi,
today I tried to run the cloudera debian dist on a 4 machine cluster. I
still
have some itches, see my list below. Some of them may require a fix in the
Furthermore, Text is meant for use when you have a UTF8-encoded string.
Creating a Text object from a byte array that is not proper UTF-8 is likely
to result in some kind of exception or data mangling. You should use
BytesWritable for this purpose
-Todd
2009/12/28 Edward Capriolo
How long does the checkpoint take? It seems possible to me that if the 2NN
checkpoint takes longer than the interval, it's possible that multiple
checkpoints will overlap and might trigger this. (this is conjecture, so
definitely worth testing)
-Todd
On Wed, Dec 23, 2009 at 6:38 PM, Jason Venner
On Wed, Dec 23, 2009 at 6:55 AM, Jason Venner jason.had...@gmail.comwrote:
If your jobs are launched by separate jvm instances, the only real
persistence framework you have is hdfs.
You have to basic choices:
1. Write a summary data to a persistent store, an hdfs file being a
simple
Hi Mark,
For what it's wroth, you're unlikely to see a big difference in performance
unless you cut down the number of spills from 1 to 1, or io.sort.factor to
io.sort.factor. The difference between 3 spills and 5 spills is not huge,
in my experience, since you're still writing the same amount of
On Wed, Dec 23, 2009 at 12:46 PM, Saptarshi Guha
saptarshi.g...@gmail.comwrote:
Hello,
I re-wrote MapFileOutputFormat for use with Hadoop 0.20.1 and have a
question.
Suppose my Map sends key-value pairs to the reducers.
In my reducer, for a given key value, i emit key1,value1, key2,value2,
Hi Starry,
The assignmultiple feature in the Fair Scheduler fixes this issue after
MAPREDUCE-706. This is not in any current releases, but we will be rolling
it into our next major release of Cloudera's distribution.
-Todd
P.S. Please don't cross-post questions to 4 lists. It's not necessary
Hi Jeff,
You can tune dfs.datanode.balance.bandwidthPerSec in order to change the
speed at which the balancer moves data around. It defaults to 1MB/sec so as
to avoid using significant cluster resources, but you could certainly bump
it up for an individual balancer run if you need it to go
The DataNode actually uses the sendfile call to do the block data transfers
(through FileChannel.transferTo). The FileInputStream is just used in order
to get to the FileChannel object.
-Todd
On Thu, Dec 17, 2009 at 6:40 PM, Martin Mituzas xietao1...@hotmail.comwrote:
I checked the DataNode
Hi Eguzki,
Is one of the tables vastly smaller than the other? If one is small enough
to fit in RAM, you can do this like so:
1. Add the small file to the DistributedCache
2. In the configure() method of the mapper, read the entire file into an
ArrayList or somesuch in RAM
3. Set the input path
Hi Fu-Ming,
Looks similar to this bug:
http://issues.apache.org/jira/browse/HDFS-686
Does this problem persist, or was it a one time occurrence?
-Todd
On Tue, Dec 15, 2009 at 5:42 PM, Fu-Ming Tsai sary...@gmail.com wrote:
Hello, all,
I tried to execute 2 SecondaryNamenode in my env.
Hi Erik,
A few things to try:
- does this FS store sensitive data or would it be possible to bzip2 the
files and upload them somewhere?
- can you add logging to the replay of FSEditLog so as to be aware of what
byte offset is causing the issue?
- DO take a backup of all of the state,
in the scalability of our system. I
will test the upper limits.
Thanks a lot.
Eguzki
Todd Lipcon escribió:
Hi Eguzki,
Is one of the tables vastly smaller than the other? If one is small enough
to fit in RAM, you can do this like so:
1. Add the small file to the DistributedCache
2. In the configure
On Thu, Dec 10, 2009 at 1:15 PM, Gang Luo lgpub...@yahoo.com.cn wrote:
Hi Todd,
I didn't change the partitioner, just use the default one. Will the default
partitioner cause the lost of the records?
-Gang
Do the maps output data nondeterministically? Did you experience any
task failures in
As always, Scott provided lots of great advice below. One note to be aware
of:
The fair scheduler assignmultiple feature in 0.20 doesn't do quite what
you think it might. It gives the ability to assign one map and one reduce
per TT heartbeat, but doesn't assign multiple map tasks in a single
Hi Nick,
My guess is that the tmp/ directories of the DNs were rather full. I've
occasionally seen this on clusters where writes have been failing.
There should be some kind of thread which garbage collects partial blocks
from the DN's tmp dirs, but it's not implemented, as far as I know. This
is temporary,
at least relative to me.
-Original Message-
From: Todd Lipcon [mailto:t...@cloudera.com]
Sent: Tuesday, December 08, 2009 12:48 PM
To: common-user@hadoop.apache.org
Subject: Re: some current features in hadoop
On Mon, Dec 7, 2009 at 10:58 PM, Krishna Kumar
krishna.ku
On Mon, Dec 7, 2009 at 10:58 PM, Krishna Kumar krishna.ku...@nechclst.inwrote:
Dear All,
Can anybody please let me know about some of the current features of
hadoop on which development work is going on / or planning to go in
future, like :
1. Record append
Not implemented and
Your mail came through empty - not sure if you meant to attach something,
but if you did it didn't appear. Try pastebin.org
On Mon, Dec 7, 2009 at 12:58 AM, pavel kolodin pavelkolodinhad...@gmail.com
wrote:
On Mon, 07 Dec 2009 04:54:16 -, Todd Lipcon t...@cloudera.com wrote:
grep clone
either.
Anyone seen something similar / possible fixes?
Thanks again
- Mikko
- Original Message -
*From:* Todd Lipcon t...@cloudera.com
*To:* common-user@hadoop.apache.org ; Mikko Lahtimikko.la...@pp1.inet.fi
*Sent:* Wednesday, December 02, 2009 9:03 PM
*Subject:* Re: Trouble
Certainly looks like HADOOP-6269 to me.
Can you try Cloudera's distribution?This patch is included.
-Todd
On Wed, Dec 2, 2009 at 4:23 AM, Arv Mistry a...@kindsight.net wrote:
Hi,
I've recently upgraded hadoop to 0.20 and am seeing this concurrent mod
exception on startup which I never got
that the problem sticks around, please report back or file a
JIRA.
Thanks
-Todd
Cheers Arv
-Original Message-
From: Todd Lipcon [mailto:t...@cloudera.com]
Sent: December 2, 2009 11:39 AM
To: common-user@hadoop.apache.org
Subject: Re: 0.20 ConcurrentModificationException
Certainly looks like
Hi Vasilis,
This is seen reasonably often, and could be partly due to missed
configuration changes. A few things to check:
- Did you increase the number of tasks per node from the default? If you
have a reasonable number of disks/cores, you're going to want to run a lot
more than 2 map and 2
No problem :) Also worth noting for anyone listening on that this feature is
not in 0.20.1 - it's been backported into CDH. It will arrive in 0.21.
Thanks
-Todd
On Wed, Dec 2, 2009 at 4:55 PM, james warren ja...@rockyou.com wrote:
Todd from Cloudera solved this for me on their company's forum.
Hi Derek,
You should set poolnameproperty to pool.name, not ${pool.name}
That should fix your issues.
-Todd
On Wed, Dec 2, 2009 at 7:46 PM, Derek Brown de...@media6degrees.com wrote:
I'm using Cloudera's distribution of 0.20.1, but this seems like a general
question to I'm posting here.
Hi Mark,
Both web UIs went down but jobs can still be submitted, etc? This seems like
a problem external to Hadoop, since the JT (port 50030) and NN (port 50070)
are entirely separate processes. If they both became inaccessible at the
same time, perhaps a firewall rule was added that
Any errors in your jobtracker log? Usually you'll see something there if the
scheduler fails to start.
What errors are the jobs failing with?
