CODEWITHSUNDEEP
hivecountapache-spark-sqlimpala
jintocvg
jintocvg

Reputation: 158

Impala not able to read full table

I refreshed a table in hive using spark-sql (v 1.6.0), after which impala is not able to read complete content of the table.

select * from tablename limit 10; is giving some results where as select count(*) from tablename is giving the following error

Error seeking to 1073561824 in file: hdfs://server/products/path/db/tablename/part-00017 Error(255): Unknown error 255

Same error occurs if I execute query with where clause as well, and all these queries are working fine when accessed through Hive.

Upvotes: 1

Views: 3174

Answers (2)

Man Nguyen Dinh
Man Nguyen Dinh

Reputation: 71

Be sure you don't read and write the file at the same time Ex: Read parquet file f1, join with some others and then write it into f1 -> failed

Upvotes: 0

hlagos
hlagos

Reputation: 7947

try running

invalidate metadata yourtable.

maybe this can clarify the difference when use refresh or invalidate metadata

Upvotes: 0

Related Questions