Re: Context Cartridges

This last paragraph was the one I was waiting for, this usually seems to occur in this type of situation. See if you can track the item that is in the ctx_all_dml_queue back to the actual record and remove it (at least temporarily). Then clear anything and everything out of your "pending" and "in progress" queues, drop and recreate the indexes. Sometimes, for some reason, even with plain vanila ASCII data you come across a document that context chokes on. Nobody from Oracle has ever been to explain this satisfactorily to me except to say that the document must be corrupt in some way. In truth, I have usually found some control character or other that has somehow got itself buried in an ascii text document that I have been able to remove which fixes the problem.

Context does reindex properly, within the bounds of the architecture that has been devised (which I won't go into here - because I don't remember enough about it!). If your indexes are particularly busy, with lots of deletes in particular, it may be advisable to drop and recreate them on a fairly regular basis if your operations can afford it.

HTH Rod J. Stewart

ks <ks_at_yahoo.com> wrote in message news:37D7E819.4CA091D2_at_yahoo.com...
> Thanks for the response.
> Basically the policies are direct data stores of plain ASCII text data
with
> mostly default settings (I added a definition). The indexes are basically
the
> same -- default settings.
> I agree with you on the trace files, altho I am sure Oracle support will
want
> to see them :-).
> I guess the approach I could take would be to build my own trigger to
reindex
> when an insert, update or delete takes place, because from what I have
been
> told
> Context doesn't reindex correctly (or at all) anyhow when it needs to.
> However, I don't really feel comfortable taking that approach either,
because
> I would rather do the
> reindexes in batches to avoid fragmentation.
>
>
>
> I'll see what Oracle discovers, and see if I can narrow down a little
further
> when the DML server dies. One thing I do notice, and this could be my
> problem???, is that
> there is a pending reindex in the CTX_ALL_DML_QUEUE. I am not able to
clear
> that out.....but the policies and indexes do exist. Ideas?
> Thanks again.
Received on Thu Sep 09 1999 - 16:40:24 CDT