Bug Description
A big portion on-premise deployments like Openstack, Kubernetes and Ceph use certificates generated by Vault, and not by a publicly trusted CA
Now, there is no option in the charm to provide the SSL CA cert of an on premise Ceph S3 endpoint
Affecting most Kubeflow deployments on-premise
To Reproduce
Deploy Minio in gateway mode
Point to an on-prem S3 endpoint that requires CA certs
I tried using ssl-ca but I assume that one only works for the Minio when on server mode
minio:
charm: minio
channel: ckf-1.10/stable
revision: 459
base: ubuntu@20.04/stable
resources:
oci-image: 547
scale: 1
options:
access-key: include-file://../../secrets/kubeflow/s3-kubeflow__access
secret-key: include-file://../../secrets/kubeflow/s3-kubeflow__secret
console-port: 9001
gateway-storage-service: s3
mode: gateway
port: 9000
ssl-ca: "include-base64://../../secrets/k8s/vault-root-ca.crt" # CA of https://ceph-radosgw.maas
ssl-cert: ""
ssl-key: ""
storage-service-endpoint: https://ceph-radosgw.maas
constraints: arch=amd64
Environment
Kubeflow 1.10/stable
Minio ckf-1.10/stable - revision: 459
S3: Ceph radosgw (on premise)
Relevant Log Output
# kk logs mlflow-minio-0
Defaulted container "minio" out of: minio, juju-pod-init (init)
2025-06-06T20:31:24.410Z [pebble] Started daemon.
2025-06-06T20:31:24.415Z [pebble] POST /v1/services 3.088395ms 202
2025-06-06T20:31:24.419Z [pebble] Service "minio" starting: docker-entrypoint.sh minio [ gateway s3 https://ceph-radosgw.maas --console-address :9001 ]
2025-06-06T20:31:24.697Z [pebble] Service "minio" stopped unexpectedly with code 1
2025-06-06T20:31:24.697Z [pebble] Service "minio" on-failure action is "restart", waiting ~500ms before restart (backoff 1)
2025-06-06T20:31:24.699Z [pebble] Change 1 task (Start service "minio") failed: service start attempt: exited quickly with code 1, will restart
2025-06-06T20:31:24.703Z [pebble] GET /v1/changes/1/wait 286.729015ms 200
2025-06-06T20:31:24.703Z [pebble] Started default services with change 1.
2025-06-06T20:31:25.233Z [pebble] Service "minio" starting: docker-entrypoint.sh
[..]
2025-06-06T21:29:25.587Z [pebble] Service "minio" stopped unexpectedly with code 1
2025-06-06T21:29:25.587Z [pebble] Service "minio" on-failure action is "restart", waiting ~30s before restart (backoff 115)
2025-06-06T21:29:56.241Z [pebble] Service "minio" starting: docker-entrypoint.sh minio [ gateway s3 https://ceph-radosgw.maas --console-address :9001 ]
2025-06-06T21:29:56.557Z [pebble] Service "minio" stopped unexpectedly with code 1
2025-06-06T21:29:56.557Z [pebble] Service "minio" on-failure action is "restart", waiting ~30s before restart (backoff 116)
Additional Context
No response
Bug Description
A big portion on-premise deployments like Openstack, Kubernetes and Ceph use certificates generated by Vault, and not by a publicly trusted CA
Now, there is no option in the charm to provide the SSL CA cert of an on premise Ceph S3 endpoint
Affecting most Kubeflow deployments on-premise
To Reproduce
Deploy Minio in gateway mode
Point to an on-prem S3 endpoint that requires CA certs
I tried using ssl-ca but I assume that one only works for the Minio when on server mode
Environment
Kubeflow 1.10/stable
Minio ckf-1.10/stable - revision: 459
S3: Ceph radosgw (on premise)
Relevant Log Output
Additional Context
No response