Fluentd is in CrashLoopBackOff state after RHOCP upgrade

Issue

• Fluentd is stuck in CrashLoopBackOff after cluster upgrade.

$ oc get pods -l component=fluentd -n openshift-logging 
NAME                                            READY   STATUS             RESTARTS   AGE
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m
fluentd-xxxxx                                   0/1     CrashLoopBackOff   63         5h6m

• Below logs are observed in fluentd pods.

2021-01-19T06:11:46.340504814Z Setting each total_size_limit for 2 buffers to 10265531289 bytes
2021-01-19T06:11:46.357700887Z Setting queued_chunks_limit_size for each buffer to 1223
2021-01-19T06:11:46.357700887Z Setting chunk_limit_size for each buffer to 8388608
2021-01-19T06:11:48.965670188Z 2021-01-19 06:11:48 +0000 [error]: config error file="/etc/fluent/fluent.conf" error_class=Fluent::ConfigError error="Invalid Kubernetes API v1 endpoint https://kubernetes.default.svc: Unauthorized"