Re: LRU writes

Graeme Sargent writes: |> Can anyone explain exactly what triggers an LRU write in Informix- |> OnLine, or what the distinction is between LRU writes and Idle writes? |> |> According to the documentation (and the Informix training courses) the |> only thing that triggers an LRU write is a Foreground write, and yet I |> commonly see tbstat reporting 0 Fg Writes along with non-zero LRU and |> Idle writes. |> |> If we believe that the statement "An LRU write occurs as a result of |> tbinit prompting, instead of as a result of the page cleaners waking by |> themselves" (OnLine Administrator's Guide, P2-71) then what events will |> cause tbinit prompting, other than a Fg write? As of release 5.0, the following is true. Idle writes are performed *only* when a page cleaner wakes up, takes a look at the %dirty on a given LRU queue and realizes it exceeds the LRU_MAX_DIRTY threshold. LRU writes are performed, as you say above, when the page cleaners are prompted to do so by tbinit. Tbinit will perform that prompting: - after 16 fg writes have been performed (multiples of 16); - whenever a buffer is moved from the clean to the dirty channel of an LRU queue, and it is seen that the % dirty for that queue exceeds LRU_MAX_DIRTY. It is perfectly normal to see mostly LRU writes in a tbstat -F (depending on how you've tuned your checkpoints - in some cases, Chunk writes will predominate), or at least more LRU writes than Idle writes. Dave Disclaimer: These opinions are not those of Informix Software, Inc. ************************************************************************** "I look back with some satisfaction on what an idiot I was when I was 25, but when I do that, I'm assuming I'm no longer an idiot." - Andy Rooney