Dask - new cluster creation fails, HDFS files owed by "dask" user

Question

I have setup dask on my MapR cluster's edge node following the directions here: https://gateway.dask.org/install-hadoop.html

Per those directions, I'm testing the install by running the following in a JupyterHub spawned ipython notebook:

from dask_gateway import Gateway
gateway = Gateway("http://sa1x-hadoopedg-np1.hchc.local:9010")
cluster = gateway.new_cluster()

However, when it tries to start the new cluster via YARN, I get the following error in the YARN application's log:

Diagnostics: User a059571(user id 1425180742) does not have access to maprfs:///user/a059571/.skein/application_1605411890003_0222/809B8EAF0CC3524F90366F449C11C97E/tmpv8cbv2ag

Even though dask is supposed to be running as the requesting user (in this case a059571), it appears to be creating directories as the user running the dask-gateway-server (in this case the user mapr):

hdfs dfs -ls -d maprfs:///user/a059571/.skein/application_1605411890003_0222
drwx------   - mapr mapr  7 2021-01-19 17:37 maprfs:///user/a059571/.skein/application_1605411890003_0222

I feel like I'm missing something obvious.

Here are my configs, for full disclosure:

/etc/dask-gateway/dask_gateway_config.py

c.DaskGateway.backend_class = (
    "dask_gateway_server.backends.yarn.YarnBackend"
)
c.DaskGateway.address= '12.190.113.133:9010'
c.Proxy.address = '12.190.113.133:9011'
c.Proxy.tcp_address = '12.190.113.133:9012'
c.YarnClusterConfig.scheduler_cmd = "/opt/anaconda3/bin/dask-scheduler"
c.YarnClusterConfig.worker_cmd = "/opt/anaconda3/bin/dask-worker"
c.YarnClusterConfig.queue = 'root.default'
c.DaskGateway.log_level= 'DEBUG'

Snippet from inside my core_site.xml

  <property>
    <name>hadoop.proxyuser.mapr.hosts</name>
    <value>*</value>
  </property>
  <property>
    <name>hadoop.proxyuser.mapr.groups</name>
    <value>*</value>
  </property>

And, some interesting lines from the dask-gateway-server logs:

[DaskGateway] - HTTP routes listening at http://12.190.113.133:9011
[DaskGateway] - Scheduler routes listening at gateway://12.190.113.133:9012
[Proxy] Unexpected failure fetching routing table, retrying in 0.5s: Get http://12.190.113.133:9010/api/v1/routes: dial tcp 12.190.113.133:9010: connect: connection refused
[DaskGateway] Removed 0 expired clusters from the database
[Proxy] Unexpected failure fetching routing table, retrying in 1.0s: Get http://12.190.113.133:9010/api/v1/routes: dial tcp 12.190.113.133:9010: connect: connection refused
[Proxy] Unexpected failure fetching routing table, retrying in 2.0s: Get http://12.190.113.133:9010/api/v1/routes: dial tcp 12.190.113.133:9010: connect: connection refused
[Proxy] Unexpected failure fetching routing table, retrying in 4.0s: Get http://12.190.113.133:9010/api/v1/routes: dial tcp 12.190.113.133:9010: connect: connection refused
INFO skein.Driver: Driver started, listening on 44262
[DaskGateway] Backend started, clusters will contact api server at http://12.190.113.133:9011/api
[DaskGateway] Dask-Gateway server started
[DaskGateway] - Private API server listening at http://12.190.113.133:9010

Note: sa1x-hadoopedg-np1.hchc.local == 12.190.113.133, an RHEL 7.x server. MapR cluster is 6.x.