glance show nbuf kernel parameter at 100%

[Edwin Ruiz_2]

Hi all:

i am using glance to monitor my server HP9000 / n4000 - HP-UX 11i. nbuf parameter is always at 100% used, is that a normal situation? could this situacion become slow my server?

thanks for ur replay

[James R. Ferguson · Accepted solution]

Hi Edwin:

This simply indicates that your buffer cache size (as configured) has reached its maximum. Whethter or not you could achieve better performance cannot be determined from this. Have a look at the discussion of file system buffer parameters in the Tunable Kernal Parameters manual for more information.

http://docs.hp.com/en/TKP-90202/ch05s02.html

Regards!

...JRF...

[baiju_3]

check the below kernel parameter values

dbc_min_pct and dbc_max_pct ,you have to optimise this considering the system memory and Server memeory utilisation .


This parameter will control the amount of buffercasche reserver foir system i/o caching .

Check buffercache useage by galce .


thx,bl.

[Edwin Ruiz_2]

Ok! thanks both.

Best regards

[Kent Ostby]

Edwin --

When setting the max_dbc_pct, you want to keep in mind a few things about it.

In general, you will not get benefit from more than about 800 meg of DBC (at the most).

It's generally recommended to set it at a value of less than 50% of RAM (which is what the default setting of 0 for dbc_max_pct would give you).

Also, if you decide to use dbc_max_pct and dbc_min_pct, you need to set nbuf and bufpages to 0 since a value in either of those will override what you do in dbc_max_pct.

Best regards,

Oz

[Ted Buis]

There is nothing wrong with 100% utilization of nbuf. It doesn't mean that your buffer area is too small. To determine if it is too small, you need to look at the buffer cache hit to miss ratio, increase it and see if it improves significantly, and also decrease it and see if it gets significantly worse. To my thinking, if it isn't 100% then your buffer cache is too large.