Trying again due to over quoting. Hi Ross, Quick answer is create a baseline on the one you like. Why it flipped takes a little more work. CURSOR_SHARING is probably not your friend. And having a TO_NUMBER interjected on one side of a join is also a bad sign. Kerry Osborne Enkitec blog: kerryosborne.oracle-guy.com twitter: https://twitter.com/KerryOracleGuy On May 1, 2013, at 3:15 PM, Ross Lafferty wrote: > Running into interesting behavior with execution plans in 11.2.0.3 (running > on Exadata x3 platform). The below query was executed at roughly the same > time, against the same environment, originating from different clients > (different machines all together). The first one listed below, using hash > joins, returns results in ~1 minute. The second one listed below, using > nested loops, doesn‚t return, even after hours. No alter sessions have been > performed. OPTIMIZER_MODE for the environment is ALL_ROWS (which should > gravitate towards hash joins). > > My question is, why the varying execution path for the same query, same > environment, being run at the same time - and any way to correct it? > Thanks in advance! > -- > B. Ross Lafferty > ============== > E: ross.lafferty@xxxxxxxxx > ============== > > > -- > B. Ross Lafferty > ============== > M: 412-608-7505 > E: ross.lafferty@xxxxxxxxx > ============== > > > > > -- > //www.freelists.org/webpage/oracle-l > > -- //www.freelists.org/webpage/oracle-l