You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "jeanlyn (JIRA)" <ji...@apache.org> on 2015/05/27 04:58:17 UTC
[jira] [Created] (SPARK-7885) add config to control map aggregation
in spark sql
jeanlyn created SPARK-7885:
------------------------------
Summary: add config to control map aggregation in spark sql
Key: SPARK-7885
URL: https://issues.apache.org/jira/browse/SPARK-7885
Project: Spark
Issue Type: Improvement
Affects Versions: 1.3.1, 1.2.2, 1.2.0
Reporter: jeanlyn
For now, *execution.HashAggregation* add the map aggregation in oder to decrease the shuffle data.However,we found gc problem when we use this optimization and finally the executor crash.For example,
{noformat}
select sale_ord_id as order_id,
coalesce(sum(sku_offer_amount),0.0) as sku_offer_amount,
coalesce(sum(suit_offer_amount),0.0) as suit_offer_amount,
coalesce(sum(flash_gp_offer_amount),0.0) + coalesce(sum(gp_offer_amount),0.0) as gp_offer_amount,
coalesce(sum(flash_gp_offer_amount),0.0) as flash_gp_offer_amount,
coalesce(sum(full_minus_offer_amount),0.0) as full_rebate_offer_amount,
0.0 as telecom_point_offer_amount,
coalesce(sum(coupon_pay_amount),0.0) as dq_and_jq_pay_amount,
coalesce(sum(jq_pay_amount),0.0) + coalesce(sum(pop_shop_jq_pay_amount),0.0) + coalesce(sum(lim_cate_jq_pay_amount),0.0) as jq_pay_amount,
coalesce(sum(dq_pay_amount),0.0) + coalesce(sum(pop_shop_dq_pay_amount),0.0) + coalesce(sum(lim_cate_dq_pay_amount),0.0) as dq_pay_amount,
coalesce(sum(gift_cps_pay_amount),0.0) as gift_cps_pay_amount ,
coalesce(sum(mobile_red_packet_pay_amount),0.0) as mobile_red_packet_pay_amount,
coalesce(sum(acct_bal_pay_amount),0.0) as acct_bal_pay_amount,
coalesce(sum(jbean_pay_amount),0.0) as jbean_pay_amount,
coalesce(sum(sku_rebate_amount),0.0) as sku_rebate_amount,
coalesce(sum(yixun_point_pay_amount),0.0) as yixun_point_pay_amount,
coalesce(sum(sku_freight_coupon_amount),0.0) as freight_coupon_amount
from ord_at_det_di
where ds = '2015-05-20'
group by sale_ord_id
{noformat}
the sql scan two text files and each file is 360MB,we use 6 executor, each executor has 8GB memory and 2 cpu.
We can add a config control map aggregation to avoid it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org