Done, https://issues.apache.org/jira/browse/CASSANDRA-1188 It was find_all_by_service_id that was the culprit, and it resolves down to a multiget_slice on a super column family. The super CF is acting as an index back into a regular CF, thus I'm providing key, supercolumn name, and getting back a set of columns which have date uuid names and values as keys back into the regular CF. The leaked file descriptors seem to be for the Super CF data file.
Matt On Jun 13, 2010, at Sun Jun 13, 3:32 PM, Jonathan Ellis wrote: > Can you open a new ticket, then? Preferably with the thrift code > involved, I'm not sure what find_by_natural_key or find_all_by_service > is translating into. (It looks like just one of those is responsible > for the leak.)
