The underlying Apache Spark code uses Apache Hadoop code to lookup the current
user name, which fails. It fails because containers in OpenShift run as an
ephemeral uid for security purposes, and that uid does not show up in
/etc/passwd. To address this, we have a custom entrypoint for
openshift-spark
containers that uses nss_wrapper to add
the ephemeral uid to /etc/passwd before running any code. However, we cannot
provide coverage for all the ways that code can be run. Specifically,
oc rsh or oc exec will avoid our setup code. If you are using either of
those, please use /entrypoint <your command>.