Page 1 of 1

Agg,1: Hash table has grown to 16384 entries.

Posted: Mon Aug 07, 2006 11:17 am
by aceflyer
Hi
I am using a Aggregator stage in a PX job , Doing hash partitioning on it and sorting the data as well
I am getting these warnings

Agg,1: Hash table has grown to 16384 entries.
Agg,0: Hash table has grown to 16384 entries.
.....

Can anyone please suggest how to get rid of these warnings

Thanks

Posted: Mon Aug 07, 2006 12:12 pm
by kris007
In the options property of the aggregator stage select Method=Sort not Hash. This should take care of the warning.

Posted: Mon Aug 07, 2006 12:37 pm
by aceflyer
Hi
I was getting this warning when i did that

Agg: When checking operator: User inserted sort "Agg.ToAgg_Sort" does not fulfill the sort requirements of the downstream operator "APT_SortedGroup2Operator in Agg"

But when i unchecked the Perform Sort in the Aggregator it worked

Thanks

Posted: Mon Aug 07, 2006 2:39 pm
by kris007
Make sure Data is sorted and key partitioned before passing it on to aggregator stage to get the accurate results.

Posted: Mon Aug 07, 2006 4:36 pm
by clickart
You can use a SORT stage before the AGGR stage to sort the data and the partitioning must be on the same fields as the grouping keys in the AGGR stage.

Posted: Mon Aug 07, 2006 10:57 pm
by ray.wurlod
Hash table aggregation method is recommended for 1000 or fewer distinct grouping values per MB. Looks like you've exceeded this limit.

Posted: Tue Aug 08, 2006 2:48 am
by kumar_s
Do you initially have any partiton and sort on the key on which the aggregation is done?

aggregator

Posted: Mon Jan 14, 2008 8:33 pm
by just4u_sharath
ray.wurlod wrote:Hash table aggregation method is recommended for 1000 or fewer distinct grouping values per MB. Looks like you've exceeded this limit.
How can we know how much MB our aggregator stage is using. If we know that we can count of record length and chosse between hash or sort.
please reply

aggregator

Posted: Mon Jan 14, 2008 8:35 pm
by just4u_sharath
ray.wurlod wrote:Hash table aggregation method is recommended for 1000 or fewer distinct grouping values per MB. Looks like you've exceeded this limit.
How can we know how much MB our aggregator stage is using. If we know that we can count of record length and chosse between hash or sort.
please reply

Posted: Mon Jan 14, 2008 9:12 pm
by ray.wurlod
APT_PM_PLAYER_MEMORY