I'm using tables - dba_hist_sqlstat, dba_hist_snapshot and dba_hist_sqltext to get stats for groups of similar sql statements over the entire period of time since instance start up. I just found the same sql_id belonging to two different snapids (snapshots).
+-------+---------+---------------+-------------+---------------+--------------+--------------+
|SNAP_ID|DBID |INSTANCE_NUMBER|SQL_ID |PLAN_HASH_VALUE|OPTIMIZER_COST|OPTIMIZER_MODE|
+-------+---------+---------------+-------------+---------------+--------------+--------------+
|63618 |622294766|1 |0ps2wsx1rjv8q|2871982686 |4 |ALL_ROWS |
|63522 |622294766|1 |0ps2wsx1rjv8q|2871982686 |4 |ALL_ROWS |
+-------+---------+---------------+-------------+---------------+--------------+--------------+
So, is a "group by sql_id" sufficient ? should i group by "sql_text" ? Is there a better way ?
[UPDATE]
Here is the query i've written so far -
select
dh_sql.sql_id,
dh_sql.plan_hash_value,
dh_sql.parsing_schema_name,
dh_sql.module,
trunc(dh_snap.end_interval_time) "DATE",
max(dh_sql.snap_id),
max(dh_sql.dbid),
sum(dh_sql.executions_delta),
.
.
.
.
from
dba_hist_sqlstat dh_sql,
dba_hist_snapshot dh_snap
where
dh_sql.snap_id = dh_snap.snap_id
and dh_snap.begin_interval_time > to_date(?, 'MM/DD/YYYY')
and dh_snap.begin_interval_time < to_date(?, 'MM/DD/YYYY') + 1
and dh_sql.executions_delta > 0
and dh_sql.module is not null
group by
dh_sql.sql_id,
dh_sql.plan_hash_value,
dh_sql.parsing_schema_name,
dh_sql.module,
trunc(dh_snap.end_interval_time)