ping: sendto: No buffer space available

[Bill Moran]

jaime at snowmoon.com wrote:
> On Tue, 17 Jun 2003, Bill Moran wrote:
> 
>>>	I think that the NIC is on the logic board.  I can try to install
>>>a PCI card and use that in its place to see if the problem goes away.
>>>Should I bother?
>>
>>I would.  There are two possibilities that I would consider here:
>>a) The NIC has gone flaky with age
>>b) Newer drivers don't talk to that particular NIC as well as the old

Another possibility that bites me in the ass when I'm not looking is
link-level problems.  Occasionally I've had weird issues that were resolved
by replacing a switch or patch cable, or by moving to a different port on
a switch.
As usual ... just throwing ideas at you.

>>Never helped for me either.  You may want to check, but in my experience
>>the output of 'netstat -m' will also tell you that you have plenty of
>>network buffers available.
> 
> 
> bash-2.05b$ netstat -m
> 144/768/26624 mbufs in use (current/peak/max):
>         139 mbufs allocated to data
>         5 mbufs allocated to packet headers
> 138/572/6656 mbuf clusters in use (current/peak/max)
> 1336 Kbytes allocated to network (6% of mb_map in use)
> 0 requests for memory denied
> 0 requests for memory delayed
> 0 calls to protocol drain routines
> 
> 	That was durring normal operation.  The following are at the tail
> end of one of the outages:
> 
> bash-2.05b$ netstat -m
> 477/768/26624 mbufs in use (current/peak/max):
>         386 mbufs allocated to data
>         91 mbufs allocated to packet headers
> 384/572/6656 mbuf clusters in use (current/peak/max)
> 1336 Kbytes allocated to network (6% of mb_map in use)
> 0 requests for memory denied
> 0 requests for memory delayed
> 0 calls to protocol drain routines

<snip additional netstat -m output>

> 144/768/26624 mbufs in use (current/peak/max):
>         139 mbufs allocated to data
>         5 mbufs allocated to packet headers
> 136/572/6656 mbuf clusters in use (current/peak/max)
> 1336 Kbytes allocated to network (6% of mb_map in use)
> 0 requests for memory denied
> 0 requests for memory delayed
> 0 calls to protocol drain routines
> 
> 	It looks like something is causing it to pile up packets in the
> buffers temporarily.  Any thoughts?  In the mean time, I will see if I can
> dig up a PCI ethernet card.

Yes, but it doesn't look like the pile is deep enough that it should have run
out of buffer space.

This one is a bit of a shot in the dark, but try using rndcontrol to increase
the entropy collection.  I'm not sure why I think this might help, but I have
some vague memory of it helping somewhere.

-- 
Bill Moran
Potential Technologies
http://www.potentialtech.com