Hi adnan I've met similar problem, the reducer output file length is zero and missing some bytes at the end of the output file. the cause is I use MultipleOutputs and forgot to close it at reducer cleanup method.
hope it helps On 26 May 2015 at 17:13, Adnan Karač <[email protected]> wrote: > Hi Brahma, > > Thanks for the quick response. I assumed that running file check without > *openforwrite* option would yield file in this block whether it was open > for write and not. However, I have just tried it as well, unfortunately no > success. > > Adnan > ᐧ > > On Tue, May 26, 2015 at 10:12 AM, Brahma Reddy Battula < > [email protected]> wrote: > >> >> Can you try like following..? >> >> * hdfs fsck -openforwrite -files -blocks -locations / | >> grep blk_1109280129_1099547327549* >> >> >> Thanks & Regards >> >> Brahma Reddy Battula >> >> >> ------------------------------ >> *From:* Adnan Karač [[email protected]] >> *Sent:* Tuesday, May 26, 2015 1:34 PM >> *To:* [email protected] >> *Subject:* Cannot obtain block length for LocatedBlock >> >> Hi all, >> >> I have an MR job running and exiting with following exception. >> >> java.io.IOException: Cannot obtain block length for LocatedBlock >> {BP-1632531813-172.19.67.67-1393407344218:blk_1109280129_1099547327549; >> getBlockSize()=139397; corrupt=false; offset=0; locs=[172.19.67.67:50010, >> 172.19.67.78:50010, 172.19.67.84:50010]} >> >> Now, the fun part is that i don't know which file is in question. In >> order to find this out, i did this: >> >> *hdfs fsck -files -blocks / | grep blk_1109280129_1099547327549* >> >> Interestingly enough, it came up with nothing. >> >> Did anyone experience anything similar? Or does anyone have a piece of >> advice on how to resolve this? >> >> Version of hadoop is 2.3.0 >> >> Thanks in advance! >> >> -- >> Adnan Karač >> ᐧ >> > > > > -- > Adnan Karač > -- All the best Liu Bo
