Copy Local file using a Shared Cluster

[Kaizen]

Hi, 

I am saving some files locally on my cluster and moving them after my job. These are log files of my process so I cant directly reference a DBFS location. 

However the dbutils.fs.cp command does not work on the shared cluster. This does however work on a individual cluster. I believe this is related to how the clusters are split amongst users.

File location: "/home/spark-daed4064-233f-446c-b9f2-5b/log.txt''

Copy command: 

import os 

#path gets set to /home/spark-daed4064-233f-446c-b9f2-5b/ 
path = os.getcwd() 

new_path = f"{path}/logs.txt" 

# output printed out -> /home/spark-4c17311c-654a-4c71-b551-2e/logs.txt 
print(new_path) 

dbutils.fs.cp(new_path, "dbfs:/databricks/scripts/logs.txt")

[Kaizen]

For reference when doing this on a single user (personal) cluster - the file is store in:

/databricks/driver/logs.txt

Which has no issue accessing and copying to dbfs after using the dbutil commands

[Kaniz]

Hi @Kaizen , When using shared clusters, it's important to note that there are specific configurations that may affect the availability of dbutils.fs commands. If you come across the error message "INSUFFICIENT_PERMISSIONS," it usually means that your user account does not have the appropriate permissions to access the DBFS root location. It's best to consult your system administrator or cluster owner to ensure that your user account has the necessary permissions to access DBFS resources.

[Kaizen]

Hi @Kaniz - thanks for mentioning that. The issue is accessing the local file on the cluster not the dbfs location.

But it is still like you said a cluster config issue:
org.apache.spark.api.python.PythonSecurityException: Path 'file:/home/spark-c989284b-a795-4ca0-858e-84/logs.txt' uses an untrusted filesystem 'com.databricks.backend.daemon.driver.WorkspaceLocalFileSystem'