How do you track sql queries in Oracle 12c? - oracle

Question

We are in the process of upgrading to Oracle 12c and I need to track the queries being executed by the application. In other words if the application executes a query like select 'foobar' from dual; I would like see the text "select 'foobar' from dual" in the output file.
If I follow the instructions here: https://docs.oracle.com/database/121/TGSQL/tgsql_trace.htm#TGSQL809 I get files that contain statistics like the following but not the actual sql queries.
WAIT #0: nam='rdbms ipc message' ela= 2999770 timeout=300 p2=0 p3=0 obj#=-1 tim=1103506389
WAIT #0: nam='rdbms ipc message' ela= 9854 timeout=1 p2=0 p3=0 obj#=-1 tim=1103522400
*** 2016-04-07 15:07:20.715
WAIT #0: nam='rdbms ipc message' ela= 2999585 timeout=300 p2=0 p3=0 obj#=-1 tim=1106522506
WAIT #0: nam='rdbms ipc message' ela= 9690 timeout=1 p2=0 p3=0 obj#=-1 tim=1106532500
If I look for the query like this I get 0 results: grep -rnw "foobar" --include=*.trc ./

Answer 1

One option is looking up the AWR repositories for it. It will keep a few days worth of SQLs. There's plenty of additional information in the system views, so this is strictly the text, but feel free to explore.
SELECT DISTINCT u.username, to_char(substr(h.sql_text, 1, 4000)) sqltxt
FROM dba_hist_sqltext h
JOIN dba_hist_active_sess_history a ON a.sql_id = h.sql_id
JOIN dba_users u ON u.user_id = a.user_id
WHERE username = 'SYS';
I filtered the results for SYS just as an example, but you can change it as you wish.

Answer 2

If you would like to see all the activity the best thing you should do is have an EM (Enterprise Manager) set for you.
If you don't, gv$activity_session_history would be a good call, it's better seen when using grouping functions. Simply selecting on it would be a mess depending on the number of calls your application is pushing.
Another way, you could see it in an average manner:
`
select s.parsing_schema_name,
inst_id,
sql_id,
plan_hash_value,
child_number,
round(nullif(s.ELAPSED_TIME, 0) / nullif(s.EXECUTIONS, 0) / 1000000, 4) elap_per_exec,
round(s.USER_IO_WAIT_TIME / nullif(s.ELAPSED_TIME, 0) * 100, 2) io_wait_pct,
round(s.CLUSTER_WAIT_TIME / nullif(s.ELAPSED_TIME, 0) * 100, 2) cluster_wait_pct,
round(s.application_wait_time / nullif(s.ELAPSED_TIME, 0) * 100, 2) app_wait_pct,
round(s.CPU_TIME / nullif(s.ELAPSED_TIME, 0) * 100, 2) cpu_time_pct,
round(s.PHYSICAL_READ_BYTES / nullif(s.EXECUTIONS, 0) / 1024 / 1024, 2) pio_per_exec_mb,
round(s.PHYSICAL_READ_BYTES / nullif(s.PHYSICAL_READ_REQUESTS, 0), 2) / 1024 read_per_request_kbytes,
round(s.buffer_gets / nullif(s.executions, 0), 4) BufferGets_per_Exec
s.executions,
to_char(s.last_active_time,'dd/mm/yyyy hh24:mi:ss') last_act_time,
s.first_load_time,
s.sql_fulltext,
s.sql_profile,
s.sql_patch,
s.sql_plan_baseline
FROM gv$sql s
WHERE 1=1
and s.parsing_schema_name in ('LIST OF DATABASE USERS YOU WANT TO MONITOR')
order by s.last_active_time desc;
`
It would give a good perspective of how well your doing based in your average thresholds.