Bob
2006-09-26 15:21:47 UTC
This problem will surely tax the brains of all of the experts on this list.
We have had an Index Server Catalog running for three years on Windows 2000
SP4. It has slowly built up ober time to 600,000+ documents, all in a dated
folder scheme, like 2006/09/30 ie by year then by month then by day . Each
daily folder will contain about 300 html and jpg files.
We index the html documents using metatags and the 'htmlprop.dll' tag
filter.
Up until a couple of months ago , it ran smoothly.. Each day a new folder
was dropped in , IS would come in , take about two minutes to index it and
that day's html files were then available to our web application.
Then, one day, it started to act very funny....when we added a day to the
catalog tree, IS would take hours ( its set as a 'dedicated server') to do
'master merge' after master merge , non-stop . If you stopped and restarted
IS it would go back to normal.
After a couple of weeks of this, I tore down the catalog completely and
re-created the web catalog, using just a few days worth of data, then
gradually added files back in to be indexed again. After only a month of
data was added, it started going 'funky' again.
You would add a day's worth of files then IS ( cisvc.exe) would start going
back over ALL the files in the catalog before it came back and answered
queries, sometimes a couple of hours later. Then we found that , for the
month of days that we had added, IS had NOT actually indexed all of the
content. Now, whenever you add a few days , IS indexes normally....but
sooner or later it goes into this mode where it seemingly 're-indexes' ALL
of the files, but doesn't really, it misses some anyway.
I have been using NT Filemon to watch CISVC open files , so I can see what
files it is hammering on at any point in time.
To demonstrate even funkier behaviour, if you throw the catalog into 'Query
Only' mode when it goes into one it's funks, then 'start' it again later,
it starts back at the beginning of the catalog, looking at ALL the files all
over again.......
Does anybody have any suggestions at all? I am basically at wit's end , run
out of things to try..... thanks in advance to anyone who answers.....
We have had an Index Server Catalog running for three years on Windows 2000
SP4. It has slowly built up ober time to 600,000+ documents, all in a dated
folder scheme, like 2006/09/30 ie by year then by month then by day . Each
daily folder will contain about 300 html and jpg files.
We index the html documents using metatags and the 'htmlprop.dll' tag
filter.
Up until a couple of months ago , it ran smoothly.. Each day a new folder
was dropped in , IS would come in , take about two minutes to index it and
that day's html files were then available to our web application.
Then, one day, it started to act very funny....when we added a day to the
catalog tree, IS would take hours ( its set as a 'dedicated server') to do
'master merge' after master merge , non-stop . If you stopped and restarted
IS it would go back to normal.
After a couple of weeks of this, I tore down the catalog completely and
re-created the web catalog, using just a few days worth of data, then
gradually added files back in to be indexed again. After only a month of
data was added, it started going 'funky' again.
You would add a day's worth of files then IS ( cisvc.exe) would start going
back over ALL the files in the catalog before it came back and answered
queries, sometimes a couple of hours later. Then we found that , for the
month of days that we had added, IS had NOT actually indexed all of the
content. Now, whenever you add a few days , IS indexes normally....but
sooner or later it goes into this mode where it seemingly 're-indexes' ALL
of the files, but doesn't really, it misses some anyway.
I have been using NT Filemon to watch CISVC open files , so I can see what
files it is hammering on at any point in time.
To demonstrate even funkier behaviour, if you throw the catalog into 'Query
Only' mode when it goes into one it's funks, then 'start' it again later,
it starts back at the beginning of the catalog, looking at ALL the files all
over again.......
Does anybody have any suggestions at all? I am basically at wit's end , run
out of things to try..... thanks in advance to anyone who answers.....