spark-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Amit Sharma <resolve...@gmail.com>
Subject Re: Memory Limits error
Date Fri, 16 Aug 2019 11:16:02 GMT
Increasing your driver memory as 12g.

On Thursday, August 15, 2019, Dennis Suhari <d.suhari@icloud.com.invalid>
wrote:

> Hi community,
>
> I am using Spark on Yarn. When submiting a job after a long time I get an
> error mesage and retry.
>
> It happens when I want to store the dataframe to a table.
>
> spark_df.write.option("path", "/nlb_datalake/golden_zone/
> webhose/sentiment").saveAsTable("news_summary_test", mode="overwrite")
>
> The error is (after long time):
>
>  Hive Session ID = be590d1b-ed5b-404b-bcb4-77cbb977a847 [Stage 2:> (0 +
> 16) / 16]19/08/15 15:42:08 WARN BlockManagerMasterEndpoint: No more
> replicas available for rdd_9_2 ! 19/08/15 15:42:08 WARN
> BlockManagerMasterEndpoint: No more replicas available for rdd_9_1 !
> 19/08/15 15:42:08 WARN BlockManagerMasterEndpoint: No more replicas
> available for rdd_9_4 ! 19/08/15 15:42:08 WARN BlockManagerMasterEndpoint:
> No more replicas available for rdd_9_6 ! 19/08/15 15:42:08 WARN
> BlockManagerMasterEndpoint: No more replicas available for rdd_9_7 !
> 19/08/15 15:42:08 WARN BlockManagerMasterEndpoint: No more replicas
> available for rdd_9_0 ! 19/08/15 15:42:08 WARN BlockManagerMasterEndpoint:
> No more replicas available for rdd_9_5 ! 19/08/15 15:42:08 WARN
> BlockManagerMasterEndpoint: No more replicas available for rdd_9_3 !
> 19/08/15 15:42:08 WARN YarnSchedulerBackend$YarnSchedulerEndpoint:
> Requesting driver to remove executor 2 for reason Container killed by YARN
> for exceeding memory limits. 9.1 GB of 9 GB physical memory used. Consider
> boosting spark.yarn.executor.memoryOverhead. 19/08/15 15:42:08 ERROR
> YarnScheduler: Lost executor 2 on nlb-srv-hd-08.i-lab.local: Container
> killed by YARN for exceeding memory limits. 9.1 GB of 9 GB physical memory
> used. Consider boosting spark.yarn.executor.memoryOverhead. 19/08/15
> 15:42:08 WARN TaskSetManager: Lost task 0.0 in stage 2.0 (TID 17,
> nlb-srv-hd-08.i-lab.local, executor 2): ExecutorLostFailure (executor 2
> exited caused by one of the running tasks) Reason: Container killed by YARN
> for exceeding memory limits. 9.1 GB of 9 GB physical memory used. Consider
> boosting spark.yarn.executor.memoryOverhead. 19/08/15 15:42:08 WARN
> TaskSetManager: Lost task 5.0 in stage 2.0 (TID 26,
> nlb-srv-hd-08.i-lab.local, executor 2): ExecutorLostFailure (executor 2
> exite
>
> Do you have a rough idea where to tweak ?
>
> Br,
>
> Dennis
>

Mime
View raw message