Re: Checkpoint take so long
Sean Kelsey wrote:
>
> Why $2.73 ???
>
> Kevyn Schneider wrote in message <36d6ced6.236919452@news.onramp.net>...
> >I'm dumb. CHKPTINTVL is obviously set to 300. I'll bet $2.73 that
> >MIN=50, MAX=60, PHYSDBS=rootdbs and PHYSFILE=1000. ;-{>
> >
> >On Fri, 26 Feb 1999 14:41:03 -0000, "Sean Kelsey"
> ><chilliinc@hotmail.com> wrote:
> >
> >>Taking on board what Kevyn has already said, it looks like there was some
> >>sort of data load happening during the long checkpoints. I also noticed
> that
> >>your logs aren't being archived. Whats the story ??
> >>
> >>What are your current settings for LRU_MAX/MIN_DIRTY & CHKPTINTVL and how
> >>big is the physical log ?? ( this is usually the reason on data loads as
> the
> >>physical log will get rather full )
> >>
> >>Sean
> >>
> >>
> >>trieu@ti.L-3Com.com wrote in message <7b4o00$s3j$1@nnrp1.dejanews.com>...
> >>>Hi all,
> >>> Can some one please tell me how to reduce checkpoint time.
> >>>My checkpoint completed sometime go up to 80seconds. Please let know
> >>>how to turning this. Thanks
> >>>
> >>>onstat -m> >>>14:59:00 Checkpoint Completed: duration was 29 seconds.
> >>>14:59:50 Logical Log 7927 Complete.
> >>>15:01:04 Logical Log 7928 Complete.
> >>>15:02:29 Logical Log 7929 Complete.
> >>>15:04:27 Checkpoint Completed: duration was 27 seconds.
> >>>15:08:16 Logical Log 7930 Complete.
> >>>15:08:54 Logical Log 7931 Complete.
> >>>15:09:38 Checkpoint Completed: duration was 10 seconds.
> >>>15:09:43 Logical Log 7932 Complete.
> >>>15:10:18 Logical Log 7933 Complete.
> >>>15:10:54 Logical Log 7934 Complete.
> >>>15:11:38 Logical Log 7935 Complete.
> >>>15:14:57 Checkpoint Completed: duration was 20 seconds.
> >>>15:19:59 Checkpoint Completed: duration was 1 seconds.
My guess? You were doing a large data load and did not have enough
CLEANERS and/or LRUS and you have LRU_MAX_DIRTY and LRU_MIN_DIRTY set
too high for that level of activity. Increase the number of LRUS (try
at least 32 but avoid 64 and 96), increase CLEANERS to equal LRUS, and
reduce the LRU_MAX/MIN_DIRTY parameters to 10/5 or even as low as 1/0.
Art S. Kagel