[Info-Ingres] Ingres 6.4(07) urgently required
Karl & Betty Schendel
schendel at kbcomputer.com
Thu Nov 15 07:52:50 CST 2007
At 12:40 PM +0000 11/15/07, Roy Hann wrote:
><james.goodwill at googlemail.com> wrote in message
>news:8d22f941-e27b-4abd-93c0-c4439aa3401c at b32g2000hsa.googlegroups.com...
> > Over the last couple of day we've seen the following messages in the
>> Ingres error log.
>>
>> Can anyone suggest a way forward. I've increased the stack size, and
>> increased the opf.memory (300,000 bytes * 32 active_sessions). So, the
> > rundbms.opt file looks like this:
>[snip]
> > -qsf.pool_size 2097152
>> -opf.memory 9600000
>>
>> Any help would be much appreciated.
>
>You're not exactly pushing the boat out with the resouce allocations. (I
>bet you've got a DMF cache that can't quite hold the contents of a 3.5"
>floppy disk.)
>I don't have a lot to go on but you could try setting psf.memory to
>something generous, maybe -psf.memory=5000000 to start. I would probably
>also double the connect_sessions and active_sessions just on spec.
In addition to what Roy said, try multiplying that qsf.pool_size
by 4 or 5. The parser builds the final qtree into qsf memory,
so wrongdoing by qsf can look like a psf problem.
I'll also suggest setting opf.active_limit (I think that was
the name) to some small number, like 4 or 5. The little that
I can get out of the error message points to psf, not opf, but
300K per session is really tight.
Other questions: is ACTION a base table or a view? If a base
table, what indexes exist? if a view, how complex? and finally,
have the permits on action changed recently? are there a lot
of grants/permits on that table?
Karl
More information about the Info-Ingres
mailing list