Skip to content

Commit 17beb0d

Browse files
committed
Improved 'executor_threads' description with integrations from #409
1 parent d7dc1f1 commit 17beb0d

File tree

1 file changed

+11
-1
lines changed

1 file changed

+11
-1
lines changed

docs/index.asciidoc

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -231,7 +231,17 @@ Refer to <<plugins-{type}s-{plugin}-ecs_metadata,ECS mapping>> for detailed info
231231
* Value type is <<number,number>>
232232
* Default value is 1 executor thread per CPU core
233233

234-
The number of parallel executor threads to be used to process incoming beats requests.
234+
The number of threads to be used to process incoming beats requests.
235+
By default Beats input will create a number of threads equals to 2*CPU cores to handle incoming connections,
236+
reading from the established sockets and execute most of the tasks related to network connection managements,
237+
except the parsing of Lumberjack protocol that's offloaded to a dedicated thread pool.
238+
239+
Generally you don't need to touch this setting.
240+
In case you are sending very large events and observing "OutOfDirectMemory" exceptions,
241+
you may want to reduce this number to half or 1/4 of the CPU cores.
242+
This will reduce the number of threads decompressing batches of data into direct memory.
243+
However, this will only be a mitigating tweak, as the proper solution may require resizing your Logstash deployment,
244+
either by increasing number of Logstash nodes or increasing the JVM's Direct Memory.
235245

236246
[id="plugins-{type}s-{plugin}-host"]
237247
===== `host`

0 commit comments

Comments
 (0)