Hey David, > Here is what I get with 11.2.0.3 when I do the recommended actions specific > to SYS.X$KCCRSR (fixed table stats have been gathered): > > SQL> exec dbms_stats.DELETE_TABLE_STATS('SYS','X$KCCRSR'); [snip] > Statistics > ---------------------------------------------------------- > 26 recursive calls > 4 db block gets > 41 consistent gets [snip] > SQL> exec dbms_stats.gather_Fixed_objects_stats() [snip] > Statistics > ---------------------------------------------------------- > 1 recursive calls > 0 db block gets > 0 consistent gets [snip] > To be honest I'd be worried that deleting the stats (whatever they may be) > and locking them will bite you at some point down the line. Since fixed > tables don't get dynamic sampling in the absence of statistics keeping the > computed stats would seem to be the best course of action. I read that dynamic sampling on fixed tables was allowed in 11.2 (bug 6608941). Also, wouldn't you have to UNLOCK_TABLE_STATS before collecting stats again for it to take effect? I don't know how fixed tables are affected by the buffer cache, but I get similar autotrace results just running the SELECT (against V$RMAN_STATUS) twice in a row, regardless of the state of stats on X$KCCRSR. Thanks! Rich -- //www.freelists.org/webpage/oracle-l