Same for us fluentd was very resource hungry, heka brought down utilization significantly. Regarding the buffering we just added kafka in between for graylog/ES

Same for us fluentd was very resource hungry, heka brought down utilization significantly. Regarding the buffering we just added kafka in between for graylog/ES scaling/maintenance.