You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hive.apache.org by Prabhu Joseph <pr...@gmail.com> on 2016/02/17 13:55:51 UTC
Alter Index on Column Query consumes huge memory
Hi All,
When rebuilding index on a table, where abc is an index on a column
which has 8.5 lakh count for a single column value, the mapreduce job ran
and the reducers timing out, when we monitored the reducers, it is taking
7.3GB RSS and 7.5GB Virt, the reducer memory configured is 7GB. The task is
doing lot of GC and CPU was 2000% and stop the world signal would have
caused the time out of tasks.
ALTER INDEX abc ON table1 REBUILD;
For 8.5 lakh of BigInt Columns ( 8.5 * 8 bytes = 6MB) , why it needs more
than 7GB. I suspect there is some memory issue with alter index query.
Count abc
*854959 110335*
395551 80558
340020 275122
208852 308319
192140 139578
179191 49070
165588 346486
133544 199021
130692 270755
103451 275368
Thanks,
Prabhu Joseph