Re: Spark Push-Based Shuffle causing multiple stage failures

2022-05-28 Thread Ye Zhou
Hi, Han. The configuration for External Shuffle Service(ESS) in YARN has to be configured in yarn-site.xml for NodeManagers, as it is an auxiliary service in NodeManager. We will try to improve the documentation for enabling push based shuffle. Thanks for the feedback. For the straggler issue, is

Re: Spark Push-Based Shuffle causing multiple stage failures

2022-05-25 Thread Han Altae-Tran
Hi Ye, This is super super helpful! It wasn't obvious to me from the documentation that this property needed to be set in the yarn-site.xml file, as all other configurations in the main spark configuration page are set through spark conf. It was particularly confusing because this property, like m

Re: Spark Push-Based Shuffle causing multiple stage failures

2022-05-24 Thread Ye Zhou
Hi, Han. Thanks for trying out the push based shuffle. Please make sure you configure both the Spark client side configuration and server side configurations. The client side configuration looks good, and from the error message, looks like you are missing the server side configurations. Please refe

Re: Spark Push-Based Shuffle causing multiple stage failures

2022-05-24 Thread Mridul Muralidharan
+CC zhouye...@gmail.com On Mon, May 23, 2022 at 7:11 AM Han Altae-Tran wrote: > Hi, > > First of all, I am very thankful for all of the amazing work that goes > into this project! It has opened up so many doors for me! I am a long > time Spark user, and was very excited to start working with th

Spark Push-Based Shuffle causing multiple stage failures

2022-05-23 Thread Han Altae-Tran
Hi, First of all, I am very thankful for all of the amazing work that goes into this project! It has opened up so many doors for me! I am a long time Spark user, and was very excited to start working with the push-based shuffle service for an academic paper we are working on, but I encountered som