spark sql启动优化

./spark-sql --conf spark.driver.maxResultSize=8g --driver-memory 20g --conf  spark.kryoserializer.buffer.max=1G  --conf spark.kryoserializer.buer=64m

测试语句:

select h02.pvcode, h02.empi, h02.stay, h02.create_time,diag.name_diag,h02.remark chief, h05.remark present, h03.remark vital,lis.name_lis,pacs.name_pacs, treat.name_treat,
h06.remark past, h04.remark person, h07.remark family, h01.remark allergic
from emrdata_h02_parquet h02
left join emrdata_h05_parquet h05 on h02.pvcode = h05.pvcode
left join emrdata_h01_parquet h01 on h02.pvcode = h01.pvcode
left join emrdata_h03_parquet h03 on h02.pvcode = h03.pvcode
left join emrdata_h04_parquet h04 on h02.pvcode = h04.pvcode
left join emrdata_h06_parquet h06 on h02.pvcode = h06.pvcode
left join emrdata_h07_parquet h07 on h02.pvcode = h07.pvcode
left join (select pk_dcpv, concat_ws(',',collect_set(name_diag)) name_diag from diag_final group by pk_dcpv) diag on diag.pk_dcpv = h02.pvcode
left join (select pk_dcpv, concat_ws(',',collect_set(name_orditem)) name_lis from ord_lis group by pk_dcpv) lis on lis.pk_dcpv = h02.pvcode
left join (select pk_dcpv, concat_ws(',',collect_set(name_orditem)) name_pacs from ord_pacs group by pk_dcpv) pacs on pacs.pk_dcpv = h02.pvcode
left join (select pk_dcpv, concat_ws(',',collect_set(name_orditem)) name_treat from ord_treat group by pk_dcpv) treat on treat.pk_dcpv = h02.pvcode limit 5;

测试通过!!!!!