[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: High Network Utilization?



Thank you for response to my question.

So based on what you said, I don't need to do anything?

If there is a way to reduce the number of % utilization, I want to do.

This machine will be used as bridged firewall with 4 network segments
attached.

(1 external, 1 internal, 2 DMZs)

I don't know that you read my previous messages.

I was using the box with v3.1 loaded for two month. And it crashed since
then with _pool_get error.

Thank you,


Young.


-----Original Message-----
From: Philipp Buehler [mailto:OpenBSD_(_at_)_fips_(_dot_)_de] 
Sent: Sunday, October 13, 2002 8:01 AM
To: misc_(_at_)_openbsd_(_dot_)_org
Subject: Re: High Network Utilization?


On 12/10/2002, Park, Young K <ypark_(_at_)_ola_(_dot_)_state_(_dot_)_md_(_dot_)_us> wrote To
misc_(_at_)_openbsd_(_dot_)_org:
> Even though I set NMBCLUSTERS = 32768, it still indicates that high 
> network utilization(94%).

*sigh*

> olafw# netstat -m
> 324 mbufs in use:
>         321 mbufs allocated to data
>         1 mbuf allocated to packet headers
>         2 mbufs allocated to socket names and addresses 320/332 mapped 
> pages in use
> 760 Kbytes allocated to network (94% in use)   <----- ????

well, it uses 94%, but that's somewhat misleading. With NMBCLUSTERS you set
'Maxpg' of 'mclpl'. In your case to 16384 pool items, and the kernel uses
320 items in mclpool right now; and this pool has currently a population of
332 items (166 pages).

> olafw# vmstat -m
> Memory resource pool statistics
> Name        Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg
Maxpg
> mbpl         256     3918    0     3573    24     0    24    24     1
inf
> 2
> mclpl       2048     2867    0     2546   166     0   166   166     4
16384
> 5

Regarding the number of Requests and Releases, I'd say, that the machine had
a rather short uptime before these commands have been issued. Furthermore
the count between mbpl and mclpl, and the mbuf type usage 
(321 mbufs allocated to data) leads me to the conclusion, that a certain
application is sending a lot of large network packets (or should receive
it).

Something is wrong, since the clusters dont get freed by either sending them
to the network or getting received from the kernel to the application.

You can find this out by 'netstat -nfinet', look where big numbers for
Send-Q and/or Recv-Q appear.

After all your current pool usage would fit into the default of 2048 for
NMBCLUSTERS, so dont get worried and dont touch stuff, you obviously dont
understand for now.


ciao
-- 
Philipp Buehler, aka fips | sysfive.com GmbH | BOfH | NUCH | <double-p> 

#1: Break the clue barrier!
#2: Already had buzzword confuseritis ?