Cluster-logging pods with tolerations getting frequently evicted in OCP 4.x

Issue

• Elasticsearch pods restarting in almost every 2 hours, even if the health of ES cluster is green and the cluster-logging components are working as expected.
• Events observed in openshift-logging namespace where taint manager is evicting the elasticsearch pod.

9m13s      Normal   TaintManagerEviction    pod/elasticsearch-cdm-xxx   Marking for deletion Pod openshift-logging/elasticsearch-cdm-xxx

• Some pods inside openshift-logging namespace are getting recreated after certain amount of time, and having zero restartCount.

$ oc get pods -n openshift-logging -owide
NAME                              READY   STATUS   RESTARTS   AGE     IP              NODE                      
cluster-logging-operator-xxx       1/1       Running    0    10d     10.1.1.1       worker-1    
elasticsearch-cdm-xxx             2/2      Running      0     52m   10.1.1.2      logging-1   
elasticsearch-cdm-xxx             2/2      Running      0      52m   10.1.1.3      logging-2  
elasticsearch-cdm-xxx            2/2      Running      0      52m   10.1.1.4      logging-3   
fluentd-xxx                        1/1       Running     0      52m   10.1.1.5       master-0        
fluentd-xxx                       1/1       Running     0      52m   10.1.1.5      master-1        
kibana-xxx                        2/2    Running       0       52m    10.1.1.2      logging-1