Mohammad Arshad created HDFS-17958:
--------------------------------------
Summary: Support NameNode-scoped configuration for
dfs.namenode.observer.enabled
Key: HDFS-17958
URL: https://issues.apache.org/jira/browse/HDFS-17958
Project: Hadoop HDFS
Issue Type: Bug
Components: namenode
Affects Versions: 3.4.1
Reporter: Mohammad Arshad
Assignee: Mohammad Arshad
*Background:*
I installed the HDFS service with HA enabled, using two NameNodes—one active
and one standby. All HDFS components currently share the same set of
configuration files.
I now want to add a third NameNode as an Observer NameNode. However, I've run
into a limitation:
* If I set dfs.namenode.observer.enabled=true to designate the third NameNode
as an observer, all the NameNodes start as observers, since they share the same
configuration.
* Alternatively, I can start the third NameNode as an observer by passing the
-observer flag at startup. But since I'm using a cluster installation tool,
this approach requires me to separately track which NameNode is the observer
and apply that information during startup—which adds unwanted complexity.
*Proposed Change:*
Similar to other existing NameNode-scoped configurations, I propose adding a
NameNode-scoped configuration:
dfs.namenode.observer.enabled.<nameserviceId>.<namenodeId>
This configuration would be set only in the configuration of the specific
NameNode we want to designate as the observer.
As a result, only that NameNode would start as an Observer NameNode, leaving
the other NameNodes unaffected.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
