1) How many mappers generated in your MR job? i have lots of input file ,each size is below 64m ,if i set the block size to 128m or 256m,it will help? my job total map tasks number is 2717
2) Are they all finished? (Check them in the jobtracker through web or command line) yes, all map tasks done 3) How many reducers in this job? all reducer in running state ,and the number of reducer is 15 4) Are reducers starting? What stage are they in? Copying/Sorting/Reducing? i do not know how to judge if the reducer is in copying / sorting / reducing ,the web ui not tell me,i use yarn framework ,and i read the syslog of the container which reducer running in ,as following output,so i judge the fetch thread get all output from map task,and the stuff from map task also sorted and merged, so what the reducer do now is reducing 2013-12-05 18:14:09,546 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.Fetcher: fetcher#3 about to shuffle output of map attempt_1386139114497_0034_m_009498_0 decomp: 2 len: 6 to MEMORY 2013-12-05 18:14:09,546 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.InMemoryMapOutput: Read 2 bytes from map-output for attempt_1386139114497_0034_m_009498_0 2013-12-05 18:14:09,546 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.MergeManagerImpl: closeInMemoryFile -> map-output of size: 2, inMemoryMapOutputs.size() -> 5641, commitMemory -> 656458998, usedMemory ->656459000 2013-12-05 18:14:09,547 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.Fetcher: fetcher#3 about to shuffle output of map attempt_1386139114497_0034_m_009499_0 decomp: 2 len: 6 to MEMORY 2013-12-05 18:14:09,547 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.InMemoryMapOutput: Read 2 bytes from map-output for attempt_1386139114497_0034_m_009499_0 2013-12-05 18:14:09,547 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.MergeManagerImpl: closeInMemoryFile -> map-output of size: 2, inMemoryMapOutputs.size() -> 5642, commitMemory -> 656459000, usedMemory ->656459002 2013-12-05 18:14:09,547 INFO [fetcher#3] org.apache.hadoop.mapreduce.task.reduce.ShuffleScheduler: CHBM224:8080 freed by fetcher#3 in 3s 2013-12-05 18:14:09,547 INFO [EventFetcher for fetching Map Completion Events] org.apache.hadoop.mapreduce.task.reduce.EventFetcher: EventFetcher is interrupted.. Returning 2013-12-05 18:14:09,555 INFO [main] org.apache.hadoop.mapreduce.task.reduce.MergeManagerImpl: finalMerge called with 5642 in-memory map-outputs and 6 on-disk map-outputs 2013-12-05 18:14:09,598 INFO [main] org.apache.hadoop.mapred.Merger: Merging 5642 sorted segments 2013-12-05 18:14:09,610 INFO [main] org.apache.hadoop.mapred.Merger: Down to the last merge-pass, with 5566 segments left of total size: 656260591 bytes 2013-12-05 18:14:13,614 INFO [main] org.apache.hadoop.mapreduce.task.reduce.MergeManagerImpl: Merged 5642 segments, 656459002 bytes to disk to satisfy reduce memory limit 2013-12-05 18:14:13,615 INFO [main] org.apache.hadoop.mapreduce.task.reduce.MergeManagerImpl: Merging 7 files, 24536123780 bytes from disk 2013-12-05 18:14:13,628 INFO [main] org.apache.hadoop.mapreduce.task.reduce.MergeManagerImpl: Merging 0 segments, 0 bytes from memory into reduce 2013-12-05 18:14:13,628 INFO [main] org.apache.hadoop.mapred.Merger: Merging 7 sorted segments 2013-12-05 18:14:13,883 INFO [main] org.apache.hadoop.mapred.Merger: Down to the last merge-pass, with 7 segments left of total size: 24536123501 bytes 2013-12-05 18:14:14,021 WARN [main] org.apache.hadoop.conf.Configuration: mapred.skip.on is deprecated. Instead, use mapreduce.job.skiprecords 2013-12-05 18:14:14,076 INFO [main] com.alibaba.dubbo.common.logger.LoggerFactory: using logger: com.alibaba.dubbo.common.logger.log4j.Log4jLoggerAdapter 5) If in the reducing stage, check the userlog of reducers. Is your code running now? which the userlog of reducers located? On Thu, Dec 12, 2013 at 9:58 AM, java8964 <java8...