You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-user@hadoop.apache.org by Gang Luo <lg...@yahoo.com.cn> on 2010/01/22 18:03:50 UTC
fine granularity operation on HDFS
Hi all,
now that I can get the splits of a file in hadoop, is it possible to name some splits (not all) as the input to mapper? Or can I manually read some of these splits (not the whole file) using HDFS api?
-Gang
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
Re: map side only behavior
Posted by Gang Luo <lg...@yahoo.com.cn>.
Thanks Jeff.
But what if the mapper's output size surpass the capacity of the output buffer? There will be multiple spills for this mapper in this case. If there is no merging, how to ensure there is *only one* output file?
-Gang
----- 原始邮件 ----
发件人: Jeff Zhang <zj...@gmail.com>
收件人: common-user@hadoop.apache.org
发送日期: 2010/1/29 (周五) 11:05:33 上午
主 题: Re: map side only behavior
No, the merge and sort will not happen in mapper task. And each mapper task
will generate one output file.
2010/1/29 Gang Luo <lg...@yahoo.com.cn>
> Hi all,
> If I only use map side to process my data (set # of reducers to 0 ), what
> is the behavior of hadoop? Will it merge and sort each of the spills
> generated by one mapper?
>
> -Gang
>
>
> ----- 原始邮件 ----
> 发件人: Gang Luo <lg...@yahoo.com.cn>
> 收件人: common-user@hadoop.apache.org
> 发送日期: 2010/1/29 (周五) 8:54:33 上午
> 主 题: Re: fine granularity operation on HDFS
>
> Yeah, I see how it works. Thanks Amogh.
>
>
> -Gang
>
>
>
> ----- 原始邮件 ----
> 发件人: Amogh Vasekar <am...@yahoo-inc.com>
> 收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
> 发送日期: 2010/1/28 (周四) 10:00:22 上午
> 主 题: Re: fine granularity operation on HDFS
>
> Hi Gang,
> Yes PathFilters work only on file paths. I meant you can include such type
> of logic at split level.
> The input format's getSplits() method is responsible for computing and
> adding splits to a list container, for which JT initializes mapper tasks.
> You can override the getSplits() method to add only a few , say, based on
> the location or offset, to the list. Here's the reference :
> while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
> int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining);
> splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
> blkLocations[blkIndex].getHosts()));
> bytesRemaining -= splitSize;
> }
>
> if (bytesRemaining != 0) {
> splits.add(new FileSplit(path, length-bytesRemaining,
> bytesRemaining,
> blkLocations[blkLocations.length-1].getHosts()));
>
> Before splits.add you can use your logic for discarding. However, you need
> to ensure your record reader takes care of incomplete records at boundaries.
>
> To get the block locations to load separately, the FileSystem class APIs
> expose few methods like getBlockLocations etc ..
> Hope this helps.
>
> Amogh
>
> On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
>
> Thanks Amogh.
>
> For the second part of my question, I actually mean loading block
> separately from HDFS. I don't know whether it is realistic. Anyway, for my
> goal is to process different division of a file separately, to do that at
> split level is OK. But even I can get the splits from inputformat, how to
> "add only a few splits you need to mapper and discard the others"?
> (pathfilters only works for file, but not block, I think).
>
> Thanks.
> -Gang
>
>
>
> ___________________________________________________________
> 好玩贺卡等你发,邮箱贺卡全新上线!
> http://card.mail.cn.yahoo.com/
>
>
>
> ___________________________________________________________
> 好玩贺卡等你发,邮箱贺卡全新上线!
> http://card.mail.cn.yahoo.com/
>
--
Best Regards
Jeff Zhang
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
task-specific parameter
Posted by Gang Luo <lg...@yahoo.com.cn>.
Hi all,
Usually we set the parameter before the job starts. I am wondering whether we can do that after the job starts, e.g. in a configure() method of a map task. Or can we even do that in map function to dynamically tune the parameters? How about reduce side?
Besides, If I can do that in a specific task, are these modifications only visible to that task, or do they have a global impact? (currently, I am focusing on io.sort.record.percent).
Thanks.
-Gang
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
Re: map side only behavior
Posted by Aaron Kimball <aa...@cloudera.com>.
In a map-only job, map tasks will be connected directly to the OutputFormat.
