Full Population - completes but never ends

We rebuilt a catalog this weekend after upgrading to Adobe iFilter
version 6. Have attempted the rebuild twice, both times the result has
been that all rows are indexed, but the population never 'ends'. The
status remains at '1 - Full Population in Progress'. The item count
indicates 74,926 - which is one more than our (current) row count of
74,925. The mssearch and mssdmn processes continue to consume CPU
(~50%).
We're on Win2K, Sql2K, sp3A. There do not appear to be any error
messages in the event log.
The catalog contains one table, and it is configured with Change
Tracking. The table has one indexed column, an image column containing
primarily Word and PDF binaries.
Any help/suggestions appreciated!
Dave Hensler
dhensler@.prestongates.com
The item count reflects the numbers of rows being indexed and the numbers of
tables being full text indexed, so the number seems correct.
Can you use profiler to see if anything is being pulled? Look for events
coming with an Application Name of PKM.
Hilary Cotter
Looking for a SQL Server replication book?
http://www.nwsu.com/0974973602.html
Looking for a FAQ on Indexing Services/SQL FTS
http://www.indexserverfaq.com
"Dave Hensler" <dhensler@.prestongates.com> wrote in message
news:1118069691.983721.165480@.o13g2000cwo.googlegr oups.com...
> We rebuilt a catalog this weekend after upgrading to Adobe iFilter
> version 6. Have attempted the rebuild twice, both times the result has
> been that all rows are indexed, but the population never 'ends'. The
> status remains at '1 - Full Population in Progress'. The item count
> indicates 74,926 - which is one more than our (current) row count of
> 74,925. The mssearch and mssdmn processes continue to consume CPU
> (~50%).
> We're on Win2K, Sql2K, sp3A. There do not appear to be any error
> messages in the event log.
> The catalog contains one table, and it is configured with Change
> Tracking. The table has one indexed column, an image column containing
> primarily Word and PDF binaries.
> Any help/suggestions appreciated!
> Dave Hensler
> dhensler@.prestongates.com
>
|||Profiler shows lots of these under PKM:
exec sp_fulltext_getdata 29, 1835153583, 0x00002D31, 1
They are firing many times per second and seem to be originating in the
master database.
|||This is interesting - I have found 3 pdf files in the FTDATA directory
that are corrupt. They will not open successfully in Acrobat. It
looks like they have been there for about 29 hours.
Seems reasonable to try to remove these files from the database and see
if it frees up the search service. However, I need to identify the
rows containing these binaries. The file names have the form
'0000001f7d78a4e70000BF9D00000003.pdf'. Any idea how to locate these
in the table? Or is there another log that may identify the row where
these binaries came from?
Thanks for any advice.
|||This means that the population is still in progress.
go to
http://msdn.microsoft.com/library/de...tml/sp04f9.asp
Download the code sample and run it in VB 6. It will tell you where you are
in your population.
Hilary Cotter
Looking for a SQL Server replication book?
http://www.nwsu.com/0974973602.html
Looking for a FAQ on Indexing Services/SQL FTS
http://www.indexserverfaq.com
"Dave Hensler" <dhensler@.prestongates.com> wrote in message
news:1118078599.081054.208540@.g49g2000cwa.googlegr oups.com...
> Profiler shows lots of these under PKM:
> exec sp_fulltext_getdata 29, 1835153583, 0x00002D31, 1
> They are firing many times per second and seem to be originating in the
> master database.
>
|||Thanks for the tool, Hilary. Using it I can see that the population is
cycling within a small set of database rows. The rows are not
necessarily recent, in fact they are spread all over the table. I've
looked at the individual binaries and do not see any pattern - some are
newer versions of the pdf file format, some are older, etc.
Have you seen anything like this before? How about Adobe's latest
filter (ifilter 6.0) - any known issues?
Thanks - Dave
|||The problem seems to be solved. The indexer was repeatedly cycling
over 5 rows containing pdf binaries. Once the 5 rows were identified
and the binaries removed from the table, the indexer went into the Idle
state. When they are reinserted into the table, the indexer hangs up
again until they are removed. These pdf binaries do not appear to be
damaged or unusual when they are opened in Acrobat, however.