Best practice tuning "buffer"


#1

Hi,

i’m trying to tune my graylog configuration and i’m playing with processbuffer_processors / outputbuffer_processors / inputbuffer_processors.
But there is some rules regarding the number of vCPU ram etc…?

My setup is composed of one graylog (~1500msg/sec) and one cluster of 3 elsatic search .
ES vm is 4vCPU (not really loaded) 16GB with HEAP at 12GB
Graylog is 2vCPU (30/40% loaded) 8GB with HEAP at 4GB.

Thank for help.


(Jan Doberstein) #2

The rule of thumb would be that those three settings processbuffer_processors / outputbuffer_processors / inputbuffer_processors should be equal to the number of CPUs you have.

If you are unsure about, leave the defaults.


#3

Ok thank you for help, is it the sum of the three settings which should be equal to the number of CPU?


(Jan Doberstein) #4

yes - the sum of all.


(Mark M) #5

I have a VM running Elastic and GL with 2 CPU’s and 2 cores each. From what I can tell the “default” totals 12 processors. What is your recommendation? We only collect from about a dozen syslog senders.


(Jan Doberstein) #6

leave the defaults - until you have real issues or the need of changing.


(system) #7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.