[ https://issues.apache.org/jira/browse/HADOOP-15336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16418201#comment-16418201 ]
Íñigo Goiri commented on HADOOP-15336: -------------------------------------- Thanks [~shahrs87] for your comments. The reason for the Router to decode the request is because it needs to know the path the request is going to. For example, if you ask the router to create a file in /user/user1, the Router needs to get the request, check the mount table to get the destination subcluster (e.g., /user is in /data in subcluster0), and then send a new request with the right file destination to the Namenode (e.g., /data/user1 in subcluster0). Regarding the backwards compatibility issue, I don't think this is introducing one but fixing one. In any case, this is a general issue and you can repro by having a client and a server (no need for a Router). Note that the reverse convert does actually check fs.hasKeyProviderUri(). Actually I would switch the null check by hasKeyProviderUri(). > RBF: NPE for FsServerDefaults.getKeyProviderUri() for clientProtocol > communication between 2.7 and 3.2 > ------------------------------------------------------------------------------------------------------ > > Key: HADOOP-15336 > URL: https://issues.apache.org/jira/browse/HADOOP-15336 > Project: Hadoop Common > Issue Type: Bug > Affects Versions: 3.1.0, 3.2.0 > Reporter: Sherwood Zheng > Assignee: Sherwood Zheng > Priority: Major > Labels: backward-incompatible, common > Attachments: HADOOP-15336.000.patch, HADOOP-15336.001.patch > > > KeyProviderUri is not available in 2.7 so when 2.7 clients contact with 3.2 > services, it cannot find the key provider URI and triggers a > NullPointerException. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org