Filebeat is able to automatically install the required ingest pipelines when it has Elasticsearch as output.
Since filebeat 6.3.0 filebeat setup is able to install pipelines using the --pipelines flag, this allows to install the pipelines by an operator when filebeats don't have Elasticsearch as output (in architectures with logstash, kafka...).
There is documentation in the guide about installing index templates and kibana dashboards, but not about pipelines, and it can be difficult to know when it is needed to do it manually and how to do it.
Note that specifying --modules modulename is required (along with --pipelines) to load the modules. I'll clarify this in the docs because it's not obvious, especially since we recommend that users use the modules.d directory to enable/disable modules. They will expect the setup command for pipelines to work similarly to the setup command for dashboards.
Related to https://github.com/elastic/beats/issues/9268
I've added some docs for LS that contain more info about loading and using ingest pipelines with Logstash. I'll add a link to the Filebeat doc after the LS PR is merged: https://github.com/elastic/logstash/pull/10438
I've also created a PR to add some links in the Beats docs that point the new content in the Logstash docs: https://github.com/elastic/beats/pull/10859.
Closing this issue because I think it's covered now.
Most helpful comment
I've added some docs for LS that contain more info about loading and using ingest pipelines with Logstash. I'll add a link to the Filebeat doc after the LS PR is merged: https://github.com/elastic/logstash/pull/10438