Re: NPE in TableInputFormatBase$TableRecordReader.restart() (line 110)

[stack <st...@duboce.net>]

On Tue, Jan 12, 2010 at 1:05 PM, Buttler, David <bu...@llnl.gov> wrote:

>
> Looking at the hbase source code suggests that for some reason the htable
> field is not being initialized.  However, according to my console, I get
> through approximately 26% of the map tasks and 9% of the reduce tasks before
> it fails [although I am not sure how accurate that is as I don't see any
> changes to my hbase table].
>
> I get this result with both my own code an running
> org.apache.hadoop.hbase.mapreduce.RowCounter
>


My guess is that its something to do with your environment -- that and then
we're not handling whatever it is gracefully.  RowCounter at least works for
me.   Its odd that you are 26% of the ways through the map task when this
happens (Is this so w/ rowcounter?).  That would almost imply the problem
transient since a bunch of map tasks have completed already.  What if you
added a bit of debug to TIF#setConf and see what its passing into setHTable.

St.Ack