-Todd
On Mon, Nov 30, 2009 at 3:49 PM, Mike Kendall mkend...@justin.tv wrote:
no dice... and the default configuration from
where? copy paste the startup sequence and the job submission logs from the
JT log?
You gotta provide some details here :)
-Todd
On Mon, Nov 30, 2009 at 3:53 PM, Mike Kendall mkend...@justin.tv wrote:
java runtime error, exit code 1..
On Mon, Nov 30, 2009 at 3:52 PM, Todd Lipcon t
line. So, give the input:
Amy Sue Fred John
Jack Joe Sue John
Alice Bob Fred Sue John
The output should be:
Sue
John
because Sue and John appear on every line. I don't know Sue and John in
advance.
Thanks,
Jim
Todd Lipcon wrote:
Hi James,
Something like the following pseudocode
Also, keep in mind that, when you open a block for reading, the DN
immediately starts writing the entire block (assuming it's requested via the
xceiver protocol) - it's TCP backpressure on the send window that does flow
control there. So, although it's not explicitly reading ahead, most of the
On Tue, Nov 24, 2009 at 10:33 AM, Brian Bockelman bbock...@cse.unl.eduwrote:
On Nov 24, 2009, at 12:06 PM, Todd Lipcon wrote:
Also, keep in mind that, when you open a block for reading, the DN
immediately starts writing the entire block (assuming it's requested via
the
xceiver protocol
On Tue, Nov 24, 2009 at 10:35 AM, Raghu Angadi ang...@gmail.com wrote:
Sequential read is the simplest case and it is pretty hard to improve upon
the current raw performance (HDFS client does take more CPU than one might
expect, Todd implemented an improvement for CPU consumed).
Just to
Hi Mike,
I haven't seen that problem. There is one patch in the Cloudera distribution
that does modify the behavior of that method, though. Would you mind trying
this on the stock Apache 0.20.1 release? I see no reason to believe this is
the issue, since hundreds of other people are using our
fine. I just tried an Apache
release. I copied the example source file of WordCount.java to my jar and
submitted the job. It worked with Apache's release, but failed with
CloudEra's release. But my code failed for both of the versions.
From: Todd Lipcon t
, and these
reducers get over in (1 min 30 sec on avg).
Pankil
On Tue, Nov 17, 2009 at 5:07 PM, Todd Lipcon t...@cloudera.com
wrote:
On Tue, Nov 17, 2009 at 1:54 PM, Pankil Doshi forpan...@gmail.com
wrote:
With respect to Imbalanced data, Can anyone guide me how sorting
takes
()) {
iCount++;
- sValue += values.next().toString() + '\t';
+ sb.append(values.next().toString()).append('\t');
}
Hope that helps, Pankil.
-Todd
On Mon, Nov 23, 2009 at 9:32 PM, Todd Lipcon t...@cloudera.com wrote:
Interesting. I
On Tue, Nov 17, 2009 at 7:52 AM, Edward Capriolo edlinuxg...@gmail.comwrote:
Todd,
I think this is very important. From the grid on Hadoop the
Definative guide 78, it appears that bzip2 and zip are the only
formats the are splittable. As a result bzip2 would be my format of
choice to
/kevinweil/hadoop-lzo
http://github.com/kevinweil/hadoop-lzoMD
On Tue, Nov 17, 2009 at 8:08 AM, Todd Lipcon t...@cloudera.com wrote:
On Tue, Nov 17, 2009 at 7:52 AM, Edward Capriolo edlinuxg...@gmail.com
wrote:
Todd,
I think this is very important. From the grid on Hadoop
Coincidentally, we *just* posted a blog entry about this, courtesy of Kevin
Weil from Twitter:
http://www.cloudera.com/blog/2009/11/17/hadoop-at-twitter-part-1-splittable-lzo-compression/
-Todd
On Tue, Nov 17, 2009 at 8:36 AM, Todd Lipcon t...@cloudera.com wrote:
On Tue, Nov 17, 2009 at 8:33
On Tue, Nov 17, 2009 at 1:54 PM, Pankil Doshi forpan...@gmail.com wrote:
With respect to Imbalanced data, Can anyone guide me how sorting takes
place
in Hadoop after Map phase.
