search

stackabletech / hdfs-operator

Apache Hadoop HDFS operator for Stackable
Other
40 stars 4 forks source link

HDFS env overrides not working correctly #497

Closed maltesander closed 7 months ago

maltesander commented 8 months ago

If using ENV overrides of variables set by the operator, the mechanism is not working correctly. When trying to apply something like:

nameNodes:
  config:
    listenerClass: cluster-internal
    logging:
      enableVectorAgent: False
  roleGroups:
    default:
      envOverrides:
        HDFS_NAMENODE_OPTS: "blabla"
      replicas: 2

The operator errors out because there are multiple entries of HDFS_NAMENODE_OPTS in the final EnvVar vector.

2024-03-12T16:44:52.520017Z ERROR hdfs_controller: stackable_operator::logging::controller: Failed to reconcile object controller.name="hdfsclusters.hdfs.stackable.tech" error=reconciler for object HdfsCluster.v1alpha1.hdfs.stackable.tech/hdfs.
kuttl-test-needed-kingfish failed error.sources=[cannot create role group stateful set "hdfs-namenode-default", Kubernetes reported error: ApiError: failed to create typed patch object (kuttl-test-needed-kingfish/hdfs-namenode-default; apps/v1,
 Kind=StatefulSet): .spec.template.spec.containers[name="namenode"].env: duplicate entries for key [name="HDFS_NAMENODE_OPTS"]:  (ErrorResponse { status: "Failure", message: "failed to create typed patch object (kuttl-test-needed-kingfish/hdfs-
namenode-default; apps/v1, Kind=StatefulSet): .spec.template.spec.containers[name=\"namenode\"].env: duplicate entries for key [name=\"HDFS_NAMENODE_OPTS\"]", reason: "", code: 500 }), ApiError: failed to create typed patch object (kuttl-test-n
eeded-kingfish/hdfs-namenode-default; apps/v1, Kind=StatefulSet): .spec.template.spec.containers[name="namenode"].env: duplicate entries for key [name="HDFS_NAMENODE_OPTS"]:  (ErrorResponse { status: "Failure", message: "failed to create typed 
patch object (kuttl-test-needed-kingfish/hdfs-namenode-default; apps/v1, Kind=StatefulSet): .spec.template.spec.containers[name=\"namenode\"].env: duplicate entries for key [name=\"HDFS_NAMENODE_OPTS\"]", reason: "", code: 500 }), failed to cre
ate typed patch object (kuttl-test-needed-kingfish/hdfs-namenode-default; apps/v1, Kind=StatefulSet): .spec.template.spec.containers[name="namenode"].env: duplicate entries for key [name="HDFS_NAMENODE_OPTS"]: ]

This is because the operator treats env vars internally mostly as a Vector, therefore not replacing but appending overrides. This is fine until env vars set by the operator are overriden.

This whole method should be reworked https://github.com/stackabletech/hdfs-operator/blob/c80034261ccce65eea2b5f38dee68349ef8e9964/rust/operator-binary/src/container.rs#L737 into using maps until the final override at the end.

lfrancke commented 7 months ago

Is this in 24.3?

adwk67 commented 7 months ago

Yes, it's included.