[ https://issues.apache.org/jira/browse/HADOOP-973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12472826 ]
Raghu Angadi commented on HADOOP-973: ------------------------------------- I think we should have an explanation if this is expected. Currently it might be possible for namenode to ask to invalidate a block twice. This is not necessarily an error. In that case, not finding a file for a block might not be an exception but just a warning written to log. Also once we have a file for a block, there is no reason why 'volume' or 'parent' could be null. In that sense those conditions should not be handled as IOExceptions. > NPE in FSDataset during heavy Namenode load > ------------------------------------------- > > Key: HADOOP-973 > URL: https://issues.apache.org/jira/browse/HADOOP-973 > Project: Hadoop > Issue Type: Bug > Components: dfs > Affects Versions: 0.10.1, 0.11.0 > Reporter: Nigel Daley > Attachments: invalidDeletion.patch > > > Running namenode delete benchmark, I saw many of these NPE's traces in a > datanode log: > ... > 2007-02-04 16:17:53,744 INFO org.apache.hadoop.dfs.DataNode: Deleting block > blk_-949112377379974621 > 2007-02-04 16:17:53,744 INFO org.apache.hadoop.dfs.DataNode: Deleting block > blk_748220495971269869 > 2007-02-04 16:17:54,619 INFO org.apache.hadoop.dfs.DataNode: Deleting block > blk_658746956471887871 > 2007-02-04 16:17:56,752 ERROR org.apache.hadoop.dfs.DataNode: Exception: > java.lang.NullPointerException > at org.apache.hadoop.dfs.FSDataset.invalidate(FSDataset.java:519) > at org.apache.hadoop.dfs.DataNode.offerService(DataNode.java:410) > at org.apache.hadoop.dfs.DataNode.run(DataNode.java:1033) > at java.lang.Thread.run(Thread.java:595) > 2007-02-04 16:18:01,754 INFO org.apache.hadoop.dfs.DataNode: using > BLOCKREPORT_INTERVAL of 3329089msec > 2007-02-04 16:18:02,577 INFO org.apache.hadoop.dfs.DataNode: Deleting block > blk_1657305102826276927 > 2007-02-04 16:18:02,577 INFO org.apache.hadoop.dfs.DataNode: Deleting block > blk_589129256174830133 > ... -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.