Kbase P76140: Why in Promon R&D 3/4 the values in 'Flushes' column can be larger than in 'CPT Q'?
Autor |
  Progress Software Corporation - Progress |
Acesso |
  Público |
Publicação |
  4/9/2004 |
|
Status: Unverified
GOAL:
Why in Promon R&D 3/4 the values in 'Flushes' column can be larger than in 'CPT Q'?
CAUSE:
Ckpt ------ Database Writes ------
No. Time Len Dirty CPT Q Scan APW Q Flushes
9 19:58:57 234 23854 20667 0 0 2683
8 19:55:31 206 22227 17731 0 0 3085
7 19:51:46 225 31210 20536 0 0 8502
6 19:46:05 341 36003 4149 3754 0 12996
5 19:43:06 179 37149 14443 0 0 19780
4 19:38:10 296 25869 3741 1588 0 8798
3 19:34:17 233 37691 18567 0 0 15602
2 19:30:12 245 27570 12472 0 0 10613
1 19:27:56 136 24743 11314 0 0 13429
0 19:22:36 320 0 0 7747 0 0
FIX:
Because the CPT Q column is not the number of buffers in the checkpoint queue, but the number of buffers that were /written from/ the queue. The last four columns are all under the heading "database writes".
The "Dirty" column is the initial number of buffers that were placed on the queue at the beginning of the checkpoint. And the "Flushes" column is the number of buffers that had to be written at the end of the checkpoint because they could not be written during the checkpoint. No database changes or transaction state changes can occur during the flushing.
NB: there is a fifth kind of write that can occur that is not shown here. That is when we want to read a new block into the buffer pool, but had to write one out first to make room for it.