AWS EMR [System Error] Fail to delete the temp folder

Question

In AWS EMR, I encountered the following error message when running a pyspark job, which ran successfully on my local machine.

> [System Error] Fail to delete the temp folder

Is there a way to troubleshoot this? Is this a permissions issue with the temp folder access by EMR, accessible across all jobs?

Accepted Answer

Hello,

Yes, it looks like either the permission issue or the tmp files might be in use. 
1. Please check if you try opening the pyspark shell as hadoop user in EMR or use sudo pyspark in hadoop user
2. Try checking the `spark-shell` working without issue instead of pyspark shell. 
3. Include `spark.local.dir` to different local directory in primary node to see if this fix the issue.
4. Restart the spark service(`sudo systemctl restart spark-history-server.service `
5. Set the log level to debug `rootLogger.level = debug` in log4j file `/etc/spark/conf/log4j2.properties` and retry the pyspark shell. This might give more insight to understand the issue.

AWS EMR [System Error] Fail to delete the temp folder

Contenuto pertinente