So calling output.collect() / context.write() in the mapper will emit data
straight to files in HDFS without sorting the data. There is no sort buffer
involved. If you want exactly one output file, follow Nick's advice.
- Aaron
On Fri, Jan 29, 2010 at 8:32 AM, Jones, Nick <ni...@amd.com> wrote:
> A single unity reducer should enforce a merge and sort to generate one
> file.
>
> Nick Jones
>
> -----Original Message-----
> From: Jeff Zhang [mailto:zjffdu@gmail.com]
> Sent: Friday, January 29, 2010 10:06 AM
> To: common-user@hadoop.apache.org
> Subject: Re: map side only behavior
>
> No, the merge and sort will not happen in mapper task. And each mapper task
> will generate one output file.
>
>
>
> 2010/1/29 Gang Luo <lg...@yahoo.com.cn>
>
> > Hi all,
> > If I only use map side to process my data (set # of reducers to 0 ), what
> > is the behavior of hadoop? Will it merge and sort each of the spills
> > generated by one mapper?
> >
> > -Gang
> >
> >
> > ----- 原始邮件 ----
> > 发件人: Gang Luo <lg...@yahoo.com.cn>
> > 收件人: common-user@hadoop.apache.org
> > 发送日期: 2010/1/29 (周五) 8:54:33 上午
> > 主 题: Re: fine granularity operation on HDFS
> >
> > Yeah, I see how it works. Thanks Amogh.
> >
> >
> > -Gang
> >
> >
> >
> > ----- 原始邮件 ----
> > 发件人: Amogh Vasekar <am...@yahoo-inc.com>
> > 收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
> > 发送日期: 2010/1/28 (周四) 10:00:22 上午
> > 主 题: Re: fine granularity operation on HDFS
> >
> > Hi Gang,
> > Yes PathFilters work only on file paths. I meant you can include such
> type
> > of logic at split level.
> > The input format's getSplits() method is responsible for computing and
> > adding splits to a list container, for which JT initializes mapper tasks.
> > You can override the getSplits() method to add only a few , say, based on
> > the location or offset, to the list. Here's the reference :
> > while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
> > int blkIndex = getBlockIndex(blkLocations,
> length-bytesRemaining);
> > splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
> > blkLocations[blkIndex].getHosts()));
> > bytesRemaining -= splitSize;
> > }
> >
> > if (bytesRemaining != 0) {
> > splits.add(new FileSplit(path, length-bytesRemaining,
> > bytesRemaining,
> > blkLocations[blkLocations.length-1].getHosts()));
> >
> > Before splits.add you can use your logic for discarding. However, you
> need
> > to ensure your record reader takes care of incomplete records at
> boundaries.
> >
> > To get the block locations to load separately, the FileSystem class APIs
> > expose few methods like getBlockLocations etc ..
> > Hope this helps.
> >
> > Amogh
> >
> > On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
> >
> > Thanks Amogh.
> >
> > For the second part of my question, I actually mean loading block
> > separately from HDFS. I don't know whether it is realistic. Anyway, for
> my
> > goal is to process different division of a file separately, to do that at
> > split level is OK. But even I can get the splits from inputformat, how to
> > "add only a few splits you need to mapper and discard the others"?
> > (pathfilters only works for file, but not block, I think).
> >
> > Thanks.
> > -Gang
> >
> >
> >
> > ___________________________________________________________
> > 好玩贺卡等你发,邮箱贺卡全新上线!
> > http://card.mail.cn.yahoo.com/
> >
> >
> >
> > ___________________________________________________________
> > 好玩贺卡等你发,邮箱贺卡全新上线!
> > http://card.mail.cn.yahoo.com/
> >
>
>
>
> --
> Best Regards
>
> Jeff Zhang
>
RE: map side only behavior
Posted by "Jones, Nick" <ni...@amd.com>.
A single unity reducer should enforce a merge and sort to generate one file.
Nick Jones
-----Original Message-----
From: Jeff Zhang [mailto:zjffdu@gmail.com]
Sent: Friday, January 29, 2010 10:06 AM
To: common-user@hadoop.apache.org
Subject: Re: map side only behavior
No, the merge and sort will not happen in mapper task. And each mapper task
will generate one output file.
2010/1/29 Gang Luo <lg...@yahoo.com.cn>
> Hi all,
> If I only use map side to process my data (set # of reducers to 0 ), what
> is the behavior of hadoop? Will it merge and sort each of the spills
> generated by one mapper?
>
> -Gang
>
>
> ----- 原始邮件 ----
> 发件人: Gang Luo <lg...@yahoo.com.cn>
> 收件人: common-user@hadoop.apache.org
> 发送日期: 2010/1/29 (周五) 8:54:33 上午
> 主 题: Re: fine granularity operation on HDFS
>
> Yeah, I see how it works. Thanks Amogh.
>
>
> -Gang
>
>
>
> ----- 原始邮件 ----
> 发件人: Amogh Vasekar <am...@yahoo-inc.com>
> 收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
> 发送日期: 2010/1/28 (周四) 10:00:22 上午
> 主 题: Re: fine granularity operation on HDFS
>
> Hi Gang,
> Yes PathFilters work only on file paths. I meant you can include such type
> of logic at split level.
> The input format's getSplits() method is responsible for computing and
> adding splits to a list container, for which JT initializes mapper tasks.
> You can override the getSplits() method to add only a few , say, based on
> the location or offset, to the list. Here's the reference :
> while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
> int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining);
> splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
> blkLocations[blkIndex].getHosts()));
> bytesRemaining -= splitSize;
> }
>
> if (bytesRemaining != 0) {
> splits.add(new FileSplit(path, length-bytesRemaining,
> bytesRemaining,
> blkLocations[blkLocations.length-1].getHosts()));
>
> Before splits.add you can use your logic for discarding. However, you need
> to ensure your record reader takes care of incomplete records at boundaries.
>
> To get the block locations to load separately, the FileSystem class APIs
> expose few methods like getBlockLocations etc ..
> Hope this helps.
>
> Amogh
>
> On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
>
> Thanks Amogh.
>
> For the second part of my question, I actually mean loading block
> separately from HDFS. I don't know whether it is realistic. Anyway, for my
> goal is to process different division of a file separately, to do that at
> split level is OK. But even I can get the splits from inputformat, how to
> "add only a few splits you need to mapper and discard the others"?
> (pathfilters only works for file, but not block, I think).
>
> Thanks.
> -Gang
>
>
>
> ___________________________________________________________
> 好玩贺卡等你发,邮箱贺卡全新上线!
> http://card.mail.cn.yahoo.com/
>
>
>
> ___________________________________________________________
> 好玩贺卡等你发,邮箱贺卡全新上线!
> http://card.mail.cn.yahoo.com/
>
--
Best Regards
Jeff Zhang
Re: map side only behavior
Posted by Jeff Zhang <zj...@gmail.com>.
No, the merge and sort will not happen in mapper task. And each mapper task
will generate one output file.
2010/1/29 Gang Luo <lg...@yahoo.com.cn>
> Hi all,
> If I only use map side to process my data (set # of reducers to 0 ), what
> is the behavior of hadoop? Will it merge and sort each of the spills
> generated by one mapper?
>
> -Gang
>
>
> ----- 原始邮件 ----
> 发件人: Gang Luo <lg...@yahoo.com.cn>
> 收件人: common-user@hadoop.apache.org
> 发送日期: 2010/1/29 (周五) 8:54:33 上午
> 主 题: Re: fine granularity operation on HDFS
>
> Yeah, I see how it works. Thanks Amogh.
>
>
> -Gang
>
>
>
> ----- 原始邮件 ----
> 发件人: Amogh Vasekar <am...@yahoo-inc.com>
> 收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
> 发送日期: 2010/1/28 (周四) 10:00:22 上午
> 主 题: Re: fine granularity operation on HDFS
>
> Hi Gang,
> Yes PathFilters work only on file paths. I meant you can include such type
> of logic at split level.
> The input format's getSplits() method is responsible for computing and
> adding splits to a list container, for which JT initializes mapper tasks.
> You can override the getSplits() method to add only a few , say, based on
> the location or offset, to the list. Here's the reference :
> while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
> int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining);
> splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
> blkLocations[blkIndex].getHosts()));
> bytesRemaining -= splitSize;
> }
>
> if (bytesRemaining != 0) {
> splits.add(new FileSplit(path, length-bytesRemaining,
> bytesRemaining,
> blkLocations[blkLocations.length-1].getHosts()));
>
> Before splits.add you can use your logic for discarding. However, you need
> to ensure your record reader takes care of incomplete records at boundaries.
>
> To get the block locations to load separately, the FileSystem class APIs
> expose few methods like getBlockLocations etc ..
> Hope this helps.
>
> Amogh
>
> On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
>
> Thanks Amogh.
>
> For the second part of my question, I actually mean loading block
> separately from HDFS. I don't know whether it is realistic. Anyway, for my
> goal is to process different division of a file separately, to do that at
> split level is OK. But even I can get the splits from inputformat, how to
> "add only a few splits you need to mapper and discard the others"?
> (pathfilters only works for file, but not block, I think).
>
> Thanks.
> -Gang
>
>
>
> ___________________________________________________________
> 好玩贺卡等你发,邮箱贺卡全新上线!
> http://card.mail.cn.yahoo.com/
>
>
>
> ___________________________________________________________
> 好玩贺卡等你发,邮箱贺卡全新上线!
> http://card.mail.cn.yahoo.com/
>
--
Best Regards
Jeff Zhang
map side only behavior
Posted by Gang Luo <lg...@yahoo.com.cn>.
Hi all,
If I only use map side to process my data (set # of reducers to 0 ), what is the behavior of hadoop? Will it merge and sort each of the spills generated by one mapper?
-Gang
----- 原始邮件 ----
发件人: Gang Luo <lg...@yahoo.com.cn>
收件人: common-user@hadoop.apache.org
发送日期: 2010/1/29 (周五) 8:54:33 上午
主 题: Re: fine granularity operation on HDFS
Yeah, I see how it works. Thanks Amogh.
-Gang
----- 原始邮件 ----
发件人: Amogh Vasekar <am...@yahoo-inc.com>
收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
发送日期: 2010/1/28 (周四) 10:00:22 上午
主 题: Re: fine granularity operation on HDFS
Hi Gang,
Yes PathFilters work only on file paths. I meant you can include such type of logic at split level.
The input format's getSplits() method is responsible for computing and adding splits to a list container, for which JT initializes mapper tasks. You can override the getSplits() method to add only a few , say, based on the location or offset, to the list. Here's the reference :
while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining);
splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
blkLocations[blkIndex].getHosts()));
bytesRemaining -= splitSize;
}
if (bytesRemaining != 0) {
splits.add(new FileSplit(path, length-bytesRemaining, bytesRemaining,
blkLocations[blkLocations.length-1].getHosts()));
Before splits.add you can use your logic for discarding. However, you need to ensure your record reader takes care of incomplete records at boundaries.
To get the block locations to load separately, the FileSystem class APIs expose few methods like getBlockLocations etc ..
Hope this helps.
Amogh
On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
Thanks Amogh.
For the second part of my question, I actually mean loading block separately from HDFS. I don't know whether it is realistic. Anyway, for my goal is to process different division of a file separately, to do that at split level is OK. But even I can get the splits from inputformat, how to "add only a few splits you need to mapper and discard the others"? (pathfilters only works for file, but not block, I think).
Thanks.
-Gang
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
Re: fine granularity operation on HDFS
Posted by Gang Luo <lg...@yahoo.com.cn>.
Yeah, I see how it works. Thanks Amogh.
-Gang
----- 原始邮件 ----
发件人: Amogh Vasekar <am...@yahoo-inc.com>
收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
发送日期: 2010/1/28 (周四) 10:00:22 上午
主 题: Re: fine granularity operation on HDFS
Hi Gang,
Yes PathFilters work only on file paths. I meant you can include such type of logic at split level.
The input format's getSplits() method is responsible for computing and adding splits to a list container, for which JT initializes mapper tasks. You can override the getSplits() method to add only a few , say, based on the location or offset, to the list. Here's the reference :
while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining);
splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
blkLocations[blkIndex].getHosts()));
bytesRemaining -= splitSize;
}
if (bytesRemaining != 0) {
splits.add(new FileSplit(path, length-bytesRemaining, bytesRemaining,
blkLocations[blkLocations.length-1].getHosts()));
Before splits.add you can use your logic for discarding. However, you need to ensure your record reader takes care of incomplete records at boundaries.