I did some experiments and found that if there are two reducers which have
same number of keys to sort and one
Message-
From: Todd Lipcon [mailto:t...@cloudera.com]
Sent: Monday, November 16, 2009 9:02 PM
To: common-user@hadoop.apache.org
Subject: Re: client.Client: failed to interact with node..ERROR
Hi Yair,
This looks like a katta-specific problem. Please direct your question to
the
katta
recommend moving back to 0.18.3 at this point for new
development. For new projects, unless you have need for the absolute most
stable release, I'd recommend 0.20.1, which is being used successfully in
production by many organizations.
-Todd
On Sun, Nov 15, 2009 at 8:47 PM, Todd Lipcon t
Hi Usman/Mike,
This feature is slated for 0.21 (not 0.20.1)
We have not backported it into Cloudera's release of 0.20.1, though we'll
certainly consider doing so if there appears to be demand for it in the
community. Anecdotally we've seen that not too many people are using bzip2
since the CPU
Hi Mark,
The simple answer is yes, to be safest, they should match.
In truth, the answer is a bit more complex. Since Java is dynamically
linked (classloaded) at runtime, as long as the method signatures and
class names you're using in your code haven't changed between versions, your
jar
, right?
Thank you,
Mark
On Sun, Nov 15, 2009 at 3:54 PM, Todd Lipcon t...@cloudera.com wrote:
Hi Mark,
The simple answer is yes, to be safest, they should match.
In truth, the answer is a bit more complex. Since Java is dynamically
linked (classloaded) at runtime, as long as the method
individually and start the daemons by hand. I do not
recommend this last option :)
sorry for all of the seemingly basic questions, but want to get it right
the first time:)
Sure thing- we're here to help.
-Todd
On Nov 9, 2009, at 1:11 PM, Todd Lipcon wrote:
On Mon, Nov 9, 2009 at 7:20 AM
On Mon, Nov 9, 2009 at 7:20 AM, John Martyniak j...@beforedawnsolutions.com
wrote:
Can the NameNode/DataNode JobTracker/TaskTracker run on a server that
isn't part of the cluster meaning I would like to run it on a machine that
wouldn't participate in the processing of data, and wouldn't
, 2009 at 12:15 PM, Todd Lipcon t...@cloudera.com wrote:
Hi,
No, there is currently no public-facing API for tracking job status aside
from parsing those JSPs (or adding your own)
-Todd
On Mon, Nov 9, 2009 at 5:28 AM, pnd prafulla.daw...@gmail.com wrote:
Hi i am new to hadoop
stand alone mode does not use any daemons - it's just a way of running
mapreduce in a single process for quick testing and getting started. You
need pseudo distributed mode if you want to run the daemons.
-Todd
On Thu, Nov 5, 2009 at 12:06 AM, Mohan Agarwal mohan.agarwa...@gmail.comwrote:
Hi ,
Hi Brian,
Any chance you are using the Cloudera distribution? We did accidentally ship
a bug like this which will be ameliorated in our next release.
The temporary workarounds are:
a) edit /usr/bin/hadoop and change the $* to a $@ (including the quotes!)
or
b) use
Hi Dennis,
This is normal to see pseudodistributed much slower than standalone.
Most of the extra overhead you're seeing is due to the heartbeat mechanism
of Hadoop. When the JobTracker has tasks that need to be run, it does not
actually push them to the TaskTrackers. Instead, the TaskTrackers
Hi John,
You can see a short slide deck from the July HUG here that includes some
info about what's new in 0.19 and 0.20:
http://cloudera-todd.s3.amazonaws.com/hug-20090917.pdf
-Todd
On Tue, Oct 20, 2009 at 1:44 AM, John Clarke clarke...@gmail.com wrote:
Hi,
I currently have an app written
Hi Schubert,
Regarding the Hive problem - you should be able to build against 0.20.0 and
then run against 0.20.1. Within a release (0.20.*) the APIs are completely
compatible, so you can compile any external application against one and run
against another.
