Re: Create Index Mysteriously Slow at the End - IDS 11.10.UC2W2

Hello Andrew,

hmmm sound like there are issues with checkpointing....
i suggest to file a case @[EMAIL PROTECTED]
 this may bite all of us
someday...


anyway you could
try and generate your own stuff:
create procedure generatechkpt()
define dirty decimal (4,3);


while (1=1)

        select ( sum(lru_nmod) / sum ( lru_nfree + lru_nmod ))
                                into dirty
        from syslrus           ;

        if (dirty < 0.75 ) then
                system "sleep 1";
        else
                system "onmode -B";
        end if
        end while ;
end procedure;

execute the above as informix and set lrumax and mindirty to 99.
when the index build is done then do manually an onmode -B followed by
a onmode -c

see you

Superboer.




On 16 mei, 00:09, "Andrew Ford" <af...@[EMAIL PROTECTED]
> wrote:
> "Superboer" <superbo...@[EMAIL PROTECTED]
> wrote in message
>
> news:475f7a43-274a-4512-ab0f-6d872018f293@[EMAIL PROTECTED]
>
> > what happens when you try onmode -B
>
> With LRUs set to 127, Cleaners set to 127, LRU min/max set to 70/80,
> RTO_SERVER_RESTART disabled and CKPTINTVL set to 8 hours I execute
onmode -B
> and continuously execute onmode -B each time I see the cleaner stop
writing
> to a chunk (state = C).
>
> Doing this I get about the same performance as my previous test of
>
> RTO_SERVER_RESTART 0, aka disabled
> CKPTINTVL 28800, 8 hours
> LRUs 1
> CLEANERS 1
> LRU min dirty 0.00000
> LRU max dirty 0.00001
>
> The only difference seems to be most of the writes are chunk writes and
not
> LRU writes, no change in performance.
>
> Still 5 minutes to scan the data pages, 10 minutes to write the index
pages
> and 20 minutes to read the index pages.
>
> Thanks for the suggestion,
>
> Andrew