On Sun, Sep 15, 2002 at 12:55:15PM +0200, Markus W Kilbinger wrote:
> Hi!
> 
> Now it's the third time that one of our NetBSD/i386 machine (running
> 1.6_RCx, now 1.6 release) hangs up while displaying the above message
> in an endless loop.
> 
> NIC is (mac part. anonymous :-)):
> 
>   ti0 at pci0 dev 9 function 0: Netgear GA620 1000BASE-SX Ethernet (rev. 0x01)
>   ti0: interrupting at irq 15
>   ti0: Ethernet address: 00:02:e3:xx:xx:xx
> 
> which is running fine, otherwise.
> 
> I already increased NMBCLUSTERS:
> 
>   options         NMBCLUSTERS=16384
> 
> The machine is a nfs/web server, but these hang ups occurred randomly
> without being correlated to any severe load or network traffic.
> 
> - How to avoid these hangs? Further NMBCLUSTERS increase?

No, it's not an NMBCLUSTERS issue. It's running out of mbuf headers.
Does unplugging the network stop the message stream ?
Can you log in when this happens ? If so, maybe use netstat and vmstat
to try to find what's using that much mbufs.

> 
> - Should these error messages ever appear (== bug)?

There can always be memory shortage. However in your case it looks like
something doesn't release memory ...

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
--