Regarding the format of the md5 file,
On Fri, Oct 16, 2009 at 4:01 AM, tim robertson timrobertson...@gmail.comwrote:
Hi all,
Adding the following to core-site.xml, mapred-site.xml and
hdfs-site.xml (based on Cloudera guidelines:
http://tinyurl.com/ykupczu)
io.sort.factor: 15 (mapred-site.xml)
io.sort.mb: 150
Hi Tim,
jps should show it running if it's hung.
In that case, see if you can get a stack trace using jstack pid.
Paste that trace and we may be able to figure out where it's hung up.
-Todd
On Thu, Oct 15, 2009 at 12:12 AM, tim robertson
timrobertson...@gmail.com wrote:
Thanks for the info
page:
http://getsatisfaction.com/cloudera/products/cloudera_cloudera_s_distribution_for_hadoop
(we don't want to confuse people on this list if anything is specific to our
distro)
-Todd
On Wed, 14 Oct 2009 10:02:22 -0700, Todd Lipcon t...@cloudera.com wrote:
Hi Bryn,
Just to let you know
Hi Bryn,
Just to let you know, we've queued the patch Hairong mentioned for the
next update to our distribution, due out around the end of this month.
Thanks!
-Todd
On Wed, Oct 14, 2009 at 9:15 AM, Bryn Divey b...@bengueladev.com wrote:
Hi all,
I'm getting the following on initializing my
with extended ASCII delimiters. Following your answer,
however, I will try to use low-value ASCII, like 9 or 11, unless someone
has
a better suggestion.
Thank you,
Mark
On Fri, Oct 9, 2009 at 6:49 PM, Todd Lipcon t...@cloudera.com wrote:
Hi Mark,
If you're using TextOutputFormat, it assumes
As this topic comes up reasonably often on the list, I thought others might
be interested in this:
http://arstechnica.com/business/news/2009/10/dram-study-turns-assumptions-about-errors-upside-down.ars?utm_source=rssutm_medium=rssutm_campaign=rss
Basically, the takeaway is that RAM errors are
Hi Mark,
If you're using TextOutputFormat, it assumes you're dealing in UTF8. Decimal
254 wouldn't be valid as a standalone character in UTF8 encoding.
If you're dealing with binary (ie non-textual) data, you shouldn't use
TextOutputFormat.
-Todd
On Fri, Oct 9, 2009 at 3:09 PM, Mark Kerzner
Hi James,
This doesn't quite answer your original question, but if you want to help
track down these kinds of bugs, you should grab a stack trace next time this
happens.
You can do this either using jstack from the command line, by visiting
/stacks on the HTTP interface, or by sending the process
On Thu, Oct 1, 2009 at 10:53 AM, Stas Oskin stas.os...@gmail.com wrote:
Hi
I'm looking into Name Node high availability, but so far found only an
approach using DRBD.
I tried to make it work using Xen over DRBD, but it didn't quite work - in
fact I received a very valuable experience of
to me at this point. Any help deciphering it would be greatly
appreciated.
I have also now disabled the IB interface on my 2 test systems,
unfortunately that had no impact.
-Nick
Todd Lipcon wrote:
Hi Nick,
Figure out the pid of the DataNode process using either jps or
straight
ps auxw
Hi Kevin,
Less than $1k/box is unrealistic and won't be your best price/performance.
Most people building new clusters at this point seem to be leaning towards
dual quad core Nehalem with 4x1TB 7200RPM SATA and at least 8G RAM.
You're better off starting with a small cluster of these nicer
Hi Mayuran,
Yes, you need to run a secondary namenode.
The secondary namenode is *not* a backup mechanism. It is an important part
of the HDFS metadata system, and is responsible for periodically
checkpointing the filesystem namespace into a single file.
