RE: Saving STATSPACK Reports

  • From: "Lawie, Duncan" <duncan.lawie@xxxxxxxxxxxxxxxxx>
  • To: "'Polarski, Bernard'" <Bernard.Polarski@xxxxxxxxxxxxxx>, bnsarma@xxxxxxxxx, "_oracle_L_list" <oracle-l@xxxxxxxxxxxxx>
  • Date: Mon, 11 Sep 2006 04:08:35 -0400

as I said, this is fundamentally about trend analysis.  I see this as part of 
the arsenal, not a replacement for statspack - a way to work out which of the 
hundreds of statspack reports you may actually want to read.
The uses I have found for this are that you can look at the shape of events 
over a few hours and work out which statspack reports you want to look at.  Or 
you can see at a glance (as a marketeer would say) that you have a daily spike 
at 4am.  Or you can look at your database growth history over the last 18 
months and project outwards for capacity planning.
It is also extensible, so that probably anything numeric you find in a 
statspack could be graphed.


From: Polarski, Bernard [mailto:Bernard.Polarski@xxxxxxxxxxxxxx] 
Sent: 11 September 2006 08:07
To: duncan.lawie@xxxxxxxxxxxxxxxxx; bnsarma@xxxxxxxxx; _oracle_L_list
Subject: RE: Saving STATSPACK Reports

Nice tool, I recognize the graphic lib used their but the I have still one 
question : 
However, how is it going to help you in ANY case when one of your DB has run in 
Basically  if you don't have  the waits event and the hash_value, nor the SQL 
text nor any locking real info, in facts you have nuts. The tool is just going 
to tell you that something that is poisened is not in good health. At least in 
statpacks you have a section with higher SQL and their waits states and can 
give an hint on were to look. here you have just figures, numbers and no 
indications on which SQL to look at. AFAIK, this is 100% useless, it is just a 
diary of figures but you can't extract a direction to investigate from it.
B. Polarski


From: Lawie, Duncan [mailto:duncan.lawie@xxxxxxxxxxxxxxxxx] 
Sent: Friday, 08 September, 2006 6:06 PM
To: 'bnsarma@xxxxxxxxx'; _oracle_L_list
Subject: RE: Saving STATSPACK Reports

If your purpose is overall trend analysis, something like orca_oracle ( <>  ) may be of 
use to you (disclosure: I wrote it)
It requires some perl knowledge to make the best use of, but will retain the 
headline stats from perfstat for a long time for trend analysis.

Please access the attached hyperlink for an important electronic communications 

Other related posts: