KERBEROS_REALM is unset when creating a shell on HDFS pods

Affected Stackable version

25.11

Affected Apache HDFS version

3.4.2

Current and expected behavior

When I shell into a pod to run an HDFS admin command and run hdfs haadmin -getAllServiceState to see the namenode HA status I receive an error message:

stackable@simple-hdfs-namenode-default-0 /stackable/hadoop-3.4.2-stackable25.11.0 $ hdfs haadmin -getAllServiceState
2026-03-10 14:28:00,336 WARN  ipc.Client (Client.java:run(749)) - Exception encountered while connecting to the server simple-hdfs-namenode-default-0.simple-hdfs-namenode-default.default.svc.cluster.local/10.1.17.125:8020
javax.security.sasl.SaslException: Bad Kerberos server principal configuration [Caused by java.lang.IllegalArgumentException: Server has invalid Kerberos principal: nn/simple-hdfs.default.svc.cluster.local@KNAB.COM, expecting: nn/simple-hdfs.default.svc.cluster.local@${env.KERBEROS_REALM}]

core-site.xml contains references to the environment variable KERBEROS_REALM in place of the actual Kerberos realm name and this variable is not set in the shell.

Possible solution

Set the KERBEROS_REALM environment variable so that command line tools use the correct realm name.

Additional context

No response

Environment

No response

Would you like to work on fixing this bug?

None

Inspect how the operator generates core-site.xml and the environment for HDFS pods, starting with the KERBEROS_REALM reference and the shell context for hdfs haadmin. Done when running hdfs haadmin -getAllServiceState from a pod uses the actual Kerberos realm and no longer reports an invalid principal.