Tweaking pipeline performance

[Ossenfeld]

Hello,

I'm trying to get the hang off tweaking logstashs performance. I searched the net but unfortunately, I can't seem to find a good guide on how to tweak and actually measure the performance of a pipeline.
I can't use the pipeline management stuff inside Kibana which is why I'm doing with a curl pipelines command.

Currently, my problem is the following:

I got a dedicated redis and a dedicated logstash cluster. The logstash cluster pulls the data from the redis cluster, does it's thing and then writes to a dedicated elasticsearch cluster.
Even if the redis cluster has many keys "building up" or waiting to get pulled, the CPU utilization of the logstash cluster is only at ~ 50%.
One thing I found is by diving the queue_push_duration_in_millis through duration_in_millis, I should be able to see if Logstash is waiting for workers. I sometimes get a value of 4,X which would mean logstash is waiting a lot of its time, correct?

I'm unsure how I could tweak the settings and would really love a useful guide which I could use.

Thanks in advance
Ossenfeld

[Ossenfeld]

In on of my pipelines, I get the following results:

"distributor_lan_filebeat" : {
  "events" : {
    "duration_in_millis" : 174876784,
    "in" : 5250849,
    "queue_push_duration_in_millis" : 831997131,
    "filtered" : 5226753,
    "out" : 5226753

These seem very bad, but where would I start tweaking?