Re: Usage Statistics (important note)

> In article <3a82al$qo@emory.mathcs.emory.edu> > cwoodall@vth1.vth.colostate.edu (Con Woodall) writes:> > > > >As has been noted before a real source of information on use is the > >logical logs, but there appears to be no good tool for extracting it > >easily in summary form. In light of Joe's comment, it should be noted > >that the log include insert, update, delete stuff and is not a source > >of information on which tables are accessed in the sense mentioned > >by Joe above. > > > > There are tblog options for by-user and by-transaction retreival, which > haveproved useful. > > I was able to use this to determine when log records were created in > relationto the begin work statements - a begin work doesnt put a > transaction in thelog until the first data is actually touched. > > There is a major caveat though - DONT INTERRUPT OUT OF TBLOG. We had a This is a very important issue. NEVER INTERRUPT OUT OF ANY OF THE ONLINE UTILITIES!!!! The effect is the same as killing the engine with KILL -9 . Many of them interface at the RSAM level to achieve their effects. > fewtimes where a ^C from tblog's voluminous output crashed an engine. > Stuff theoutput into a file or pipe, and know the difference in your > stty settingsbetween interrupt and quit. > >