Just curious if you've tried using Hive's explain method to see what IT thinks of your query.
On Fri, Aug 24, 2012 at 9:36 AM, Himanish Kushary <himan...@gmail.com>wrote: > Hi, > > We have a complex query that involves several left outer joins resulting > in 8 M/R jobs in Hive.During execution of one of the stages ( after three > M/R has run) the M/R job fails due to few Reduce tasks failing due to > inactivity. > > Most of the reduce tasks go through fine ( within 3 mins) but the last one > gets stuck for a long time (> 1 hour) and finally after several attempts > gets killed due to "failed to report status for 600 seconds. Killing!" > > What may be causing this issue ? Would hive.script.auto.progress help in > this case ? As we are not able to get much information from the log files > how may we approach resolving this ? Will tweaking of any specific M/R > parameters help ? > > The task attempt log shows several lines like this before exiting : > > 2012-08-23 19:17:23,848 INFO ExecReducer: ExecReducer: processing 219000000 > rows: used memory = 408582240 > 2012-08-23 19:17:30,189 INFO ExecReducer: ExecReducer: processing 220000000 > rows: used memory = 346110400 > 2012-08-23 19:17:37,510 INFO ExecReducer: ExecReducer: processing 221000000 > rows: used memory = 583913576 > 2012-08-23 19:17:44,829 INFO ExecReducer: ExecReducer: processing 222000000 > rows: used memory = 513071504 > 2012-08-23 19:17:47,923 INFO org.apache.hadoop.mapred.FileInputFormat: Total > input paths to process : 1 > > Here are the reduce task counters: > > *Map-Reduce Framework* Combine input records0 Combine output records0Reduce > input groups > 222,480,335 Reduce shuffle bytes7,726,141,897 Reduce input records > 222,480,335 Reduce output records0 Spilled Records355,827,191 CPU time > spent (ms)2,152,160 Physical memory (bytes) snapshot1,182,490,624 Virtual > memory (bytes) snapshot1,694,531,584 Total committed heap usage (bytes) > 990,052,352 > > The tasktracker log gives a thread dump at that time but no exception. > > *2012-08-23 20:05:49,319 INFO org.apache.hadoop.mapred.TaskTracker: > Process Thread Dump: lost task* > *69 active threads* > > --------------------------- > Thanks & Regards > Himanish > -- "The whole world is you. Yet you keep thinking there is something else." - Xuefeng Yicun 822-902 A.D. Tim R. Havens Google Phone: 573.454.1232 ICQ: 495992798 ICBM: 37°51'34.79"N 90°35'24.35"W ham radio callsign: NW0W
