Re: SeqScans Statistics 7.3_uc6

"Art S. Kagel" wrote: > mattb5@ozemail.com.au wrote: > > > > In article <7bfr9s$uoh$1@nnrp1.dejanews.com>, > > mattb5@ozemail.com.au wrote: > > > In article <36D5C1A6.256D@bloomberg.net>, > > > kagel@bloomberg.net wrote: > > > > mattb5@ozemail.com.au wrote: > > > > Folks, > > > > I have found something interesting now: > > > > Here are two SQL's that are practically the same, if you run these SQL's with > > SET EXPLAIN ON, both report INDEX PATH as you would expect. However the second> > SQL increments the sysptprof.seqscans by 1 where tabname = "msf700" > > > > Question? Do you have any other evidence that a sequential scan is > ACTUALLY happening? Perhaps the bug is that sysptprof.seqscans is being > incorrectly updated!?!? Try zeroing stats and run the first query then > look at onstat -D for I/O's against the chunk and onstat -p for total > I/O and buffer activity. Then zero stats again and run the second > query. Is there any significant difference? > > Art S. Kagel Hi, there is a bug related to the "keyfirst" feature that was introduced with 7.3. Due to this bug the count of sequential scans as found in sysptprof (or as displayed in 'onstat -p') may be higher than the number of sequential scans actually executed. If Explain reports an index path and at the same time seqscans is incremented I am pretty sure that you hit this problem. It will be fixed in 7.31 (and probably some 7.30.UCx release as well). Heiko ============= all disclaimers apply ================