vmstat output

[bigscouse]

I am currently investigating a memory problem on a box I support.
We are running Solaris 2.6 and I am using vmstat to gather information.
One coulmn in particular gives a higher reading than we normally get on our other machines, this is the 'de' column.
The only information I can get on this column is that it represents 'anticipated short-term memory shortfal'
Could anybody ellaborate on this?
All help is much appreciated.
[sig]<p>Ged Jones<br><a href=mailto:gedejones@hotmail.com>gedejones@hotmail.com</a><br><a href= > </a><br>Top man[/sig]

 

[OpenSys]

The &quot;de&quot; parameter is a deficit counter for paging. If your output is a non-zero then memory was being used very quickly recently. When this happens any required extra memory will try to be reclaimed, which in turn raises your paging activity outputs and accesses. Which in itself may produce a resource problem on the paging. Not saying that additional wont solve your particular problem but in many cases this action actually adds to the problem causing more paging and subsequent performance degradation. The problem normally arises through short, large memory requirement bursts, so the lotsfree/memfree page scanner cant accurately calculate the timeframes to free memory up, playing with increasing the lotsfree may help, dont make it too big or you waste free memory as it never gets free'd up or altering the fastscan rate will help. Dont expect immediate results, it takes time to get the settings right and you need to watch the vmstat's over a few cycles and peak times to see if you got it right. Best case is when you see the scanner running &quot;little and often&quot; Use vmstat -s to view the scanner running through the memory. a good performance reference for vmstat and others is the Sun Press book : Sun Performance and Tuning. [sig][/sig]

 

[bigscouse]

Thanks for the reply, I hate being left in ignorance.

SUN support eventualy came through with a resolution to the fault. There was a problem with paging that was actualy causing the system to fall over from time to time.
The solution was to add a line to /etc/system to improve paging efficency. [sig]<p>Ged Jones<br><a href= > </a><br>Top man[/sig]

 

[Annihilannic]

Ged,

Was this the [tt]set priority_paging=1[/tt] setting?

Annihilannic.

 

[bigscouse]

I Believe it was, I can't confirm it as I now work for a different company.

Didn't publish it at the time in case people used it inapropriately.

Ged Jones

Top man