Expand event ordering

Backport of #11715 to 7.x
This commit is contained in:
Karen Metts 2020-04-14 15:01:16 -04:00
parent 9597ab1eea
commit 335032dc3f

View file

@ -1,9 +1,9 @@
[[processing]]
=== Processing Details
Understanding how {ls} works and how components interrelate can help you make better
decisions when you are setting up or adjusting your {ls} environment. This
section is designed to elevate concepts to assist with that level of
Understanding how {ls} works and how components interrelate can help you make
better decisions when you are setting up or adjusting your {ls} environment.
This section is designed to elevate concepts to assist with that level of
knowledge.
NOTE: This is a new section. We're still working on it.
@ -12,6 +12,21 @@ NOTE: This is a new section. We're still working on it.
[[event-ordering]]
==== Event ordering
By design and by default, {ls} does not guarantee event order. Reordering can
occur in two places:
* Events within a batch can be reordered during filter processing.
* In-flight batches can be reordered when one or more batches process faster than
others.
When maintaining event order is important, use a single worker.
This approach ensures that batches are computed one-after-the-other, and
that events maintain their order within the batch.
[float]
[[order-setting]]
===== 'pipeline.ordered' setting
The `pipeline.ordered` setting in <<logstash-settings-file,logstash.yml>>
gives you more control over event ordering for single worker pipelines.
@ -28,8 +43,3 @@ processing cost required to preserve order.
The Java pipeline initialization time appears in the startup logs at INFO level.
Initialization time is the time it takes to compile the pipeline config and
instantiate the compiled execution for all workers.
If you notice a slowdown, ...