Cause: when the "cluster-monitoring-config" config map is invalid (e.g. it can't be decoded), the cluster monitoring operator uses the default configuration.
Consequence: any change made to customize the cluster monitoring configuration (such as PVC, node selectors, toleration) would be lost and the monitoring stack would be reconciled to its default state.
Fix: when the cluster monitoring operator can't decode "cluster-monitoring-config" config map, it doesn't try to reconcile the monitoring stack. An alert fires when this happens.
Result: the monitoring stack isn't modified when an invalid configuration is given.
Description of problem:
When the 'cluster-monitoring-config' config map is updated to something that can't be parsed by CMO, the operator will log an error but it will use the default configuration, potentially reverting previous customization.
Version-Release number of selected component (if applicable):
4.4 but the same is true for earlier versions
How reproducible:
Always
Steps to Reproduce:
1. Enable user workload monitoring as described in the 4.3 docs.
2. Edit the 'cluster-monitoring-config' config map to the following:
kind: ConfigMap
apiVersion: v1
metadata:
name: cluster-monitoring-config
data:
config.yaml: |
techPreviewUserWorkload:
enabled: invalid
Actual results:
No more pods running in the user-workload-monitoring namespace.
Expected results:
CMO shouldn't do anything.
Additional info:
Invalid CMO config should trigger an alert.
Test with 4.5.0-0.nightly-2020-04-02-195956, set invalid value in cluster-monitoring-config confimap, there is error in cluster-monitoring-operator pod and trigger ClusterMonitoringOperatorReconciliationErrors/ClusterOperatorDegraded/ClusterOperatorDown alerts, no resource is created under openshift-user-workload-monitoring
kind: ConfigMap
apiVersion: v1
metadata:
name: cluster-monitoring-config
namespace: openshift-monitoring
data:
config.yaml: |
techPreviewUserWorkload:
enabled: tuer
# oc -n openshift-monitoring logs cluster-monitoring-operator-876444cb8-vfzrc -c cluster-monitoring-operator| tail
E0403 08:31:39.424739 1 operator.go:273] sync "openshift-monitoring/cluster-monitoring-config" failed: the Cluster Monitoring ConfigMap could not be parsed: error unmarshaling JSON: while decoding JSON: json: cannot unmarshal string into Go struct field UserWorkloadConfig.techPreviewUserWorkload.enabled of type bool
I0403 08:31:39.424851 1 operator.go:298] Updating ClusterOperator status to failed. Err: the Cluster Monitoring ConfigMap could not be parsed: error unmarshaling JSON: while decoding JSON: json: cannot unmarshal string into Go struct field UserWorkloadConfig.techPreviewUserWorkload.enabled of type bool
# oc -n openshift-user-workload-monitoring get pod
No resources found in openshift-user-workload-monitoring namespace.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHBA-2020:2409
Description of problem: When the 'cluster-monitoring-config' config map is updated to something that can't be parsed by CMO, the operator will log an error but it will use the default configuration, potentially reverting previous customization. Version-Release number of selected component (if applicable): 4.4 but the same is true for earlier versions How reproducible: Always Steps to Reproduce: 1. Enable user workload monitoring as described in the 4.3 docs. 2. Edit the 'cluster-monitoring-config' config map to the following: kind: ConfigMap apiVersion: v1 metadata: name: cluster-monitoring-config data: config.yaml: | techPreviewUserWorkload: enabled: invalid Actual results: No more pods running in the user-workload-monitoring namespace. Expected results: CMO shouldn't do anything. Additional info: Invalid CMO config should trigger an alert.