We have a cassandra (simple) cluster of two nodes.
During recovery from a user-created dump, we found that when we have lines like
del column_family['row_1'];
set column_family['row_1']['column1'] = '123';
set column_family['row_1']['column2'] = '456';
set column_family['row_1']['column3'] = '789';
... often the first column (column_1) is NOT set.
We used:
$ cassandra-cli -h cassandra.host.name -k keyspace_name -f dump_file
We use cassandra 1.0.10 on the Debian host.
The dump is always discarded and then recreates the key space, so it is actually empty when we execute it. We know that the delete statement is not actually needed.
We deleted it, but we still cannot understand why this is happening. I suppose this is due to the fact that this is the design, and that we just missed, but we can’t understand that EXACTLY goes “wrong.”
St0rm source