To get the block locations to load separately, the FileSystem class APIs expose few methods like getBlockLocations etc ..
Hope this helps.
Amogh
On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
Thanks Amogh.
For the second part of my question, I actually mean loading block separately from HDFS. I don't know whether it is realistic. Anyway, for my goal is to process different division of a file separately, to do that at split level is OK. But even I can get the splits from inputformat, how to "add only a few splits you need to mapper and discard the others"? (pathfilters only works for file, but not block, I think).
Thanks.
-Gang
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
Re: fine granularity operation on HDFS
Posted by Amogh Vasekar <am...@yahoo-inc.com>.
Hi Gang,
Yes PathFilters work only on file paths. I meant you can include such type of logic at split level.
The input format's getSplits() method is responsible for computing and adding splits to a list container, for which JT initializes mapper tasks. You can override the getSplits() method to add only a few , say, based on the location or offset, to the list. Here's the reference :
while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) {
int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining);
splits.add(new FileSplit(path, length-bytesRemaining, splitSize,
blkLocations[blkIndex].getHosts()));
bytesRemaining -= splitSize;
}
if (bytesRemaining != 0) {
splits.add(new FileSplit(path, length-bytesRemaining, bytesRemaining,
blkLocations[blkLocations.length-1].getHosts()));
Before splits.add you can use your logic for discarding. However, you need to ensure your record reader takes care of incomplete records at boundaries.
To get the block locations to load separately, the FileSystem class APIs expose few methods like getBlockLocations etc ..
Hope this helps.
Amogh
On 1/28/10 7:26 PM, "Gang Luo" <lg...@yahoo.com.cn> wrote:
Thanks Amogh.
For the second part of my question, I actually mean loading block separately from HDFS. I don't know whether it is realistic. Anyway, for my goal is to process different division of a file separately, to do that at split level is OK. But even I can get the splits from inputformat, how to "add only a few splits you need to mapper and discard the others"? (pathfilters only works for file, but not block, I think).
Thanks.
-Gang
Re: fine granularity operation on HDFS
Posted by Gang Luo <lg...@yahoo.com.cn>.
Thanks Amogh.
For the second part of my question, I actually mean loading block separately from HDFS. I don't know whether it is realistic. Anyway, for my goal is to process different division of a file separately, to do that at split level is OK. But even I can get the splits from inputformat, how to "add only a few splits you need to mapper and discard the others"? (pathfilters only works for file, but not block, I think).
Thanks.
-Gang
----- 原始邮件 ----
发件人: Amogh Vasekar <am...@yahoo-inc.com>
收件人: "common-user@hadoop.apache.org" <co...@hadoop.apache.org>
发送日期: 2010/1/27 (周三) 1:40:26 下午
主 题: Re: fine granularity operation on HDFS
Hi,
>>now that I can get the splits of a file in hadoop, is it possible to name some splits (not all) as the input to mapper?
I'm assuming when you say "splits of a file in hadoop" you mean splits generated from the inputformat and not the blocks stored in HDFS.
The [File]InputFormat you use gives you access to splits, locations etc. You can use this to add only a few splits you need to mapper and discard the others ( something you can do on files as a whole using PathFilters ).
>>Or can I manually read some of these splits (not the whole file) using HDFS api?
You mean you list these splits somewhere in a file beforehand so individual mappers can read one line (split) ?
Amogh
___________________________________________________________
好玩贺卡等你发,邮箱贺卡全新上线!
http://card.mail.cn.yahoo.com/
Re: fine granularity operation on HDFS
Posted by Amogh Vasekar <am...@yahoo-inc.com>.
Hi,
>>now that I can get the splits of a file in hadoop, is it possible to name some splits (not all) as the input to mapper?
I'm assuming when you say "splits of a file in hadoop" you mean splits generated from the inputformat and not the blocks stored in HDFS.
The [File]InputFormat you use gives you access to splits, locations etc. You can use this to add only a few splits you need to mapper and discard the others ( something you can do on files as a whole using PathFilters ).
>>Or can I manually read some of these splits (not the whole file) using HDFS api?
You mean you list these splits somewhere in a file beforehand so individual mappers can read one line (split) ?
Amogh