Fg Writes

I'm getting a number of Fg writes on our production instance. They don't seem to occur during the day (during oltp processing) but sometime during the night, when a few of processes run, including a couple of load routines. I know Fg writes are bad, but what exactly is the text book (or real world) solution to reducing/eliminating them. Any clue is appreciated. For the Record: Solaris 2.6, 4 processors, Informix 7.30.UC5 Thx in advance..... Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't.

lower LRU_MAX_DIRTY, LRU_MIN_DIRTY. Make checkpoints more frequent either through CKPTINTVL or resizing the physical log. Increase BUFFERS, CLEANERS, LRUS. Those are the general answers (I can think of right away). LRU_M[IA][NX]_DIRTY - Control when your buffers are cleaned out, bg writes. CKPTINTVL - Time between Checkpoints (chunk writes). BUFFERS - When there are no clean buffers, you can get a fg write. CLEANERS - How quickly the LRU queues can get cleaned. LRUS - Controls the number of buffers per LRU queue. This can help reach LRU_MAX_DIRTY faster. Physical Log - When it is 75% full, you get a checkpoint. ktroxel@my-deja.com wrote: : I'm getting a number of Fg writes on our production instance. They : don't seem to occur during the day (during oltp processing) but : sometime during the night, when a few of processes run, including a : couple of load routines. : I know Fg writes are bad, but what exactly is the text book (or real : world) solution to reducing/eliminating them. Any clue is appreciated. : For the Record: Solaris 2.6, 4 processors, Informix 7.30.UC5 : Thx in advance..... : Sent via Deja.com http://www.deja.com/ : Share what you know. Learn what you don't. -- Rob Wilson rwilson@ntsource.com