Part 02
Here I try to show you what happens and why Oracle updates the V$ASH view. Please read the pdf by Graham Wood, for more details.
What I analyze is the session "session_id=1732" with "session_serial#=31065" saw in the previous post.
This is what happens
 |
| Click for enlarge |
The session 1732 (serial# 31065) is sampled 3 times: at sample_id 50243482, 50243483, 50243484. In the last sample_id, there is a change of state of the session. In fact, querying the V$ACTIVE_SESSION_HISTORY, you see
 |
| Click for enlarge |
But what happens behind the lines is something like this. Pay attention that my conclusions are based on my experiment
 |
| Click for enlarge |
(1) SAMPLE_ID 50243482/SAMPLE_TIME 15-DEC-17 02:50:40.054
At the 15/12/2017 14:50:40.838429, one row is inserted in X$ASH with SAMPLE_TIME, 14:50:40.054
(2) SAMPLE_ID 50243484/SAMPLE_TIME 15-DEC-17 02:50:42.064
At the 15/12/2017 14:50:42.016313, one row is inserted in X$ASH with SAMPLE_TIME, 14:50:41.064
(3) SAMPLE_ID 50243484/SAMPLE_TIME 15-DEC-17 02:50:41.064
At the 15/12/2017 14:50:43.192604, one row is inserted in X$ASH with SAMPLE_TIME, 14:50:42.064
Note that, at this time, the session 1732 is waiting, but Oracle doesn't know yet how match time the session will wait
(4) SAMPLE_ID 50243484/SAMPLE_TIME 15-DEC-17 02:50:41.064 (the same of the previous one)
At the 15/12/2017 14:50:44.372309 (sample_id 50243485), the last row is updated in X$ASH with SAMPLE_TIME, 14:50:42.064
At sample_id 50243485, the session 1732 finish it works and Oracle know how much time the session waited. So he can update the state of the wait on X$ASH
Here I left the previous row, just to show what happening.
(5) The final result
What you see finally, is a consistent state of the session
1 commento:
Excellent series of posts! Well done friend!! =]
Posta un commento