Description of problem:

+ The amount of metrics information (mainly "log_collected_bytes_total") published by the collector pods on the nodes gradually increases.

+ scrape_duration_seconds for collector fluentd goes up to 10 sec triggers FluentdNodeDown alert.

Version-Release number of selected component (if applicable):

+ Cluster logging 5.5.16, 5.7, 5.8

How reproducible:

+ At both customers and Red Hat test bench

Steps to Reproduce:

<will share it in the comment below>
====

KCS: https://access.redhat.com/solutions/7050165