@hotmail.com> wrote: > Assume the block size is 128M, and your mapper each finishes within half > minute, then there is not too much logic in your mapper, as it can finish > processing 128M around 30 seconds. If your reducers cannot finish with 1 > week, then something is wrong. > > So you may need to find out following: > > 1) How many mappers generated in your MR job? > 2) Are they all finished? (Check them in the jobtracker through web or > command line) > 3) How many reducers in this job? > 4) Are reducers starting? What stage are they in? Copying/Sorting/Reducing? > 5) If in the reducing stage, check the userlog of reducers. Is your code > running now? > > All these information you can find out from the Job Tracker web UI. > > Yong > > ------------------------------ > Date: Thu, 12 Dec 2013 09:03:29 +0800 > > Subject: Re: issue about Shuffled Maps in MR job summary > From: justlo...@gmail.com > To: user@hadoop.apache.org > > hi, > suppose i have 5-worknode cluster,each worknode can allocate 40G mem > ,and i do not care map task,be cause the map task in my job finished within > half a minuter,as my observe the real slow task is reduce, i allocate 12G > to each reduce task,so each worknode can support 3 reduce parallel,and the > whole cluster can support 15 reducer,and i run the job with all 15 reducer, > and i do not know if i increase reducer number from 15 to 30 ,each reduce > allocate 6G MEM,that will speed the job or not ,the job run on my product > env, it run nearly 1 week,it still not finished > > On Wed, Dec 11, 2013 at 9:50 PM, java8964 <java8...@hotmail.com> wrote: > > The whole job complete time depends on a lot of factors. Are you sure > the reducers part is the bottleneck? > > Also, it also depends on how many Reducer input groups it has in your MR > job. If you only have 20 reducer groups, even you jump your reducer count > to 40, then the epoch of reducers part won't have too much change, as the > additional 20 reducer task won't get data to process. > > If you have a lot of reducer input groups, and your cluster does have > capacity at this time, and your also have a lot idle reducer slot, then > increase your reducer count should decrease your whole job complete time. > > Make sense? > > Yong > > ------------------------------ > Date: Wed, 11 Dec 2013 14:20:24 +0800 > Subject: Re: issue about Shuffled Maps in MR job summary > From: justlo...@gmail.com > To: user@hadoop.apache.org > > > i read the doc, and find if i have 8 reducer ,a map task will output 8 > partition ,each partition will be send to a different reducer, so if i > increase reduce number ,the partition number increase ,but the volume on > network traffic is same,why sometime ,increase reducer number will not > decrease job complete time ? > > On Wed, Dec 11, 2013 at 1:48 PM, Vinayakumar B > <vinayakuma...@huawei.com>wrote: > > It looks simple, J > > Shuffled Maps= Number of Map Tasks * Number of Reducers > > Thanks and Regards, > Vinayakumar B > > *From:* ch huang [mailto:justlo...@gmail.com] > *Sent:* 11 December 2013 10:56 > *To:* user@hadoop.apache.org > *Subject:* issue about Shuffled Maps in MR job summary > > hi,maillist: > i run terasort with 16 reducers and 8 reducers,when i double > reducer number, the Shuffled maps is also double ,my question is the job > only run 20 map tasks (total input file is 10,and each file is 100M,my > block size is 64M,so split is 20) why i need shuffle 160 maps in 8 reducers > run and 320 maps in 16 reducers run?how to caculate the shuffle maps number? > > 16 reducer summary output: > > > Shuffled Maps =320 > > 8 reducer summary output: > > Shuffled Maps =160 > > > >
