Details
Description
HADOOP-13597 migrated KMS from Tomcat to Jetty.
In Hadoop 2.x, the Accept Queue Size default value was set by environment variable KMS_ACCEPT_QUEUE with default value 500.
But after the migration, Hadoop 3.x now uses the default HttpServer2 configuration to set accept queue size (hadoop.http.socket.backlog.size), which is 128.
Should we restore the default accept queue size in kms-default.xml? Should we document it as a known issue? After all, Hadoop 2 --> 3 allows breaking changing.
Attachments
Attachments
Issue Links
- is broken by
-
HADOOP-13597 Switch KMS from Tomcat to Jetty
- Resolved