Hi,
I think I know what's going on here. It has to do with how many spills the map task performs. You are emitting the numbers in order, so if there is only one spill, they stay in order. For larger number of records, the map task will create more than one spill, which must be merged. During the merge, the original order is not preserved. If you want the original order to be preserved, you must set io.sort.mb and/or io.sort.record.percent such that the map task requires only a single spill. Cheers, Sven From: Utkarsh Gupta [mailto:utkarsh_gu...@infosys.com] Sent: 18 January 2013 18:25 To: mapreduce-user@hadoop.apache.org Subject: RE: Limitation of key-value pairs for a particular key. You are right Actually we were expecting the values to be sorted. We tried to reproduce the problem by this simple code private final IntWritable one=new IntWritable(1); private Text word=new Text(); @Override public void map(LongWritable key,Text value, Context context) throws IOException, InterruptedException { int N=30000; for(int i=0;i<N;i++) { word.set(i+""); System.out.println(i); context.write(one,word); } } For smaller N numbers were in order but for N 3000000 order was not maintained From: Harsh J [mailto:ha...@cloudera.com] Sent: Thursday, January 17, 2013 1:57 AM To: mapreduce-user Subject: RE: Limitation of key-value pairs for a particular key. We don't sort values (only keys) nor apply any manual limits in MR. Can your post a reproduceable test case to support your suspicion? On Jan 16, 2013 4:34 PM, "Utkarsh Gupta" <utkarsh_gu...@infosys.com <mailto:utkarsh_gu...@infosys.com> > wrote: Hi, Thanks for the response. There was some issues with my code. I have checked that in detail. All the values of map are present in reducer but not in sorted order. This case happens if the number of values are too large for a key. Thanks Utkarsh From: Vinod Kumar Vavilapalli [mailto:vino...@hortonworks.com <mailto:vino...@hortonworks.com> ] Sent: Thursday, January 10, 2013 11:00 PM To: mapreduce-user@hadoop.apache.org <mailto:mapreduce-user@hadoop.apache.org> Subject: Re: Limitation of key-value pairs for a particular key. There isn't any limit like that. Can you reproduce this consistently? If so, please file a ticket. It will definitely help if you can provide a test case which can reproduce this issue. Thanks, +Vinod On Thu, Jan 10, 2013 at 12:41 AM, Utkarsh Gupta <utkarsh_gu...@infosys.com <mailto:utkarsh_gu...@infosys.com> > wrote: Hi, I am using Apache Hadoop 1.0.4 on a 10 node cluster of commodity machines with Ubuntu 12.04 Server edition. I am having a issue with my map reduce code. While debugging I found that the reducer can take 262145 values for a particular key. If more values are there, they seem to be corrupted. I checked the values while emitting from map and again checked in reducer. I am wondering is there any such kind of limitation in the Hadoop or is it a configuration problem. Thanks and Regards Utkarsh Gupta **************** CAUTION - Disclaimer ***************** This e-mail contains PRIVILEGED AND CONFIDENTIAL INFORMATION intended solely for the use of the addressee(s). If you are not the intended recipient, please notify the sender by e-mail and delete the original message. Further, you are not to copy, disclose, or distribute this e-mail or its contents to any other person and any such actions are unlawful. This e-mail may contain viruses. Infosys has taken every reasonable precaution to minimize this risk, but is not liable for any damage you may sustain as a result of any virus in this e-mail. You should carry out your own virus checks before opening the e-mail or attachment. Infosys reserves the right to monitor and review the content of all messages sent to or from this e-mail address. Messages sent to or from this e-mail address may be stored on the Infosys e-mail system. ***INFOSYS******** End of Disclaimer ********INFOSYS*** -- +Vinod Hortonworks Inc. http://hortonworks.com/
