Re: onstat -g dic question

Art S. Kagel wrote: > andy lennard wrote: > > > > An aside to this. On Solaris 2.6, 7.23.UC4 we had a problem with loads > > <snip> > > > > > Can't find anything in the manuals for onstat -g dic -- except that the > > > command exists. Does anyone know what the dirty? column means? > > The dirty column indicates whether that dictionary entry was loaded > from a definition of the table that is no longer valid, in effect the > table has been altered in some way to invalidate the dictionary entry > and noone has accessed the entry since so it has not been reloaded. > If the table was altered there may be a new entry in another bucket so > the old one is marked dirty so sessions know to either rehash to find a > 'clean' version in cache or to reload cache with the new version. > > Art S. Kagel This was kind-of what I thought it would be. However, I am getting many dirty = 'yes' entries ( I just looked and this is not true right now -- but has been a usual occurrence -- maybe 10-15 of them by the end of a week's up-time). This is with no changes to the table's schema, constraints, triggers, etc. Our engine is crashing regularly -- usually Mondays when our volume goes up from the weekend. It is usually (but not always) an spipe error. When our 'Monday Crash' of two weeks ago was an assertion failure in rsinit.c, Informix support suggested it was a problem with exceeding the dictionary entries (which is when I started looking at the -g dic command). We are on 7.22UC2 (HP-UX) -- which I guess was not one of their better releases. We're moving to 7.30UC5 within two weeks, and hopefully all this will go away. Tom Grenier DBA, Biztravel.com