Re: Index "Fillfactors" - Results...

Greg Moye wrote: > > Greg Moye wrote: > > > > Has anyone noticed that when v7 (Specifically 7.12.UC1 for me) builds > > indexes it usually leaves leaf pages rather sparsely packed?? Averages > > of 700 to 1200 bytes free out of 2k bytes is common for me. > > > > tbcheck -pT output shows this pretty clearly, regardless of what > > FILLFACTOR is set to. FWIW, V5 index build looks essentially identical. > > snip...... > > Well - its a bug. With at least some kinds of data (not quite sure what > kind, but I apparently have lots of it...) No matter what you set > FILLFACTOR to, you'll frequently get indexes that are 40-60% full. Bug > is #52950. > Well, I hate to be the bearer of bad news but I just checked our bug database and this bug has been asserted as a documentation problem. The test case provided for this bug was a case where the a bottom up index build can't work with fillfactor very well. It has to do with the fact that it is a duplicated index and that causes us to put the key in a new node before we completely reach fillfactor. The documentation will be changed to state that we will only be able to approach fill factor where indices are reasonably unique or tend to have short dups lists. On a side note, there is another bug (don't remember the number) where if you used the fillfactor clause in the create index statement we ignored it and you actually had to set fill factor in the onconfig file before you'd see a difference. Jacques > On the bright side, when it gets fixed, it should cut my (and possibly > your...) index page reads by about 40-60% - i'll take that... > > -- > Greg Moye > Electronic Data Systems Corp. - Plano TX -- ******************************************************************** * Jacques P. Renaut "I'd dazzle you with brilliance * * Informix Advanced Support if I only had the knack..." * * email: jrenaut@informix.com #include <disclamier.h> * ********************************************************************