Description of problem:

When clusterlogging are deployed using deprecated format(spec.collection.type is not defined). the clusterlogging instance cat not be upgraded from 5.6 to 5.7.0

Steps to Reproduce:

1. Deploy clusterlogging without spec.collection.type.

apiVersion: logging.openshift.io/v1
kind: ClusterLogging
metadata:
  name: instance
  namespace: openshift-logging
spec:
  collection:
    logs:
      fluentd: {}
      type: fluentd
  managementState: Managed
  logStore:
    ....

2. Upgrade clusterlogging to 5.7.0

Actual results:

Clusterlogging can not be upgraded to 5.7.0

$ oc get csv

NAME                            DISPLAY                            VERSION   REPLACES                        PHASE

cluster-logging.v5.6.4          Red Hat OpenShift Logging          5.6.4     cluster-logging.v5.6.3          Replacing

cluster-logging.v5.7.0          Red Hat OpenShift Logging          5.7.0     cluster-logging.v5.6.4          Pending

$oc get ip install-c2clt -o json |jq '.status.conditions'

[
  
{     "lastTransitionTime": "2023-03-17T01:55:04Z",     "lastUpdateTime": "2023-03-17T01:55:04Z",     "message": "error validating existing CRs against new CRD's schema for \"clusterloggings.logging.openshift.io\": error validating custom resource against new schema for ClusterLogging openshift-logging/instance: [].spec.collection.type: Required value",     "reason": "InstallComponentFailed",     "status": "False",     "type": "Installed"   }
]

 

Expected results:

Add this issue as knonw issue. Or allow clusterlogging to be upgraded when  spec.collection.type is not defined.

Additional info:

Workaround

1) delete sub/cluster-logging, csv/clusterlogging.

2) upgrade clusterlogging/instance to use  spec.collection.type 

3) subscribe to clusterlogging 5.7.0 directly