>Did you start the hbase2 over an hbase1 dataset. I've seen the exception
you note when I've done this. The startup keeps going over this exception,
right? (IIRC, its a complaint reading a file written w/ hbase1... We fail
to read in the bloom filter which is not the end-of-the-world).

Checking the related code.  There seems no issue..  We are not going
to even fail reading the bloom.  In 2.0 code base we expect the
comparator class name to be null.  But in 1.x we write old KV based
Raw Bytes comparator class name.  So reading that back, we will return
class name as null and we get NPE it looks like.

 else if (comparatorClassName.equals("org.apache.hadoop.hbase.KeyValue$RawBytesComparator")
        || comparatorClassName.equals("org.apache.hadoop.hbase.util.Bytes$ByteArrayComparator"))
      // When the comparator to be used is Bytes.BYTES_RAWCOMPARATOR,
we just return null from here
      // Bytes.BYTES_RAWCOMPARATOR is not a CellComparator
      comparatorKlass = null;

We can better do a null check before trying the comparator class
instantiation so that we can avoid this scary error logs :-)   But
very minor..   So absolutely no issues here.


On Mon, Mar 5, 2018 at 8:26 PM, Chia-Ping Tsai <[EMAIL PROTECTED]> wrote: