[ https://issues.apache.org/jira/browse/OOZIE-3542?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16936957#comment-16936957 ]
Zsombor Gegesy commented on OOZIE-3542: --------------------------------------- Yes, the problem was, that the there are multiple reflective calls to the remote HDFS service, where the same error could happen - and unfortunately only one call is handled previously. The new patch adds the same checks to all calls. > Handle better old Hdfs implementations in ECPolicyDisabler > ---------------------------------------------------------- > > Key: OOZIE-3542 > URL: https://issues.apache.org/jira/browse/OOZIE-3542 > Project: Oozie > Issue Type: Bug > Components: tools > Reporter: Zsombor Gegesy > Assignee: Zsombor Gegesy > Priority: Major > Fix For: 5.2.0 > > Attachments: OOZIE-3542-2.patch, OOZIE-3542-3.patch, > OOZIE-3542-4.patch, OOZIE-3542-amend-01.patch > > > Currently, ECPolicyDisabler checks if the local hdfs implementation has the > necessary methods to get and set erasure coding policy. However, if the > namenode implementation is old, it could throw a > org.apache.hadoop.ipc.RemoteException with > RpcErrorCodeProto.ERROR_NO_SUCH_METHOD value in it. > In this case, ECPolicyDisabler fails, and prevents the installation to > succeed. > This case should be handled just like, when erasure coding is not supported. -- This message was sent by Atlassian Jira (v8.3.4#803005)