Without the secondary namenode running,
On Sun, Sep 27, 2009 at 7:39 PM, Starry SHI starr...@gmail.com wrote:
Hi Dave.
Thank you for your reply!
I have checked {dfs.data.dir}/tmp, the tmp files are there while the job is
running. However, it seems that the tmp files on each node are the same.
That is to say, the whole HDFS is
On Mon, Sep 28, 2009 at 10:44 AM, Mayuran Yogarajah
mayuran.yogara...@casalemedia.com wrote:
Hey Todd,
Note that you do not need to run the 2NN on a separate machine *if* you
have
enough RAM for two entire copies of your filesystem namespace. For small
clusters you should be fine to run
Hi Huang,
Boris's answer should work fine. If it would be useful for you to have a
single command line tool to start up a pseudo-distributed cluster for
testing, please comment on this JIRA:
http://issues.apache.org/jira/browse/MAPREDUCE-987
-Todd
On Fri, Sep 25, 2009 at 10:19 AM, Boris
Hi Marc,
Exit status 1 usually means some kind of controlled exit by the mapreduce
child task. Things like JVM crashes usually are indicated by other exit
codes (134 seems to be the code most commonly reported).
If you look at the stderr and stdout from your task (in the userlogs/
directory on
18:16:43,092 INFO org.apache.hadoop.mapred.TaskTracker:
addFreeSlot : current free slots : 2
2009-09-24 18:17:07,057 INFO org.apache.hadoop.mapred.TaskTracker: Received
'KillJobAction' for job: job_200909221656_0006
-Original Message-
From: Todd Lipcon [mailto:t...@cloudera.com]
Sent
On Mon, Sep 21, 2009 at 2:57 AM, Steve Loughran ste...@apache.org wrote:
Jeff Zhang wrote:
My cluster has running for several months.
Nice.
Is this a bug of hadoop? I think hadoop is supposed to run for long time.
I'm doing work in HDFS-326 on making it easier to start/stop the
On Mon, Sep 21, 2009 at 7:50 AM, Edward Capriolo edlinuxg...@gmail.comwrote:
Storing the only copy of the NN data into NFS would make the NFS server an
SPOF, and you still need to solve the problems of
@Steve correct. It is hair splitting but Stas asked if there was an
approach that did
Hi Gregory,
This is way outside the design parameters of HDFS. It may work, but you are
very likely to run into issues, and I don't think anyone would recommend
this as a solution. More reasonable would be a HDFS cluster spanning two
datacenters within the same metro area (1-2ms latency), but
Hi Stephen,
There was an email thread about two weeks ago announcing the release
candidate and subsequent release.
Regarding Cloudera's version, the most recent one available for download is
based on 0.20.0, but we've included many of the bug fixes that are in
0.20.1. We're currently working on
Hi Stas,
Look at the LocalDirAllocator class. It's not really meant for public
consumption, so its API might break in future releases, but that's the class
that the hadoop daemons use for round robin behavior.
-Todd
On Tue, Sep 15, 2009 at 4:51 AM, Stas Oskin stas.os...@gmail.com wrote:
Hi.
On Tue, Sep 15, 2009 at 9:20 AM, CubicDesign cubicdes...@gmail.com wrote:
We're currently working on updating our patch series to include the
entirety of 0.20.1 and should have that out next week.
Hi.
Do you think that 0.20.1 will be the next stable version (like 0.18.3?)
We will
are
specified with full name.
Do you have any idea about why this happens?
Again, thanks for your help
2009/9/14 Todd Lipcon t...@cloudera.com
That's not an error - that just means that the daemon thread is waiting
for
a connection (IO event)
The logs in $HADOOP_HOME/log/ are entirely empty
Hi,
The EC2 scripts will boot Cloudera's distribution for Hadoop. Currently they
boot our distribution of 0.18.3, but 0.20 support should be ready pretty
soon now. Here's a list of what patches are in our newest 0.18.3
distribution:
101 - 200 of 230 matches
Mail list logo