hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] Created: (HBASE-807) Lost split meta updates
Date Fri, 08 Aug 2008 20:49:44 GMT
Lost split meta updates
-----------------------

                 Key: HBASE-807
                 URL: https://issues.apache.org/jira/browse/HBASE-807
             Project: Hadoop HBase
          Issue Type: Bug
            Reporter: stack


Two different clusters suffered loss of meta table updates yesterday.  In particular, they
both lost knowledge of splits.

The manifestation of the loss was a NPE doing a get of a row that was within the range of
the split region:

{code}
NativeException: java.lang.NullPointerException: null
    from org/apache/hadoop/hbase/client/ServerCallable.java:64:in `getServerName'
    from org/apache/hadoop/hbase/client/HConnectionManager.java:900:in `getRegionServerWithRetries'
    from org/apache/hadoop/hbase/client/HTable.java:467:in `get'
    from org/apache/hadoop/hbase/client/HTable.java:449:in `get'
...
{code}

Digging in, a ghost of the daughter regions remained down in .META. store files in the form
of region historian updates.

On one cluster, there was an HDFS storm where HDFS was badly abused trying to make it work.
 On the other, original data was written with RC1 which had lots of issues.  For now, let
the HDFS storm and the fact that the data was written with RC1 'explain' the loss... but lets
keep an eye out in case we see this issue again.


-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message