[Info-ingres] Memory Utilisation 98%

[Steve]

> On Thursday, September 10, 2020 at 12:18:04 PM UTC+10, Karl Schendel wrote:
> 
> I'd start by looking at the DMF buffer cache sizes / numbers to make sure that 
> it's not overallocated.

Thanks Karl. Would I do that by setting trace point DM420 that Paul suggested?

> The other step would be to use trace point SC916 
> to print out the current sizes of the various ULM memory pools (these 
> are the ones that grow). You can also use the DBMS log, if enabled, to 
> see when the ULM pools grow.

Am I right to think most/all trace points can be set without needing to restart Ingres?

Also, if the DBMS log is not enabled (i.e. II_DBMS_LOG is not set), am I right in thinking I must specify the log file using the SET TRACE OUTPUT statement? 

Is there any issue using the SET TRACE OUTPUT statement with multiple DBMS servers running? I see, for example, one can use %d and %p when setting II_DBMS_LOG.
 
> Finally, I'd make sure that you're looking at memory actually in use by 
> the DBMS server, and not simply used as a file page cache. 

I am naively looking at the RSS figure returned by prstat -t (installation is running on Solaris).

In wanting to tweak the memory footprint of Ingres, it's not so much that there is a problem with Ingres, it's just that Ingres is the only thing under my control, the server is managed by another outfit.

Thanks
Steve