Many fluentd pods in 'MatchNodeSelector' status during deployment

Issue

• After deploying openshift logging with the following configuration:

openshift_logging_es_pvc_dynamic=true
openshift_logging_es_pvc_size=100G
openshift_logging_es_cluster_size=3
openshift_logging_es_memory_limit=4G
openshift_logging_es_number_of_replicas=1

• Expected to see a fluentd instance on each node, however we see the following:

NAME                          READY     STATUS              RESTARTS   AGE                    
logging-curator-1-1mz2x       1/1       Running             4          23m     
logging-es-20sd50o9-1-h94dm   1/1       Running             0          23m     
logging-es-602kzxam-1-0zb9t   1/1       Running             0          23m     
logging-es-b7i434tm-1-jld4n   1/1       Running             0          23m     
logging-fluentd-0wxbw         0/1       MatchNodeSelector   0          23m     
logging-fluentd-2jpz5         0/1       MatchNodeSelector   0          23m     
logging-fluentd-3b9vx         0/1       MatchNodeSelector   0          23m     
logging-fluentd-3jsm5         0/1       MatchNodeSelector   0          23m     
logging-fluentd-4cmgl         0/1       MatchNodeSelector   0          23m     
logging-fluentd-b8xlj         1/1       Running             0          23m     
logging-fluentd-jf0r1         1/1       Running             0          23m     
logging-fluentd-jv950         1/1       Running             0          23m     
logging-fluentd-lmjrn         0/1       MatchNodeSelector   0          23m     
logging-fluentd-qq8hs         0/1       MatchNodeSelector   0          23m     
logging-fluentd-x1s19         1/1       Running             0          23m     
logging-kibana-1-qbn7q        2/2       Running             0          23m     

• Fluentd only started on our application nodes.
• Other fluentd on infra and master node are in MatchNodeSelector