In case anyone is wondering, I made this go away by changing TableOutputFormat.close() to simply call table.close(). That same change was already made in trunk as part of HBASE-3777.
On Mon, Aug 29, 2011 at 5:48 PM, Patrick Marchwiak <[email protected]> wrote: > I'm trying to upgrade my cluster from CDH3b2 (based on hbase > 0.89.20100621) to CDH3u1 (based on hbase 0.90.3) and running into > trouble in one of my integration tests. This test runs against a local > hbase instance running in pseudo-distributed mode – it puts a value, > runs a (local in-process) Hadoop job which increments that value, gets > the value, and lastly, verifies that the value was updated. The > problem is that the Get does not succeed and the following exception > gets thrown: > java.io.IOException: > org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation@d3576a2 > closed > at > org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.locateRegion(HConnectionManager.java:570) > at > org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.locateRegion(HConnectionManager.java:558) > at org.apache.hadoop.hbase.client.HTable.<init>(HTable.java:172) > at org.apache.hadoop.hbase.client.HTable.<init>(HTable.java:146) > at BasicHBaseTest.testConnectionClosing(BasicHBaseTest.java:92) > > JUnit test here: http://pastebin.com/Mn17a1jW > > This same test works fine when I run it against the "minicluster" set > up by HBaseTestingUtility. It also worked in the previous version of > HBase. This seems to be a problem with how connections are cleaned up > in TableMapper or TableReducer in the LocalJobRunner case. > > Should I file a bug? >
