maobaolong commented on HDFS-13270:

[~ywskycn] Thank you for the quick reply. I have a couple of reason for this 
- As far as i know, the audit log in the namenode cannot feel the client but 
the router, because every RPC to the NamenodeRpcServer miss the client info. 
- When the router busy or in the unhealthy state, we want to know who(which 
client) attack me. This topic the from the different view, in the namenode 
audit view, we see who(which router) broke the cluster and the namenode state, 
in the router audit view, we see who(which client) do something bad and we can 
know whether the router is in a health state.

I think the auditlogger can be put into RouterRpcServer. For the namenode, the 
auditlog is in NameNodeRpc and FsNamesystem, so we cannot reuse the 

There are also two solution:
- Make the DefaultAuditLogger more abstract and common.
- Copy the logic of DefaultAuditLogger to a new class, maybe RouterAuditLogger, 
and port for router use.

> RBF: Router audit logger
> ------------------------
>                 Key: HDFS-13270
>                 URL: https://issues.apache.org/jira/browse/HDFS-13270
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: hdfs
>    Affects Versions: 3.2.0
>            Reporter: maobaolong
>            Priority: Major
> We can use router auditlogger to log the client info and cmd, because the 
> FSNamesystem#Auditlogger's log think the client are all from router.

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org

Reply via email to