When try to run simple hive query: select count(*) form table , its shows the Error: Invalid distance too far back.

[#1]

When try to run simple hive query in tez: select count(*) form table , its shows the Error: Invalid distance too far back..can anyone help me out of this error.. i have attached the error log...please help me..errorlog.png

[#2 · Accepted solution]

It seems your zip file present in table directory is corrupted. Try decompress the file directly with unzip utility(you may get the file name from the failed container logs).

[#3]

Thanks you so much@Ankit Singhal

But i have below concerns

#why should i decompress the file..?

# how can i get ..failed container logs because.. i could see 12 directory paths under "yarn.nodemanager.log-dirs" property and .i just confused where should i find the application logs ...please suggest me..

[#4]

@sankar rao,

bq. why should i decompress the file..?

In order to confirm that file selected is actually corrupted or not.

bq. how can i get ..failed container logs because.. i could see 12 directory paths under "yarn.nodemanager.log-dirs" property and .i just confused where should i find the application logs

Actually, I don't remember the actual keyword to search in the logs ,but you can check syslogs for container with id similar to _14435*_237788_1_01_000062_1 and look for line saying "processing file" or something similar.

[#5]

Thanks again for your time@Ankit Singhal

# I feel bad to say that i am unable to understand this reply"In order to confirm that file selected is actually corrupted or not."

You already said "It seems your zip file present in table directory is corrupted" so why should need this..and If corrupted is confirm..what is solution to execute this query ...

#my intense is to run query...how can i do it ..what are the steps should i take..

[#6]

@Ankit Singhal

Any updated answer to my query ?

[#7]

@sankar rao, Actually I don't have the yarn cluster ready to confirm you what log lines needs to be searched in the container logs for the file names. Probably , it will be better if you can raise a support case so that a dedicated team can look into the issue specifically. As it depends, How is the data loaded in the table/hdfs , how they are zipped, which input format you are using etc.?